_ LEW ,
mug».
i Ac.

, .5
KM“. .
3.1.“: 1

”F

'1 . at ‘

dawn. ..S 1.x a. ,
“fine A . ;
he 5.»

an. ;

 

‘ . .799

An}...
:11... .

 

 

$5

to ..

{Juikréfi nflx§_M§.,.fi ”25.35%. a;

 

mam ATE ARIES

LIBRARY L lllllllllll\lllflllllflljlllllllllll

Michigan State 31293
University

 

 

 

 

This is to certify that the

dissertation entitled

The Dynamics of Microenterprises in Jamaica:

An Analysis of Panel Data From 1990-1994
presented by

Todd W. Gustafson

has been accepted towards fulfillment
of the requirements for

PhD degree in W

 

 

Major ofessor

Date 0 q

MS U i: an Affirmatiw Action/Equal Opportunity Institution 0- 12771

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINE return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

use mass-m4

THE DYNAMICS OF MICROENTERPRISES IN JAMAICA:
AN ANALYSIS OF PANEL DATA FROM 1990-1994

by

Todd W. Gustafson

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Economics

1998

ABSTRACT

THE DYNAMICS OF MICROENTERPRISES IN JAMAICA:
AN ANALYSIS OF PANEL DATA FROM 1990-1994

By

Todd W. Gustafson

Microenterprise have been the focal point of much research, policy, and
donor agency attention in recent years. Indeed, the role and presence of
microenterprises in LDC economies is now well documented, if not well
understood. This study is the first to delve into the growth and dynamics of
microenterprises through an econometric analysis of panel data. The data for
this research come from a pioneering, five year data collection effort in the
country of Jamaica.

The prevailing theory of firm growth comes from Jovanovic (1982),
who's learning theory posits an inverse relationship between firm growth and
the firm's size and age. Additional theory and empirical work has identified
other firm, proprietor and environmental factors as important as well.

Two data sets on the same panel of firms are used for analysis. The
first are five year annual data on firm employment, and the second are two
year quarterly data from 1993 and 1994. Both track the same set of firms
over the time period. Using ordinary least squares (OLS) Instrumental

Variable econometric techniques, firm growth is first modeled as a function of

startup size, firm age, and other firm and proprietor characteristics. A
dynamic model is also estimated incorporating lagged employment into the
specification. There is an inverse relationship between firm startup size or
lagged size and growth, but these data find only a marginal relationship
between firm age and growth. Age of the proprietor is inversely related to
firm growth, however. The physical location of the firm (commercial building,
home, road-side) is related to firm growth, as well as the location of the
parish in which the firm operates. The rural or urban location of the firm
does not influence growth. Firms headed by female proprietors grow more
slowly than those headed by their male counterparts in terms of employment.
Technical assistance and access to credit are both only marginally positively
related to firm growth.

Finally, several different model specifications were examined. A cross
sectional model was estimated, and different definitions of the dependent
variable were tested with some strikingly different results across the
estimations. Importantly, a panel data model estimated in levels provided
the most robust and stable results. Although the cost of collecting panel data
may be high, these findings lend support to data collection efforts in other

countries paralleling the Jamaican project.

DEDICATION

This dissertation is dedicated to my wife Gail. Her perseverance,

encouragement, and faith carried this work to its' conclusion.

iv

ACKNOWLEDGEMENTS

In this chapter of my life, a large cast of characters have played
significant roles at various points in time. The names are too numerous to
mention, but my debt and appreciation to each is not diminished. This work
is perhaps but a footnote in their lives, but in this work their contribution is
worked out in each and every line. Specific mention is due, however, to some
key participants. Carl Liedholm, my dissertation chair, encouraged
tirelessly, persevered in all realms of the dissertation process, and rekindled
my vision for this work when the road seemed to close in. John Strauss
raised the level of my technical and econometric contributions, and Don Mead
provided both encouragement and important insights to improve the quality
of the work right to the last revision. This entire committee always believed
in me and the work, without which these words would not be written today.

Support for this project came from the United States Agency for
International Development during data collection, and from the Center for
International Business Education and Research at Michigan State University
during portions of the early write-up and analysis. My thanks to both for
their support. In Jamaica, many thanks are due, but especially to John
Owens at USAID for his valuable insights into Jamaican microenterprise.

At STATIN, Isbeth Bernard and Martin Brown worked tirelessly on

executing the Jamaica Quarterly Panel Survey.

Also, my family's support, prayers, and encouragement carried me
through, and for this I am extremely thankful. My boys Andrew, born in
Jamaica, and Will, born in the midst of Chapter III in Illinois, provided a
constant reminder to work harder and keep going. I thank my wife Gail,
who endured the birth of our first child away from home, other prolonged
absences on my part, and many of the other trials and tribulations
attributable to graduate school. Special thanks as well goes to my parents,
in-laws, and extended family for their continued understanding of my
absences at family gatherings.

Finally, I am especially grateful to Ron Mogavero, who believed,
encouraged, and bet on the completion of this work before the completion of
his hand-crafted boat. He won the bet, but this work is in part a testimony to

his friendship and many promptings.

vi

TABLE OF CONTENTS

LIST OF TABLES: ............................................................................. x
LIST OF FIGURES: ........................................................................... xii
CHAPTER I: INTRODUCTION ......................................................... 1
CHAPTER II: MICROENTERPRISES IN JAMAICA: A REVIEW OF
RESEARCH AND ISSUES ........................................... 9
2.1 Introduction ............................................................ 9
2.2 Microenterprise in the Jamaican Context ........... 11
2.2.1 Economic Environment .......................... 11
2.2.2 Policy Environment ............................... 15
2.2.3 Jamaican Microenterprise:
Past Research ......................................... 19
2.3 Methodology and Background .............................. 22
2.4 Quarterly Dynamics:
Jamaican Microenterprise .................................... 27
2.4.1 Quarterly Change in Employment ........ 27
2.4.2 Quarterly Change in the Wage Bill ....... 37
2.5 Five Year Retrospective on Employment ............ 43
2.6 Conclusions ........................................................... 51

vii

CHAPTER III:

CHAPTER IV:

DYNAMICS OF MICROENTERPRISE IN JAMAICA:
A CROSSECTIONAL ECONOMETRIC
INVESTIGATION OF LONG TERM

EMPLOYMENT GROWTH ......................................... 53
3.1 Introduction .......................................................... 53
3.2 Theory and Hypotheses ........................................ 56

3.2.1 Firm Survival .......................................... 56

3.2.2 Firm Growth ............................................ 58
3.3 The Econometrics of Panel Data .......................... 72
3.4 Data Definitions .................................................... 79

3.5 Employment Growth: A Cross Sectional Analysis
of Jamaican Microenterprise between
1990 and 1994 ........................................................ 81
3.5.1 Introduction ............................................. 81

3.5.2 Five Year Cross Sectional Results:

Log Employment ..................................... 82

3.5.3 Five Year Cross Sectional Results:
Growth in Employment ........................... 94
3.6 Conclusions ......................................................... 102

A PANEL DATA MODEL OF LONG AND
SHORT RUN DYNAMICS FOR
JAMAICAN MICROENTERPRISE .......................... 104

4.1 Introduction ........................................................ 104

4.2 A Panel Data Econometric Approach to Firm
Startup Size and Firm Growth .......................... 106

4.3 Long Term Employment Growth Revisited:

A Panel Data Lagged Dependent
Variable Approach ............................................... 1 1 1

viii

4.3.1 Introduction ........................................... 1 1 1

4.3.2 Five Year Dynamic Employment Model:
Results ................................................... 1 1 2

4.4 Short Term Employment Growth Revisited:
A Panel Data Lagged Dependent Variable

Approach ............................................................. 122

4.4.1 Introduction ........................................... 122

4.4.2 Quarterly Employment Panel Data Model:

Results ................................................... 122

4.5 Attrition Bias ...................................................... 130

4.6 Conclusions ........................................................ 135

CHAPTER V: CONCLUSIONS AND IMPLICATIONS .................. 137
APPENDICES: ......................................................................... 158
REFERENCES: ......................................................................... 181

ix

Table 2.1

Table 2.2

Table 2.3

Table 2.4

Table 2.5

Table 2.6

Table 2.7

Table 3.1

Table 3.2

Table 3.3

Table 3.4

Table 3.5

Table 3.6

LIST OF TABLES

Macroeconomic Summary for Jamaica:
1990-1994 ..................................................................... 14

Distribution of Jamaican Microenterprise in 1990:
By Employment and Industry Type ............................ 21

Summary of Sample by Quarter: Jamaican
Quarterly Panel Survey ............................................... 23

Employment Patterns: Jamaican Quarterly
Panel Survey ................................................................ 28

Year to Year Changes in Employment: Jamaican
Quarterly Panel Survey of Microenterprise ................ 29

Patterns of Change in Firm Wage Bill of Micro-
enterprises: Jamaican Quarterly Panel Survey ........ 38

Levels and Annual Change in Employment: 1990-1994,
Jamaican Microenterprises ......................................... 44

Main Sources of Starting Capital for Jamaican
Microenterprises .......................................................... 69

Cross Sectional Model of Employment: Jamaican
Microenterprise, 1990-1994 ......................................... 84

Complete Cross Sectional Model off Employment:
Jamaican Microenterprise, 1990-1994 ........................ 85

The Relationship Between Firm Size and
Firm Growth ................................................................. 91

Cross Sectional Model of Employment Growth:
1990-1994 ..................................................................... 96

Complete Cross Sectional Model of Employment
Growth: 1990-1994 ...................................................... 97

Table 4.1 Five Year OLS Employment Growth Model ............. 108
Table 4.2 Five Year ln ln Dynamic Employment Model:
Jamaican Microenterprise, 1990-1994 ...................... 114
Table 4.3 Quarterly ln ln Dynamic Employment Model:
Jamaican Microenterprise, 1993-1994 ...................... 124
Table 4.4 Evaluation of Attrition Bias in a Panel Data Model of
Employment Growth: in Levels for Jamaican
Microenterprise .......................................................... 133
Table 4.5 Evaluation of Attrition Bias in a Panel Data Model of
Employment Growth: ln ln form for Jamaican
Microenterprise .......................................................... 134
Appendix Tables
Table A.1 Cross Sectional Model: 1990-1994, Firm age
Interactions Included ................................................. 159
Table A2 Complete Cross Sectional Model: 1990-1994, Firm age
Interactions Included ................................................. 160
Table B.1 Five Year ln ln Employment Model: Jamaican
Microenterprise Panel Data, OLS and OLS IV
Results, 1993-1994 ..................................................... 162
Table 0.1 Quarterly Employment Model: OLS and

OLS IV Results, 1993-1994 ....................................... 167

xi

Figure 2.1
Figure 2.2
Figure 2.3
Figure 2.4
Figure 2.5
Figure 2.6
Figure 2.7
Figure 2.8
Figure 2.9
Figure 2.10
Figure 2.11

p
Figure 2.12

Figure 2.13
Figure 2.14
Figure 2.15

Figure 2.16

LIST OF FIGURES

Some Characteristics of the Sample ........................... 24
Index of Change in Total Employment ....................... 30
Change in Employment by Gender ............................. 32
Change in Employment by Sector ............................... 33
Change in Employment by Location ........................... 34
Change in Employment by Establishment Size ......... 35
Change in the Wage Bill: Real vs. Nominal ............... 38
Change in the Real Wage Bill by Gender ................... 40
Change in the Real Wage Bill by Sector ..................... 41
Change in the Real Wage Bill by Location ................. 42
Change in the Real Wage Bill by Size ......................... 43
Five Year Retrospective on Employment:

Jamaican Microenterprise ........................................... 46
Five Year Index Employment by Gender .................... 47
Five Year Index Employment by Sector ..................... 48
Five Year Index of Employment by Location .............. 49

Five Year Index of Employment by Size ..................... 50

xii

CHAPTER I

INTRODUCTION

The growth of microenterprises “may offer the greatest potential for
reaching the poorest of the poor by creating jobs and for generating the
greatest developmental impact by transforming marginal enterprises into
sustainable businesses (Boomgard, 1989)”. James Boomgard’s comment
reflects the under-girding motivation behind research targeting
microenterprises. Certainly the sector of its own accord — fragmented,
unaccounted for in official statistics, categorically unglamorous — doesn’t cry
for attention or demand an audience. The World Development Report in
1990, however, reports that more than 1 billion people in the developing
world are living in poverty. That is, over one billion people are struggling to
live on less than $370 per annum. The research agenda, hence, is in response
to this overwhelming global need to understand and formulate solutions to an
age old problem, which should be just that: an old (sic) problem.

Small scale industry has attracted a lot of attention in recent years, as
new empirical research has revealed the robustness of the sector even amidst
overall economic stagnation. Perhaps, as in 1964 with T.W. Schultz’s article
on agriculture, a realignment of effort will take place to capture - or set loose

- some of this robustness to contribute to economic development and the

structural transformation. In the context of development, the “structural
transformation” refers to the transformation of the structure of demand,
trade, production and employment, and is generally manifested in three
principal areas: industrialization, agricultural transformation, and migration
and urbanization (Chenery, 1989). The characteristics of the transformation
include increases in rates of accumulation, shifts in sectoral composition,
changes in income distribution and the demographic transition. The “core” of
the transformation, from Syrquin’s (1989) point of view, is manifested in the
accumulation of physical and human capital, and the shifts in the
composition of demand, production, and employment. The above comments
implicitly reflect a dynamic process of change. Thus to assess how specific
policies aid or hinder the structural transformation requires a probing look
beyond economic growth to the very core of what the transformation is all
about.

The research agenda for microenterprise grew out of such a context.
As Liedholm and Chuta report (1976), an interest in small scale firms grew
out of the disappointing results of the import substitution policies of the
1950’s and 1960’s. By the mid- to late-1970’s, the need for information and
insight into the small and microenterprise sector began to be articulated.
The interest in small scale firms was not limited to the developing context
either. In 1979, Birch’s seminal paper was published, in which he stated that

“whatever else they are doing large firms are no longer the major providers of

new jobs for Americans.” A source for much debate in the ensuing years, his
thesis proposed that most new jobs emanated from small firms. Acs and
Audretsch (1986) in fact report that in the past 20 years, the majority of new
US. employment did come from small enterprises. The race to understand
the contribution of these small and micro firms was underway.

One of the most significant early and continuing challenges to
understanding small scale and microenterprises is the paucity of data
available on these firms. As alluded to above, these firms typically slip
through the official data collection of governments or private industry
tracking performance. As several authors point out, one of the main reasons
why micro firms have received so little attention: no data! (McPherson, 1992;
Cabral, 1994; Mata, 1994; Reid, 1995) The research that has been done,
however, has focused on several very significant issues, of which a few will be
briefly reviewed here.

The most debated topic has been the validation of Gibrat’s Law (1931),
purporting the independence of firm size to growth. Several early
investigations found evidence supporting Gibrat's Law (Hymer and
Pashighian, 1962; Prais, 1956; Singh and Wittington, 1975; and Wagner,
1992). This included Mansfield's (1962) influential study, with a convincing
argument that small firms that are more prone to die could bias results in
favor of a negative relationship between firm size and growth. The debate

rigorously continued into the late 1980’s and early 1990’s, however, as new

evidence mounted refuting Gibrat’s law (most recently, Evans, 1987; Hall,
1987; Audretsch, 1992; McPherson, 1992; Mata, 1994, Reid, 1995). A
multiplicity of theoretical perspectives have been provided to account for this
and provide alternatives (see Vining, 1976; Nelson and Winter, 1978;
Jovanovic, 1982;).

In the late 1980’s, a fresh interest in the factors influencing firm entry,
exit or survival of the firm evolved. To explore these factors, hazard or
duration analysis was adopted from biostatistics and applied to data in the
United States and Europe (see Dunne and Hughes, 1994; Audretsch and
Mahmood, 1991; Dunne, Roberts, and Samuelson, 1989;) and to a lesser
degree to data from developing countries (Cabal, 1994; McPherson, 1992;
Behrman and Deolalikar, 1989). Certainly all the issues surrounding firm
survival have yet to be resolved.

Apart from these main avenues of research, several equally important
topics have received attention as well. A sampling of these include the
question of small firm efficiency, the role of innovation in small firm growth,
the entrepreneurial influence of the proprietor, data collection methodologies
and definitional issues. However, as McPherson (1992) points out, much of
the past work on small- and microenterprises has focused on static concerns
and side stepped the very important understanding of the dynamic nature of

these firms: how do the survivors grow and change over time?

The need to understand the firm growth of survivors has been clearly
articulated by many different authors. In 1956, Prais recognized that a
significant amount of effort needed to be put against understanding the role
of surviving small scale firms. Aislabie (1992) reported that there are few
stylized facts outlining firm growth. Post entry performance, according to
Mata (1994) and Reid (1995), have not been studied enough. Brock and
Evans (1989) suggested that the most important research for policy direction
is against the puzzle of the growth of small scale firms. Liedholm and Mead
(1987) identified firm dynamics as one of the main items on the research
agenda for small and microenterprises in developing countries.

Some limited research has been done on these issues for survivors with
US. data (Hall, 1987 ; Evans, 1987) and European data (Mata, 1994; Wagner,
1994, or Dobson and Gerrard, 1989), for example. In developing countries,
some of the more recent work includes McPherson (1992) and Cabral (1994).
However, the lack of panel data in developing countries has largely curtailed
the ability to address these issues effectively. In order to understand firm
dynamics adequately, panel data is required (McPherson, 1992; Evans, 1987),
and in the developing world, panel data sets have been almost non-existent
(Grootaert and Kanbur, 1995). Consequently, much recent research has
unwillingly skirted the issue of thoroughly understanding determinants of
growth and profitability for surviving small businesses. What characterizes

the growth pattern of these firms? Which factors influence this growth

pattern over time? What are the determinants of the different growth rates
among survivors? How does a firm become an incumbent? These issues are
especially salient given the additional obstacles to growth faced by micro
firms (Bregman, et al, 1995). Answers to questions such as these have direct
policy implications, and as Audretsch and Acs (1994) suggest, the most
important research to be conducted at this time relates to what happens to
firms after startup.

The purpose of this dissertation is to extend the learning on firm
dynamics. To accomplish this, simple measures of firm performance were
collected annually on a panel of microenterprises in Jamaica over a five year
period. In the final two years of enumeration, the firms were visited
quarterly,l tracking firm employment, demographic and locational factors,
and some limited flow information on firm sales or output. This represents
the first extensive panel data on microenterprise covering retail sales, output,
and employment, for Jamaica and the developing world.2 As such, this
research offers a more comprehensive picture of firm dynamics than
previously available. This dissertation will focus on the employment
dimension of firm growth. Further, Instrumental Variable and fixed effect
econometric techniques are used for the first time to unpack the growth

dynamic issues in a developing country. Finally, short and long run data are

 

1 A fairly complex issue relates to the definition of formal versus informal
enterprise. For the purpose of this research, firm size is the only qualifying
element. Any firm between 1 — 10 persons are included.

2 Mata (1995) analyszed a panel dataset on firm growth in Portugal.

available on the same subset of firms, providing policy makers and other
researchers an important window into the connection between the two.

This dissertation has two objectives. The first is to understand firm
dynamics more completely. Uniquely in this work, this is accomplished
through both a five year annual and two year quarterly analysis of
microenterprise employment. Some of the questions asked include, what
factors and firm characteristics influence a firms short and long run growth?
What is the relationship between firm size, age and growth? Do technical
assistance or access to credit affect growth? Although these questions have
been addressed in previous research for other countries, these data provide a
unique vantage point from which to address these issues. Second, how does
panel data contribute to this understanding of firm dynamics? Is panel data
a necessary element to an accurate characterization of firm growth? As such,
these results will be compared to much of the earlier research on firm growth
across both the developing and developed world. This final task in itself is
identified by Brock and Evans (1989) as one of the key research agendas of
the day.

Three different lines of analysis are pursued. In Chapter II, the
Jamaican context is described, and a stylized, dynamic descriptive picture of
the Jamaican microenterprise is presented. Specifically, the secular trends in
employment and wages are examined on a quarterly basis and annual

employment is examined over the five-year period. In both contexts, the

unique value of panel data is utilized to write the story. Chapter III
examines long run employment growth through a traditional cross sectional
OLS analysis. This approach enables direct comparison to past research as
well as the ability to understand the stability of the insights over multiple
years. Chapter IV extends the analysis of long and short-run employment
growth by introducing a panel data model and utilizing different econometric
techniques, such as Instrumental Variables, to control for measurement
error. Several other potential sources of estimation bias are addressed as

well. Conclusions and policy implications follow in Chapter V.

CHAPTER II
MICROENTERPRISES IN JAMAICA:
A REVIEW OF RESEARCH AND ISSUES

2.1 Introduction

The data for this dissertation come from a long term study of
microenterprise in Jamaica, combining several data collection exercises over
a five year period. Specifically, three elements are combined. The first are
the results from a national census of microenterprises in 1990 executed by
STATIN, the Statistical Institute of Jamaica. This census visited 20% of
microenterprises in Jamaica. The second set comes from a 1992 National
Survey, again executed by STATIN, interviewing approximately 2,400 firms
from the 1990 frame. The third is a Quarterly Panel Survey (QPS), a truly
pioneering effort to enumerate a subset of the microenterprises visited in
1992. This panel was conducted in Jamaica from the second quarter of 1993
through the fourth. quarter of 1994. Supported by funds from the
Government of the Netherlands, USAID, and the Office of the Primer
Minister and executed by STATIN (The Statistical Institute of Jamaica), this
survey was designed to trace seasonal or quarterly changes in the level and
patterns of activity of a panel of existing microenterprises in that country.
Because the quarterly data collection effort extended beyond one year, the

survey also aimed to provide some perspectives on the longer term growth

patterns of these microenterprises. As all three research projects were
connected, the linked data between the years generates key information on a
single panel of firms over a full five year period. Separately and combined,
this represents a significant analysis opportunity on microenterprise.

This Jamaican QPS of microenterprises was unique and represented a
truly pioneering effort. Although enterprise panel surveys have been
conducted in Jamaica and elsewhere for many years, they traditionally cover
only the larger firms that keep good books and records. Microenterprises
have never been included in such surveys. Most of these enterprises have
never been included in such surveys. Most of these enterprises do not keep
books and information on them is scanty at best, difficult to obtain, and
typically paints a partial picture of the firm at one point in time. The
Jamaican effort thus represented the first attempt to generate dynamic data
on key activity variables - employment, wages, and sales (output) - from the
same group of microenterprises repeatedly every quarter. Since the firms
chosen for the panel were scientifically drawn from a representative national
census, the findings have national significance.

By providing such information, policymakers, donors, and those
involved with microenterprises are alerted not only to the importance of
these enterprises, but how they are changing. Are existing microenterprises
expanding or contracting overall? Are they employing more or less workers

in particular quarters? How are trading firms doing relative to those in

10

manufacturing? Answers to these and other important questions will be
pursued in this chapter.

In this chapter, a brief overview of the Jamaican context will be given.
Following this, Section 2.3 reviews the data collection methodology. This is
followed in section 2.4 and 2.5 with the results from the survey, both for the

quarterly and five year time frame. Some concluding remarks will follow.

2.2 The Jamaican Context

2.2.1 Economic and Policy Environment

The island of Jamaica sits just south of Cuba on the western side of the
Caribbean. In the late 20th century, the island is known best for its’ beautiful
beaches, warm Caribbean waters, and natural beauty. Jamaica covers only
10,991 square kilometers in the shape of an oval, spanning 150 miles long to
as little as 40 miles wide. The terrain varies significantly in this compact
environment. Dramatic cliffs line the coast on the eastern side of the island,
while the majority of the remainder enjoys beautiful sand beaches. In the
east as well is the Blue Mountain range, with the highest peak exceeding
7,000 feet above sea level. The north side of the island experiences
significant amounts of rain fall, while the south, in the rain shadow, contain
pockets of desert flora. The island is volcanic and fairly mountainous

throughout.

ll

Jamaica is also endowed with many natural resources. Best known, of
course, is the natural beauty, attracting over 1.5 million visitors each year.
The island boasts large deposits of bauxite, used in the production of
aluminum. The climate and soil are conducive to the production of several
cash crops, such as sugar cane, bananas, and coffee. Jamaican Blue
Mountain coffee is considered one of the best coffees in the world, often
demanding the highest per pound coffee price on the international market.

The population of Jamaica is approximately 2.5 million, with a growth
rate around one percent per annum and a population density of 234.2 per
square kilometer. Roughly 31% of the current population is under 15, and the
life expectancy at birth is 73.8 years. The majority of the Jamaican
population is of African or Afro-European decent (90%), with the remainder of
East Indian, Chinese and British or American decent (STATIN, 1993). A
high number of J amaican’s migrate annually to other countries, and an equal
number of native Jamaican’s live in Canada, Great Britian and the United
States as currently populate the island. Similar to other islands in the
Caribbean, the origins of the current populace of Jamaica is mostly from
abroad, as the Spanish and British eradicated the indigenous Arawak Indian
population in the 1600’s.

The current economic reality of Jamaica, however, is one of slow
growth and stagnation, rising prices, and high unemployment. To put this in

context, a brief foray is necessary into the policy climate between 1962 and

12

the present (see below for more detailed comments). Following the peaceful
transition to independence from the British in 1962, the Jamaican economy
grew at a robust rate of 5-7%. Between 1969 and 1973, for example, the
average growth rate was 6.1% (Davies and Witter, 1989). In this early
period, the government encouraged foreign investment and shifted away from
the colonial style economy of the production of primary goods for export. In
the 1970’s, however, Jamaica sank into a period of stagnation and decline,
principally fueled by a shift in governmental policy towards redistribution of
income and the introduction of a welfare state. As Jamaica also imports 98%
of its’ energy needs, the oil crisis in 1972 contributed to this spiral as well. In
combination, the country descended into a period of economic crisis . After a
heated and violent election, the 1970’s regime gave way in 1980 to a
government focused on exports and high rates of private investment. In the
late 1980’s, the government changed hands again, but the policy climate
remained consistent, with an emphasis on private investment, exports, and a
de-emphasis on governmental intervention. With all of this change and
shifting ground, however, growth in the late 1980’s and early 1990’s has
hovered around zero percent, with unemployment rates above 15% and
inflation around 25% (World Bank estimates).

One of the key policy developments in this analysis period has been the
liberalization of the exchange rate. In the 1970’s, the Jamaican dollar was

pegged to the US. dollar (previously the British Sterling). As the

13

government set the rate, the real market value of the Jamaican dollar
quickly fell out of equilibrium. An active black market for Jamaican dollars
developed (Grosse, 1994). In September of 1991, the Government of Jamaica
liberalised the dollar, causing a dramatic jump in the exchange rate fi'om 7:1
to 25:1, soaring inflation rates (80% in 1991), and the disappearance of the
black market. This change in exchange rate policy resulted in a dramatic
shock to the Jamaican economy.

Overall, the economic context between 1990 and 1994 was one of
stagnation, although conditions did not worsen significantly. In 1991 with
the liberalisation of the exchange rate, the inflation rate hit 81%, but during
the rest of the period, inflation declined to 25% and the unemployment rate

fluctuated slightly around 15%. Table 2.1 summarizes the overall conditions.

 

 

 

 

 

 

Table 2.1
Macro Economic Summary for Jamaica:
1990 — 1994
Economic Year
Measure 1990 1991 1992 1993 1994
1) OffiCial 8.031 21.521 22.211 32.521 33.2:1
Exchange Rate (J$: 113$)
2) Unemploy- 15.7% 15.7% 15.7% 16.3% 15.4%
ment Rate
3) % Change in 29.8 80.2 40.2 30.1 26.8
Consumer
Prices
4) “/0 Real GDP 4.5% -0.3% 0.5% 0.4% -0.3%
Growth per
Capita

 

 

 

 

 

 

Source 1, 3-4: IDB estimates with data from the IMF
Source 2: Programa Regional del Empleo para America Latina y El Caribe

l4

 

2.2.2 The Political Climate and Microenterprise

Modern Jamaica can be traced back to the formation of the two key
political parties in the 1930’s and early 1940’s. The JLP, or Jamaican Labor
Party, has traditionally been the more conservative, with an emphasis on
social and economic stability rather than change. The PNP, or People’s
National Party, has been more left of center, focused on redistribution of
income and interested in sweeping change (Stone, 1989). These two parties
have traded leadership roles in Jamaica’s modern history, often resulting in
dichotomous policy from one elected administration to the next.

Jamaica gained independence from it’s colonial power, Great Britian,
on August 6 of 1962. “Out of many, One people,” the adopted national motto
of Jamaica, reflects an ideal of social cohesion, political unity, and peaceful
socio-political interaction as the goal of independence. Yet since Jamaica’s
independence, the path has been marred by lack of identity, political and
social violence, and continued polity in public policy. The violent elections of
1967 and 1980; the Rodney riots in 1968; student unrest in the 1970’s; the
development of the urban ghetto’s serving as a flashpoint of unrest in the
1970’s and 80’s; all underscore the tensions present in the social, political and
economic fabric of the country.

In broad stroke, public policy has directly reflected the reigning
ideology of the reigning political power. The overarching paradigm, however,

can best be described as an “inward looking policy,” which has plunged the

15

country into stark economic times (Balassa, 1989). Policy objectives
promoted the expansion of an export industry through tax holidays, raw
material import holidays, and market monopolies. This policy tract has lead
to slow growth in manufacturing while simultaneously leading to a bias
against small scale industry (Fisseha, 1982). Growth of small scale firms
flourished nonetheless, leading to a statement found in the five year plan
(1978-1982) to pay “serious attention to small scale manufacturing.”

From 1962 to 1970, the JLP party was in office, implementing
economic plans to encourage foreign investment for manufacturing and
encourage private investment. As alluded to above, the policies reflected the
accepted structural transformation “export” policies prevalent at the time. In
the 1970’s the PNP came into power, shifting the emphasis towards income
redistribution, welfare reform, and a more active role of government in
business and development. The country fell in to a period of economic
decline, however, with real GDP falling by 13% between 1975 and 1980
(Fisseha, 1982). This reflects in part the effect of the global recession in the
period and wayward economic policies. After this period of deteriorating
economic conditions in the late 1970’s, the JLP again came to power after a
violent election in 1980, implementing an even more conservative version of
their policy of the 1960’s.

In 1989, the PNP once again came to power. Their economic policy

platform, however, coincided in broad terns with the JLP platform. Both

16

parties advocated a diminished role of government in business, the
encouragement of private enterprise as a means to economic development,
and liberalization of the Jamaican dollar. The adoption of these measures
reflects in part the influence of IMF and World Bank policy directives for
Jamaica. The differences in economic and social welfare reform between the
two parties appears to have converged. Even with an increasing degree of
similarity in ideology between the parties, the change in political parties
continues, however, to define the changes in the economic landscape
(Anderson and Witter, 1991).

During the period 1990 — 1994, the period of this analysis, the PNP
party held power in Jamaica. As alluded to above, the policy platform of the
PNP reflected a new ideology, with a shift towards smaller government, and
an emphasis on private enterprise, trade liberalization, and currency reform.
Significantly in this period, specific programs were put in place which
focused on microenterprises.

The Government of Jamaica (GOJ) participated in several initiatives
to support microenterprises, beginning with MIDA in 1991. MIDA, the Micro
Investment Development Agency, was formed to assist microenterprises
through technical training and credit support. In 1992, for example, MIDA
funneled 40 million Jamaican dollars to microenterprises through more than

20 lending agencies. The funds supported agricultural projects as well.

17

Other programs include CAST, an entrepreneurial training center
establish in 1986, training over 400 micro entrepreneurs in 1992. This
program is currently co-sponsored by the GOJ/GON Micro Enterprise Project,
a two year project aimed at providing financial services to the sector, increase
the fact base on micro firms to help guide policy decision, provide training,
support women and in general raise the status of micro enterprise activities.
Finally, JAMPRO Entrepreneurial Centers provide a wide range of non-
financial support to micro enterprises. This quasi-governmental organization
assisted 99 businesses expand/start in 1992, with a projected employment
impact of 732.

In summary, the years between 1990 and 1994 contain policies
liberalizing the macro economic context, with a continued emphasis on
private enterprise as a cornerstone of economic growth. Further, several new
programs were put in place which specifically addressed the needs of the

microenterprise.

18

2.2.3 Jamaican Microenterprise: Past Research

If research on microenterprise and small business in LDC’s have
suffered from a paucity of good data, then Jamaica has the good fortune of
generating above average attention over the past 15 years. Several research
and data collection programs have been executed. To begin with, in 1978 a
national census was conducted which developed a sample frame of
microenterprises. This sample frame was used for subsequent research by
Yacob Fisseha. This line of research explored management characteristics of
the microentrepreneurs and the constraints facing these business people on a
daily basis. In 1983, the World Bank sponsored the Small Establishment
Survey. This survey sub-sampled from the same frame, and collected
rudimentary information on approximately 2,000 micro firms. Seven years
lapsed, and then in 1990, STATIN again conducted a national census,
collecting basic information on microenterprise business activity and
employment. From this frame, a National Survey was conducted in 1992.
This survey, administered to 2,400 micro firms, asked detailed questions on
constraints to firm growth, historical questions on firm formation, and sought
some basic information on employment. As will be detailed below, this
current research is connected to both the 1990 census and 1992 national
survey. Finally, in 1993, a dynamic study by Yacob Fisseha tracked 142

firms included in his original sample in 1980.

19

In 1978, the micro sector was estimated at roughly 38,000 firms,
providing a livelihood for some 80,000 individuals. By 1990, this number is
estimated to have jumped to 88,000 firms employing over 150,000 (STATIN,
1992), almost tripling the number of firms and doubling the employment.
These are sizeable numbers for a country of only 2.3 million. Liedholm and
Mead (1987) estimate, for example, that small scale manufacturing
represents roughly 22% of total manufacturing GDP, or roughly 3.5% of total
GDP. More significantly, 74% of all industrial employment is carried by
these firms. Finally, about one-eight of the total Jamaican population are
“fully or substantially” supported by these industries. Thus without even
considering the linkages between this sector and agriculture or with larger
firms, the microenterprise sector plays a significant role in the economic life
of the country.

How can the microenterprise sector in Jamaica be characterized? To
begin with, the sector is somewhat lopsided, with 52% of the firms in the
sector own account and roughly 66% of the firms defined as wholesale or
retail trade (see Table 2.1). By gender, the micro sector is split almost in
half, with a slightly higher percentage of women working in micro firms.
Finally, more firms were located in urban areas, and these firms enjoyed a
slightly higher average employment than their rural counterparts. Average

employment across all of these firms averaged 1.7%.3

 

3 Summary statistics from STATIN, 1992.

20

Table 2.2
Distribution of Jamaican Microenterprises in 1990:
By Employment and Industry Type

 

 

 

 

 

 

 

 

 

 

Business Category Percent of Industry Group Percent of
Businesses Businesses
Own-Account 73.4% Manufacturing 7.4%
Small (1-4 21.8% Construction 0.6%
employees)
Small (5-9 4.8% Wholesale, Retail 66.4%
employees) Trade and Hotels
- - Transport and 3.3%
Communication
- - Finance, 1.9%
Insurance and
Real Estate
- - Personal Services 20.4%
Total 100.0% 100.0%

 

 

 

SOURCE: STATIN, 1990 National Census

Other than the distributional characteristics of these firms, several
other significant dimensions have been revealed through the recent research.
To begin with, Jamaican firms utilize little family labor (F isseha, 1982),
which accounts for the high percentage (two thirds) of micro firms which
closed between 1980 and 1992 due to death, retirement or migration of the
proprietor (Fisseha, 1993). No family labor was in place to take up the reigns
and carry on. Fisseha also described the surviving firms as tenacious,
surviving the turmoil of the 1980’s to survive into the 1990’s, and points out
that roughly 40% of the firms sampled rely on their business as a primary

source of income.

21

 

Anderson's 1994 study reveals that capital shortages and poor market
demand are the two most prevalent problems facing micro-enterprises in
Jamaica today. Very few firms access formal credit at startup (6.1%), with an
even smaller percentage obtaining working capital. The majority did not
apply for loans, due to high interest rates, access to family money, or a
percieved lack of loan collateral.

In sum, the microenterprise sector in Jamaica plays a major role in the
local and national economy. Policy directives have not been keenly focused
on the sector, in part due to a lack of relevnat data to guide the decision
makers (Anderson, 1994). One of the purposes of this research is to shed

more light on the microenterprise landscape leading to better policy.

2.3 Methodology and Background

The motivation for this research resulted in a Quarterly Panel Survey
(QPS) of Jamaican microenterprises, one of a series of three linked studies
exploring the microenterprise sector in Jamaica. The data for this work
draws upon all three. In 1990, STATIN conducted a nationwide survey
visiting a random sample of 20% of the country's enumeration areas. In each
of the locations selected in the sample, STATIN conducted a complete
enumeration of all microenterprises, collecting information on the business

type and employment in each enterprise. In 1992, STATIN again conducted

22

fieldwork, under the direction of the University of the West Indies, on this
occasion collecting more detailed information on a random sample of 2,394
enterprises chosen from the original 1990 sample of 16,000 firms. For these
firms, more detailed background questions were asked, including information
on the educational and training experience of the entrepreneur, access to
credit, employment in the firm as well as its output. From this frame of 2,394
firms, 700 firms were randomly selected, and comprise the sample for the
QPS. Table 2.3 details the sample.
Table 2.3

Summary of Sample by Quarter
Jamaican Quarterly Panel Survey

 

 

 

 

 

 

 

1993 1994
Quarter Quarter Quarter Quarter Quarter Quarter Quarter

II III IV I II III IV
Firms in the sample 361 391 292 334 281 313 255
this quarter (F)
Firms active this 361 363 275 322 274 305 243
quarter
(D)
Temporarily closed na 28 17 12 7 8 12
this quarter
(E)
New Firms added na 133 75 52 27 10 9
this quarter (A)
Firms reentering na na 27 89 51 89 46

this quarter from an
earlier period (B)

 

 

Number of firms in na 258 190 193 203 214 200
the sample this
quarter that were
also in previous
quarter ' (C)
Firms temporarily na 11 8 6 3 5 1
closed that were in
previous quarter
(subset of above)

 

 

 

 

 

 

 

 

1 Firms used for analysis of quarter to quarter change.
Note 1: RowT=A+B+C
Note2: RowT=D+E

23

 

The data in Table 2.3 reflect some of the birth pains of the QPS, with
its evolving understanding of data collection difficulties, as well as the fluid
nature of the microenterprise sector in Jamaica.4 The second round, for
instance, shows a high number of entrants coming into the sample, due in
part to the initial difficulties encountered in locating these firms during the
first round of data collection. Some of these characteristics of the sample are

portrayed in Figure 2.1.

Figure 2.1 Some Characteristics of the Sample

 

 

Qtl’ ll Qt! Ill Qtr IV Qtrl Qtrll Qtr Ill Qt! IV

Col l: Total Sample

Col 2: Firms Active in this Quarter and in Previous Quarter
Col 3: Temporarily Closed

Col 4: Out of Business

Many of the firms moved in and out of the sample over the period
covered by this report. Enterprises dropped out of the dataset in any quarter

for one of four reasons: 1) enterprises were temporarily closed that quarter; 2)

 

4 700 firms were originally sampled, but due to data collection difficulties
mentioned in the text, the active sample in any one quarter was roughly half
that. These 700 firms can be accounted for by adding 361, the total number
of firms active in Q1, with the sum of row A. This total of 667 firms is the

24

enterprises refused to answer the questionnaire that quarter; 3) enterprises
were known to have permanently closed; and 4) those with no data, for other
or unknown causes. The bulk of the firms with no data fall into the latter
category.5

A particularly thorny problem plaguing this sample is the large
number of firms that drop out of the sample for a quarter with no information
on their status (refusal, temporarily closed, etc) during their absence. This
is a particular problem in quarter IV where almost 40% of the sampled firms
fell into this category.6 The percentage of firms that fall into this category,
however, declined from quarter to quarter, drOpping to roughly 14% by the
final quarter of 1994. During the earlier quarters, some firms were not
located due to relocation or closures, whereas during the latter periods firms
were more likely to be excluded as the proprietor was not available for the
interview on the day(s) of the intended enumeration. This difficulty
ultimately decreases the analyzable sample size and may introduce some

degree of selectivity bias into the analysis.

 

active sample throughout data collection. The difference between roughly

700 and 667 is due to refusals, closures, and unable to locate the business.

5 The "unknown" firms in turn fall into three categories: 1) unknown, with no
information available from the field; 2) unable to locate the business or proprietor
in the given period; and 3) contact made, but not able to complete an interview in
the required time period.

5 Due to a period of bad rains with much flooding in the rural areas, many
firms were not visited during this particular quarter.

25

A number of firms also closed their doors temporarily due to slow
business conditions. The number of enterprises temporarily closed in this
way reached over 7% of the enterprises with valid data in the third quarter.
It is likely that a number of the “unknown” firms also fall in this category.
Particularly noteworthy as well is the constant percentage of firms that fell
into this “temporarily closed” category on an on going basis. This seasonal
pattern of business activity suggests an underlying weakness in the economy,
with firms having to shut down operations for months at a time. These
seasonal fluctuations can be particularly stressful for microenterprises, as
multiple family members often draw their income from one microenterprise.

The main focus of analysis in the section is on quarter-to—quarter
changes among the sample firms. This analysis makes use of information for
all enterprises with the relevant data for any two successive quarters. Table
2.3 shows, for example, that there were 258 enterprises that provided
responses in both quarter II and quarter III. The analysis in the following
sections is based on patterns of changes in such pairs of data for individual
enterprises from one quarter to the next. Also note that the analysis period

begins with quarter II, 1993. 7

 

7 This reflects the decision to link the reference weeks, for which the
respondents provided the relevant weekly sales (output) and wage bill data,
to the quarter in which that reference week occurred. Originally, it was
envisioned that the reference week would have been in the last week of the
previous quarter, but timing delays and limitations in the accuracy of the
respondents memory recall necessitated that the reference week would fall in
the current quarter. Thus, the data collected for the first complete panel

26

2.4 Quarterly Dynamics: Jamaican Microenterprise

2.4.1 Quarterly Change in Employment

Employment is the indicator typically used to measure change in the
level of economic activity of existing firms. It is most easily and accurately
remembered by the entrepreneur and also does not need to be deflated.
Employment is thus the first indicator that will be examined in the report.
The term "employment" includes working proprietors and unpaid family
members as well as paid employees, trainees and apprentices; paid
employment is also analyzed and discussed separately.

The employment data for Jamaican microenterprise for all seven
quarters are presented in Table 2.4. The data, as in all sections of the report,
are presented as indices, with the second quarter of 1993 set equal to 100 in
each case. The percentage change between the current and previous quarter
is noted below each index.8 The first two rows of the table report on total
employment and total paid employment, and then total employment is

further broken down by gender, sector, location, and establishment size.9

 

period (April, 1993) ultimately related to quarter II rather than quarter I as
originally planned. Moreover, the monthly and quarterly figures proved to be
unreliable, requiring the use of the weekly data.

8 For the disaggregated data (gender, sector, etc), each index has been linked back

to employment in quarter 2 of 1993.

9 In examining quarter to quarter change, all firms which provided valid data
in both periods were included in the calculations. “Valid data” includes firms
reporting positive business activity as well as those temporarily closed.

27

Table 2.4
Employment Patterns
Jamaican Quarterly Panel Survey

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1993 1994
Quarter Quarter Quarter Quarter Quarter Quarter Quarter
11 III IV I II III IV
Total employment 100 92.7 90.2 93.2 87.5 85.9 81.4
I (7.3%) (2.8%) (+33%) (6.1%) (1.8%) (5.2%)
Paid employment 100 114.3 111.6 109.0 117.5 117.0 105.0
@4304) (2.4%) (2.3%) (+78%) (0.4%) (10.3%)
Total employment, by gender of owner
Male-owned 100 91.1 92.0 91.3 87.7 92.5 85.8
enterprises (8.9%) (+10%) (8%) (3.9%) (+55%) (2.4%)
Female-owned II 100 95.2 86.4 95.6 86.4 78.3 76.4
enterprises (48%) (9.2%) (+10.6%) (9.6%) (9.4%) (2.4%)
Total employment1 by sector
Manufacturing 100 99.3 98.8 96.5 92.0 91.0 80.8
(7%) (1.2%) (1.1%) (4.7%) (1.1%) (11.2%)
Trade and Commerce 100 89.5 84.2 88.6 82.5 80.0 77.5
(10.5%) (6.0%) (53%) (6.9%) (3.0%) (3.1%)
Services, Transport, 100 104.4 107.7 108.8 102.9 104.1 96.7
and Finance (+45%) (+32%) (+10%) (-5.4%) (+12%) (-7.l%)
Total employment, by location
Urban 100 96.1 98.5 97.1 89.7 89.1 86.1
(3.9%) (+25%) (1.4%) (7.6%) (0.7%) (3.3%)
Rural I 100 91.0 84.7 88.3 83.8 81.9 75.1
(-9.0%) (-6.9%) (+43%) (-5.1%) (2.3%) (-8.3%)
Total employment, by size
Own Account 100 111.0 116.3 133.4 125.4 125.5 118.1
(+1 1.0%) (+4.8%) (14.7%) (-6.0%) (0.1%) (5.9%)
1-4 Employees 100 88.3 85.7 88.4 83.2 82.3 78.3
(-11.7%) (-3.0%) (+32%) (-5.9%) (1.0%) (-4.9%)
5-9 Employees 100 86.3 76.5 67.9% 63.0% 60.6% 55.6%
(13.7%) (11.4%) (11.1%) (7.3%) (3.8%) (8.2%)

 

Note: figures in parentheses report the percentage change in employment for firms in this category since the

previous quarter.

Source: Jamaican Panel Survey of Microenterprise

Overall, from quarter II 1993 to quarter IV 1994 , there is a gradual

pattern of decline in employment for all the existing enterprises covered by

 

Although this method allows the sample to shift over time, it generates the

largest possible sample size with which to examine quarter to quarter

changes.

28

 

the survey“). Indeed, employment in existing microenterprises was
approximately 19 percent lower in quarter IV 1994 than in quarter II 1993.

The overall downtrend is reinforced when the data are examined over
rolling one year time periods. In Table 2.5, the annual percentage change in
employment between comparable quarters in 1993 and 1994 are portrayed.
The sharpest decline, 12.5%, occurred between the 2nd quarter of 1993 and
1994. Yet, even when quarters III and IV were used as the bases for the
calculation of the annual change in employment, the decline was sizable.
Clearly, there was a downward trend in the employment of existing
microenterprises in Jamaica over the 1993-1994 period.

Table 2.5

Year to Year Changes in Employment
Jamaican Quarterly Panel Survey of Microenterprise

 

Annual Change in Employment for Jamaican Microenterprise
1993 - 1994

 

 

 

 

 

Percent Change in
Employment
Quarter 11 1993 to Quarter II 1994 -12.5%
Quarter III 1993 to Quarter III 1994 -7.3%
Quarter IV 1993 to Quarter IV 1994 -9.8%

 

 

Source: Table 3.1
An irregular seasonal pattern of contraction and expansion is also evident

when the data are examined on a quarter by quarter basis. The largest

 

10 Consistent results in the decline in employment were similarly obtained by
examining a sample of firms for which we had data for all seven quarters.

29

 

decline takes place between the second and third quarters of 1993, the first
two quarters of 1994, and the last two quarters of 1994. Significantly,
employment expanded in only one quarter, quarter I of 1994, which
experienced a 3.3 % increase.11 Figure 2.2 illustrates these quarterly as well

as seasonal patterns.

Figure 2.2 Index of Change in Total Employment

Index of Employment

 

Qtr ll Qtr III Qtr IV Qttl Qtr ll Qtr III Qtr IV
Quarter II 1993 to Quarter III 1994

UIHdeIOf Totéi Employment HindexiOf P216 ETn’ployment

A breakdown of employment into some of its constituent parts reveals
important differences in the patterns of seasonal variations and secular

downtrend of employment. The discussion below examines changes in

 

The pattern of change from quarter to quarter was similar. The index for the
seven quarters were: 100, 93.2, 86, 89.2, 82.6, 81.7, and 79.6.

11 STATIN reports a decrease in national employment levels in 1993, but an
increase in employment in 1994, particularly in the latter half of the year. In
1994, the index of employment jumped by 10 points over 1993, while the
unemployment rate dropped by 5 % to 15.4%. The gainers in 1993 came

30

employment by labor type, gender, sectors, location, and size of
establishment.

In striking contrast to the decline in overall employment is the
increase in the level of paid employment over the same period, rising 5 points
by the last quarter of 1994. Was this because firms with paid employment
did better than those without? Our analysis could find no significant
differences in the overall employment performance of those firms with paid
employment and those firms with none. This result suggests that for firms
with paid workers, the brunt of the declining employment must have been
borne by unpaid family members, trainees, and proprietors. Of further
interest is the fourth quarter for which paid employment declines sharply
compared with earlier quarters (see Figure 2.2 above). This is due primarily
to a decline in the manufacturing sector, which employs a higher percentage
of paid workers than the other sectors.12 Most of the growth in paid
employment came from own account businesses expanding their operation by
one or two employees, representing a significant percentage increase in their

business.

Gender: It is noteworthy that female owned firms experienced a slightly

larger decline in employment than their male counterparts over the entire

 

primarily in the financial services sector while the manufacturing sector
continued its’ decline of 1993.

31

period, decreasing by 23% compared with 14% for male headed firms. Female
headed firms also experienced more quarter to quarter volatility. Male
owned firms declined by a large 8.9% initially but then were relatively stable,
while female owned firms fluctuated between negative 9.6% to positive 10.5%
in any one period. Thus, female owned firms declined somewhat more
dramatically overall and experienced more variability than their male
counterparts. This pattern can be explained somewhat by the increased
likelihood of female headed firms operating only on a part-time basis in
comparison with their male counterparts.13 Figure 2.3 below illustrates

these patterns.

Figure 2.3 Change in Employment by Gender

 

 

Index of Employment
«1 a
O C

so 1} ~

70 f

60

50 L , - , - 1 1 ._ .. 1.1., _
02 03 Q4 Q1 02 03 04

Quarter II 1993 to Quarter IV 1994

 

I'I-_"Male — FTamale Ref

 

12 Manufacturing, for example, contains 44% own account firms vs trade
with over 66% own account.

13 Patricia Anderson in “The 1992 Jamaican Microenterprise Survey” reports
that overall 15.0% of female entrepreneurs work less than 7 hours a day
compared with 9.7 % for male entrepreneurs. In manufacturing, this number
jumps up to 27.3% for females compared to 9.7% for males.

32

Sector: Employment in the trade and commerce sector declined the most, a
23 percent fall, and experienced the greatest quarter to quarter fluctuations.
This cyclical variation coincides with the peak in tourism during the winter
months.14 By contrast, service sector employment declined only slightly over
the period, 3.7%, and actually increased in 4 of the 6 quarters. Employment
in manufacturing was quite stable, typically declining only 1 percent per

quarter; it did experience, however, a sizable decline, -11.2 %, in the last

 

 

 

 

 

quarter.
Fi ure 2.4 Chan ein Em lo mentb Sector
y
120
o-v 110 - -
E100 -"_-:_-__-_- .-""""-_
E _ ~ ~ ~ -— —- -
W 80 \
'6
g 70
U
E 60
so
02 03 Q4 01 Q2 03 04

Quarter II 1993 to Quarter IV 1994
’—’ J ~1ng 1‘ 1.334 -T -TSe~?e

 

 

ref‘

 

14 Roughly 30% of petty trade firms, for example, shut down their operation
entirely during the “off season” compared with only 10% for wholesale trade.
(The 1992 Jamaican Microenterprise Survey, pg 41)

33

Location: Urban15 firms experienced less quarterly variability and declined
less over the entire period than their other rural counterparts. Except for
quarter II 1994, urban firms declined by less than 4% in any one period.
Employment in rural firms declined 9 percent more than urban firms over

the entire period with quarterly swings ranging from -9% to +4.3%.

Figure 2.5 Change in Employment by Location

120

 

 

 

 

 

“110
5
E100 -‘____-—-—-—I-—.
O \.
3,90 ‘-—-—_
E —_
E 80»
0
5 70»
'U
C
_60
50 1
Q2 Q3 Q4 01 Q2 Q3 Q4

Quarter II 1993 to Quarter IV 1994

Ref

 

; ’- Urban —Rural

Firm Size: there were also significant patterns of change in employment by
firm size (see Table 2.2). The size categories, - own account, 1-4, and 5-9
employees - are consistent with the size classification used by STATIN in its
reporting of the 1990 sampling of small businesses and the 1992 Jamaican
National Survey. For the purpose of this analysis, the classification of each

firm’s size was determined as of the time it first entered the panel. The

 

15 For this report, STATIN’s definition of rural and urban are used (see 1990
publication on microenterprises).

34

strong performance displayed by the own account firms is particularly
noteworthy, showing positive employment growth in 4 of the six periods and
ending 1994 18 points above the starting point in 1993. This robustness may
be somewhat misleading, however, because, at least initially, one person
firms cannot decline and continue to exist. Employment in both medium (1-4
employees) and larger size (5-9 employees) microenterprises declined during
the period. The larger sized firms fared particularly poorly, declining three
consecutive periods by double digits and ending 40+ points below the initial

period. Figure 2.6 below illustrates these findings.

Figure 2.6 Change in Employment by Establishment Size

140

 

120 —'—--"'"'-.
.. — —

100 ‘

 

———~
———~

80

 

60 " x
40

Index of Employment

20»

 

 

0
02 03 Q4 01 02 Q3 Q4

Quarter II 1993 to Quarter IV 1994

 

 

7' ’7' TOwniAcct 7—7 1 1074 —. 5'16 9 3f

Examining the employment data by gender, sector, location, and size, it
would appear that firms headed by females, operating in the trade sector,

and in rural areas experienced the most significant employment declines

35

across the period. This pattern may reflect the higher concentration of
female headed firms in the trade sector as well as the higher concentration of
trade firms in rural localities. It is noteworthy, however, that male firms in
rural localities performed as poorly as well, and that female headed firms in
urban areas were similar to their male counterparts. This suggests that
location may be of overriding importance.

Overall, it is striking that only the service sector and own account
firms avoided the significant overall decline in employment from the second
quarter of 1993 to the end of 1994. The volatile nature of employment for
firms headed by females and those operating in the trade sector is also

noteworthy.

36

2.4.2 Quarterly Change in the Wage Bill

Wage data on microenterprises are also typically quite difficult to
obtain. The QPS generated weekly wage information for each individual
working at the establishment. The detailed individual data were added
together to determine the entire wage bill of the firm. This wage bill data,
shown in Table 6.1, paints a similar picture portrayed by the employment
and sales data; yet there are some differences. Overall, it is noteworthy that
the nominal wage bill declined by 12 percent over the period, suggesting a
secular decline in wages even in current dollars. The real wage bill, which
has been calculated by subtracting the quarterly change in inflation from the
quarterly change in nominal wages,16 declined by almost 42% over the period.
Although not quite as steep as the decline in real sales, this drop was
nonetheless substantial. Finally, recognizing that the amount of paid
employment increased over the period by 5 percent, there is evidence that the
decline in the real wage Ltes of paid employees was severe. Table 6.1 and

Figure 2.7 below summarize the results.

 

37

Index of Wage Bill

Figure 2.7 Change in the Wage Bill: Real Vs Nominal

1201

.0...
go-
60-
40»

20»

 

Quarter II 1993 to Quarter IV 1994

ONSmEaI gReall

Table 2.6
Patterns of Change in Firm Wage Bill of Microenterprises
Jamaica Quarterly Panel Survey

 

38

There were important variations in real wages by gender, sector, location and

firm size. These are examined below.

Gender: Contrasting sharply with earlier results, real wages in female
headed firms out performed male headed firms. The real wages of female
headed firms declined by only 30% while those of their male headed
contemporaries declined by 45%. Female headed firms also experienced less
quarter to quarter variability, with a steep decline only in the second period
of 1994 (-21.3%); their male counterparts, by contrast, experienced extreme
quarter to quarter fluctuations, ranging from -23.1% to +10.1%. Real wages
in male headed firms increased in only one quarter, while those in female
headed firms increased in two. Overall, this reverses the pattern
demonstrated in both the sales and employment data where female headed
firms experienced the greater degree of fluctuation. Figure 2.8 below

displays the patterns.

39

Figure 2.8 Change in the Real Wage Bill by Gender

 

 

 

 

 

 

 

120
E
e
2
3
3
E 40 4
8
E 20 .
o 1
02 Q3 Q4 01 02 Q3 Q4

Quarter II 1993 to Quarter IV 1994

Ref ',

 

" FaEE'i‘fgsie - -
Sector: the real wage bill for manufacturing takes a roller coaster ride
between 1993 and 1994. In 1993, manufacturing wages increased by 112
index points and in 1994 declined by 125 points. This dramatic pattern
contrasts sharply to the fairly stable pattern observed in manufacturing sales
and employment data. Nonetheless, manufacturing ends 1994 in a better
position than either trade or service. Real wages in the trade sector, for
instance, dropped overall by almost 50%, a bigger decline than in its
employment. Further, all three sectors experienced similar erratic
percentage swings in their real wage bill, with manufacturing increasing by
83% one quarter and declining 44% the next. Finally, both trade and service
experienced declines in their real wage bill in each and every quarter. (See

Figure 2.9 below.)

40

Figure 2.9 Change in the Real Wage Bill by Sector

 

225
200 ~
175 ~
150 ~~
125 ,,

‘mfir:_

75 . ‘
25 4»

 

 

 

Index of Real Wage

 

 

 

Q2 Q3 Q4 01 02 Q3 Q4
Quarter II 1993 to Quarter IV 1994

L—* I — fir; ’—'_'Tr;<i:e_: "-1 5831;

 

Location: another striking contrast are the differences between the real
wage performance of rural and urban firms. Although urban firms faired
better than their rural counterparts in employment and sales, rural firms
demonstrated a more stable, less volatile real wage bill than their urban
components. Urban firms, by contrast, closed the year with a wage bill
nearly 50% below where they began while that of their rural counterparts
decreased by roughly 25%. The patterns of real wage changes from quarter
to quarter do not correspond very well to either the employment or real

sales/output pattern.

41

Figure 2.10 Change in Real Wage Bill by Location

 

 

 

 

 

 

 

120
1°0\
g ’ N \
_ 80 ~ ‘ ’ ~
g Q — §
1: 60» " ~
.5 § I I I I I I
x 40»
'3
E
20.?
0 1 1
Q2 Q3 Q4 QI 02 Q3 Q4

Quarter II 1993 to Quarter IV 1994
'- 5 Urfian—V ' "1465* ’

 

_ Ref

Firm Size: closely paralleling the pattern of change in employment, own
account firms outperformed medium and large size microenterprises in wage
bill growth, ending 1994 almost 12 points above its 1993 level. Further, real
wages in own account firms increased in all but one period (quarter 2 1994),
while in the larger microenterprises, real wages decreased in each and every
quarter. Clearly evident across employment, real sales, and now real wage
bill growth is the downward spiral of the larger microenterprises. Across all
three measures, the larger firms declined, experiencing volatile swings in

their quarter to quarter changes. Figure 2.11 below illustrates the results.

42

Figure 2.11 Change in the Real Wage Bill by Size

 

 

 

 

 

 

 

 

 

120 ‘__‘
en‘—

'3' 803 ‘~\..\
- \.-----.----..-
3 60 _—_~~ .-.----
m ~——~\
:40- ‘1
E20.

0 1 1

OZ Q3 Q4 01 Q2 Q3 Q4

Quarter 11 1993 to Quarter IV 1994
—'~ _ "Own Acct"- 74 7* T104 ;——_5 169

 

7 (Ref 1

2.5 Five Year Retrospective on Employment

How do these quarterly findings for the 1993/1994 period fit into a
longer term perspective on the performance of existing microenterprises in
Jamaica? A potential clue comes from additional employment data on this
same panel of microenterprises, extending back to 1990 that provides an
annual picture of employment for each enterprise for June of the given year.
For 1990, employment data were captured from the 1990 Census of
Microenterprises which included data on total firm employment. For 1991
and 1992, data were collected through the 1992 Jamaican Microenterprise
National Survey, which queried proprietors with in depth questions on firm
employment for both June of 1991 and June of 1992. Since the 1990 Census

Survey provided the sample frame for the Quarterly Panel Survey along with

43

the 1992 National Survey, all three data sources were easily linked. Over the
five year period it was possible to link together employment data on 348

firms. Table 7.1 below displays the results.

Table 2.7
Levels and Annual Change in Employment: 1990 to 1994
Jamaican Microenterprise

 

Change in Employment between 1990 and 1994
Jamaican Microenterprise

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

n=348 firms 1 1990 1991 1992 1993 1994

Overall I 100 111.8 110.5 94.7 85.8
(+11.8%) (-1.2%) (-l4.3%) (-9.3%)

Employment by Sector

Manufacturing 100 107.0 110.6 91.1 84.6
(+7 0%) (+3.3%) (17.6%) (-7.1%)

Trade 100 111.3 108.3 95.2 83.9
(”1.30/9 (2.7%) (12.0%) (11.9%)

Service 100 115.4 116.1 95.2 91.4
(+15.4%) (+0.63%) (-18.0%) (3.9%)

Employment by Gender

Male Owned I 100 115.4 113.8 96.6 86.8
(+15.4%) (-1.4%) (-15.1%) (-10.2%)

Female Owned I 100 103.0 102.3 92.1 83.2
(+30%) (-.7%) (10.0%) (53.6%)

Employment by Location

Urban 100 108.7 112.9 94.4 87.3
(+87%) (+3.9%) (16.3%) (-7.6%)

Rural I 100 114.1 109.5 97.2 84.0
(+14.1%) (-4.0%) (-1 1.2%) (-l3.6%)

Size of Firm

Own Account 100 85.4 85.1 66.7 69.9
(-14.6%) (0.4%) (21.6%) (+4.9%)

1-4 Employees 100 125.9 126.4 105.4 99.4
(+25.9%) (+04%) (16.6%) (5.7%)

5-9 Employees 100 125.1 119.7 124.9 94.0
(+25.1%) (4.3%) (+43%) 4.247%)

 

 

Source: Survey Data

 

Overall, the five year employment trend for this panel of microenterprises

provides another picture of general decline. Employment in existing

44

enterprises in 1994, for example, was 14.2 points lower than in 1990. Yet,
this decline was not even from year to year. Employment actually increased
in 1991 and decreased only slightly in 1992. Indeed, it was only in the last
two years that the sharp drop in employment took place, particularly in 1993.
Thus the QPS, which covered seven quarters of 1993/94, was providing a
picture of microenterprises at a time when they were apparently experiencing
their worst performance since 1990.

What factors might have accounted for the relatively poor performance
of microenterprises during the 1990’s? The macro economy clearly played an
important role. Real GDP, for example, increased relatively little over the
entire period, ranging from 0.7% to 1.5% per year. With population
increases, real GDP per capita was stagnant, and in some years, actually was
declining. Currency devaluations, increases in the inflation rate (ranging
from 27 to 80 percent per year), along with increases in the sales (GCT) tax
contributed to the worsening performance of microenterprises as the period
progressed. Additional insights into the longer run trends can be gained
from a more disaggregate analysis. Variations by gender, sector, location

and firm size will thus be scrutinized.

4S

Figure 2.12 Five Year Retrospective on Employment:
Jamaican Microenterprise

"ll

1990 1991 1992 1993 1994
1990 to 1994

 

 

 

 

Index of Employment

 

 

 

 

Refi

VJ-TEmpToymeth

 

Gender: male headed firms declined less overall than their female
counterparts over the five year period, the same pattern revealed in the QPS.
Male headed firms, however, experienced wider year to year fluctuations
over the longer period than female headed firm, opposite the pattern revealed

in the QPS.

46

Figure 2.13 Five Year Index of Employment by Gender

120

110+ /
100 --.---------

 

 

 

 

  

7o .3

Index of Employement
l
l

 

 

50 ‘
1990 1991 1992 1993 1994

1990to 1994

 

 

 

Males - -" ?- ’Feriweles

Rei

Sectors: all three sectors declined between 1990 and 1994, the service sector
declining the least and trade declining the most. In striking contrast to the
quarterly data, all three sectors between 1990 and 1994 followed a similar
pattern varying together almost in lock-step. In particular, manufacturing
departs from its quarterly pattern of near zero change, the trade sector does
not demonstrate the steep decline of the quarterly periods, nor does the
service sector follow a positive, stable rate of growth. Strikingly, the service
sector experienced the most volatility across the five years, with a 15.4% gain
in 1991 and a 18% decline in 1993. The trade sector, by contrast, fluctuated
the least. Most noteworthy, all the sectors moved in tandem, increasing and

decreasing by moderate swings in each of the five years.

47

Figure 2.14 Five Year Index of Employment by Sector

120

 

1104 ‘n';-__.I—If-_~\.

 

80 «»
70 -
60

Index of Employment

 

 

 

1990 1991 1992 1993 1994
1990 to 1994

 

 

— 6M9 'i'ii "Eda.
Location: rural firms outpaced urban firms in employment between 1990
and 1993, dropping below the urban employment index only in 1994. In 1991
the rural sector experienced a large percentage gain (14%) that drove this
result. Notably, both urban and rural firms decreased significantly over the
period and experienced wide year to year percentage swings. This
corresponds fairly closely with quarterly data, which show urban firms

declining less (overall), although with less quarter to quarter variability.

48

Figure 2.15 Five Year Index of Employment by Location

120

 

110 » / K

 

80 »
70

Index of Employment
8
/
I

 

 

50
1990 1991 1992 1993 1994

1990 to 1994

 

 

Urban —’ —7Rurali

 

Ref ’

Firm Size: a distinctly different pattern of change is observed in firms of
different sizes. Although all firms experience significant yearly employment
volatility, the larger firms clearly outperform the own account category over
the five years. Strikingly, employment in the larger firms generates an index
above 100 in all years except 1994, and the intermediate sized
microenterprises end 1994 with the highest index level overall (99.4). This
pattern contrasts with the QPS data where the larger microenterprises fared

the worst.

49

Figure 2.16 Five Year Index of Employment by Size

 

 

 

 

 

 

 

130
1 I I I I I I g

120L ’2’ —--..___:_..—--—""\
E /’ ~ h \
e 110 . ‘ ~ \
e / ~ . . _
3. 100 - ' LL...
3 \
“El 90 4»
X
3 70 ~
5 7

60 »

50 I I

1990 1991 1992 1993 1994

1990101994
FOw’nAcct -’- 2-110'4 :—2 —stos

 

 

Ref .

In summary, this disaggregated View of the five year picture has
provided several insights not captured in the QPS. For the five year picture,
however, several important nuances jump out at the disaggregated level.
Most notably, manufacturing tracks much closer to the other sectors in its
pattern of decline; male headed firms experience as much year to year
volatility in employment as their female counter parts; urban firms as well as
rural firms fluctuate significantly between years; and the larger

microenterprises out-perform own account firms.

50

2.6 Conclusions

In this chapter, the secular trends of Jamaican microenterprises were
examined both on a quarter to quarter basis in 1993 and 1994, and on an
annual basis fi'om 1990 to 1994. The analysis revealed several important
variations in firm growth and change over the period, with sector, gender,
and geography the most notable. Perhaps the most important insight from
the analysis, however, is the wide fluctuation in quarter to quarter
microenterprise performance.

Overall, employment tended down by 20 percent for this sample of
firms. Female owned enterprises declined by 25%, 10 points more than their
male counterparts. Urban firms faired better than their rural counterparts,
and the largest firms in the sample (59 employees) declined by almost 50%.

By linking the QPS to earlier surveys, it was also possible to
ascertain how the existing Jamaican microenterprises performed over a
longer time period from 1990 to 1994. This exercise revealed that although
overall microenterprise employment was lower in 1994 than in 1990, this
decline was not even. Indeed, employment in existing microenterprise
increased in 1991 and the sharpest decline occurred 1993 and 1994, the years
in which the QPS was conducted. Thus, the quarterly results must be

interpreted with this fact in mind. At the more disaggregate level, the longer

51

term findings revealed that the major sectors moved together much more in

lock-step than indicated from the QPS.

52

CHAPTER III
DYNAMICS OF MICROENTERPRISE IN JAMAICA: A CROSS

SECTIONAL ECONOMETRIC INVESTIGATION OF LONG TERM
EMPLOYMENT GROWTH

3.1 Introduction

Firm dynamics have been the focus of much research and speculation
since Alfred Marshall debated the advantages and disadvantages of small
firms in the 1890’s. In the context of developing economies, most of the
research until recently has focused on the performance of the medium and
large scale firms. Even within the United States and Europe, only larger
firms have attracted careful scrutiny, unraveling the factors affecting firm
growth and change. However, as a result of empirical work in recent years,
attention has shifted towards the magnitude and characteristics of small and
microenterprises.

In examining the growth of microenterprises, two hurdles have
typically encumbered a robust examination of the characteristics and
determinants of growth dynamics. Most notably, especially in the context of
LDC’s, is the availability of any data on these types of enterprises. Until just

a few years ago, the significance of the contribution of microenterprises to a

53

growing economy was skirted. Further, data collection is expensive, time
consuming, and difficult to manage over large areas, such as an entire
country. Second, most of the study designs to collect data on
microenterprises rely on retrospective surveys to generate data. This results
in a high potential for measurement error, especially for data specific to firm
performance.

Chapter II developed a descriptive profile of microenterprises in
Jamaica, and began to flesh out how these firms have changed both in the
short and long run. The snapshot of firm growth and dynamics developed
there represents a new and unique perspective, due to the types of data
available and the manner in which this panel data was collected.

In this chapter, long run firm dynamics will be examined for Jamaica,
focusing specifically on determinants of employment growth through a cross
sectional OLS analysis.16 The objectives are threefold. The first is to
examine the key relationships between firm age, size and growth in a manner
consistent with other research on these issues. This will allow for direct
comparisons with findings from other countries. Second, this analysis will
incorporate additional human and firm capital dimensions to determine their
importance to firm growth. Finally, an assessment of long run dynamics will
be made, as these data cover a five year period. How do the growth equations

change over time?

 

15 Parker (1991) shows that in at least 2 cases employment growth is highly
correlated with growth in sales.

54

Two unique elements distinguish these results from earlier research.
First, as a direct application of the validity (or lack of validity) of Gibrat’s
law, much of the recent theoretical literature posits firm growth as a function
of firm startup size. However, most empirical research use beginning period
size as the definition of firm startup size”. In most cases, this proxy is
necessary as data on firm startup size are not available. As a unique
contribution, the Jamaican data overcome this shortcoming as information on
firm startup size are available.

Second, as discussed above, the Jamaican data provide detailed
information on these micro firms over a five year period, whereas the
majority of research consists of the results from one time surveys. These
data provide a more complete dynamic picture of growth, capturing the
change and evolution of these firms over this longer time frame. Further, the
data collection instrument did not rely on retrospective data collection. This
reduced the effects of measurement error in the data.

In section 3.2, an overview of the research and theory regarding firm
growth will be reviewed. In section 3.3, the econometric techniques and
relevant issues will be discussed. The data collection methodology and
variable definition will be reviewed in section 3.4. Finally, section 3.5
summarizes the results from the cross sectional models. Conclusions are

presented in section 3.6.

 

17 Beginning period size refers to the size of the enterprise at the start of the
analysis period.

55

3.2 Theory and Hypotheses

Two interrelated lines of inquiry have received extensive attention in
the literature of firm dynamics: what causes a firm to grow, and what affects
firm survival (or failure). These questions are interrelated as a lack of
growth has much to say of individual firm survival, and the ability of firms to
survive affects the growth of firms in the aggregate. Further, both issues
have significant ramifications for policy directions. The examination of firm
growth has consumed much of the early research on firms, but most recently,

a significant amount of emphasis has focused upon firm survival.

3.2.1 Firm Survival

At the aggregate level, average firm growth is influenced by the growth
(or lack thereof) of existing firms, as well as by the complex interplay
between firm growth and decline, firm death and entry. Manfield (1962)
posited the importance firm exit might play in empirical investigations of
Gibrat’s Law, and recent empirical research has confirmed the large role firm
death plays in understanding firm growth. Liedholm and Mead (1991),
reviewing results of research across multiple African countries, show that
death rates of small firms average around 910 percent per year. Behrman
and Deolalikar (1989) in a study of medium and large scale manufacturing in

Indonesia, found that only 50% of the firms existing in 1975 survived 10

56

years. In Europe and North America, research has revealed similar patterns
for both large and small firms (Mata and Portugal 1994; Dunne, Roberts and
Samuelson, 1989; Wagner 1994).

Duration analysis (or hazard analysis) has been utilized in several
studies, both in developing and developed countries to determine which
factors affect firm survival. Several important hypotheses have been
explored. The most prolifically tested are the relationships between firm
survival with firm age and size. For the most part, an inverse relationship
between firm age and size with survival has been confirmed (Cabal, 1994;
Mata and Portugal, 1994; Audretsch and Mahmood, 1991; Behrman and
Deolalikar, 1989), although some exceptions have been found (McPherson,
1992).

Some other key relationships to survival have been explored as well,
mostly due to unique datasets built in developing countries. Most notably,
McPherson (1992) found in several African countries that neither gender or
credit affected the hazard, but firm location, sector, and country did play a
role. Cabal (1994) with data from the Dominican Republic, in contrast, found
gender significant and location insignificant, but consistent with McPherson,
different sectors obtained different hazard rates. Access to credit had no
affect on survival.

A significant amount of empirical effort has been put against firm

survival, exit and entry in recent years. Most of the research in Europe and

57

North America has centered on larger firms, for which ample data exists, but
some more recent work has been focused on microenterprises in developing
countries. Without going into detail or noting the exceptions, the empirical
work has generally confirmed the recent theoretical innovations of

Jovanovic’s 1982 seminal paper. This will be discussed in more detail below.

3.2.2 Firm Growth

Firm growth, or explorations of the “life cycle” of existing firms, has
received extensive treatment in the literature, principally along two lines of
inquiry. The first has its’ roots in what is termed “Gibrat’s Law” or the Law
of Proportionate Effect. This relationship has been debated and explored
since the 1950’s. Second, the relationship between firm age (or learning) and
growth has developed within the past 15 years, principally due to Jovanovic’s
1982 theoretical “learning model.” Much of the research on these
relationships have focused on large firms, due to the availability of data as
well as a reflection of the research agenda of the day (Brock and Evans,
1989). More recently, a body of literature has surfaced examining these
issues with smaller firms, in both developed as well as developing countries.
McPherson (1992), however, points out that an encompassing theory of firm
growth really doesn’t exist. This remains the case at this writing as well.

The first of these lines of research, Gibrat’s Law, proposes that firm

growth is independent of firm size. If the market for a given industry

58

expands at a rate of 5%, then all firms in that industry, whether large or
small, will enjoy a corresponding 5% growth rate. This implies a log—normal
distribution of companies by size, as is considered a stylized fact by

Schmalensee (1989), and can be understood in the following terms:

A variate subject to a process of change is said to obey the law of
proportionate effect if the change in the variate at any step of the

process is a random proportion of the previous value of the variate.18

The attention given Gibrat’s Law can be readily identified with two causes.
The first concerns policy considerations, while the second relates to the
economic theory of small business growth.

Regarding policy implications, imagine a world where small firms grew
more quickly than larger firms. In an economic downturn, policies targeted
to encourage small business growth would naturally lead toward a quicker
sparking of employment opportunities. On the other hand, if Gibrat’s Law
holds, policy makers could target any business size segment and expect an
equal payback.

Economic theory is equally subject to the validity (or lack thereof) of
the Law of Proportionate Effect. Several major theories, for example, have

presupposed it. The most significant, perhaps, is Lucas’ (197 8) theory on the

 

13 This definition, attributed to Aitchison and Brown (1969), is quoted in
Chesher (1979).

59

size distribution of firms. Lucas posits the distribution of managerial ability
as the driver for the coexistence of small and larger firms within a market.
Firm costs are defined as c(q)/0, where 0 reflects managerial ability. Firms

with higher 0’s have lower costs. This results in a log-normal size

distribution of firms in market in equilibrium. Lucas (1967) also proposed a
model of adjustment costs which could account for the distribution of firm

size.

Another line of attack is characterized by Herbert Simon’s comments:

Ijiri and I have suggested that the growth of organizations may have
only a little to do with efficiency..., but may be produced mainly by
simple stochastic growth mechanism.... Without the introduction of
empirical evidence, neoclassical theory provides no explanation for the
repeated appearance of Pareto distributions of firm sizes in virtually
all situations where size distributions have been studied. These
observed distributions are difficult to reconcile with any notions that
have been proposed for optimal firm size, but are easily explained by
simple, plausible probabilistic mechanism that make no appeal to

optimality.19

 

19 Found in Simon, 1991.

60

These comments capture the thinking on the “stochastic” theory of firm
growth. This class of models characterized firm growth as the outcome of a
series of lucky draws from a random distribution. Scherer (1980), for
example, matched outcomes of probabilistic simulations with marketplace
reality and found a consistent pattern across experiments.

Some of the empirical evidence supporting Gibrat’s law include Hart
and Prais (1956), Simon and Bonini (1958), and Contini and Revelli (1989)
and Wagner (1992) in some instances. McPherson (1992) points out,
however, that recent empirical evidence has more generally refuted Gibrat’s
law, leaving the door open for new theories of firm growth. Some of these
empirical studies will be reviewed in more detail below.20

Several new models of firm growth debuted in the late 70’s and early
808”, but Boyan Jovanovic’s (1982) represents the most influential new
arrival. The key parameter to the model, similar to Lucas’ (1978) model,

relates managerial efficiency to firm cost:

 

20 Some of the studies rejecting Gibrat’s Law include Hymer and Pashigian
(1962), Singh and Whittington (197 5), Chesher (1979), Evans (1987), Hall
(1987), Dunne, Roberts and Saumuelson (1989), Dunne and Hughes (1990),
Mata (1994), and Reid (1995). The classic approach to addressing the
question was established by Prais (1956), and followed by Chescher (1979)
and others. Hall (1987) and Evans (1987) followed a different tact in
modeling growth as the dependent variable. A very innovative approach is
followed by Stanley, et a1 (1994), who borrowed Zipf Plots from Physics, and
proved Gibrat’s law in fact does not hold.

21 Also consider Kihlstrom and Laffont (197 9) or Pakes and Ericson (1987).

6I

TotalCost(q,) = C(q, ) * S(0 + 8,)

where q refers to output, 0 refers to managerial ability and s are firm specific

costs. Firms only learn their level of managerial ability (or firm efficiency)
over time, so as time passes and they learn more of their own immutable
ability, their guesses as to the value of 0 becomes more accurate. Younger
firms, not knowing the value of 0, might possibly grossly over or
underestimate it’s value, resulting in the potential for higher rates of growth
to hit the optimal level of q. On the other hand, older firms which have
successfully determined the value of 9, produce at or close to their optimal
level and consequently experience slow growth.

The appeal of Jovanovic’s model is both intuitive and empirical. At
the intuitive level, the theory suggests a learning disequilibrium, resulting in
the potential for rapid small firm growth to achieve a stable path. On the
empirical side, firm “learning” can easily be proxied by firm age. This
approach has been followed by many authors (Hall, 1987; Evans, 1987;
McPherson, 1992; Cabal, 1994), and is adopted here as well. Several other
recent theoretical frameworks have been proposed more recently and bear

discussion even though they do not offer hypotheses testable by these data.22

 

22 None of the new theories are specific to either micro firms or on LDC’s.

62

A slightly different formulation of Jovanovic’s model, for example,
comes from Pakes and Ericson (1988), who term Jovanovic’s theory a “passive
learning” model. Fundamentally, Jovanovic depicts 0, or managerial ability,
as immutable. Managers who have extraordinary “ability” observe and
correct output correctly and grow more quickly. Other firms do not. Pakes
and Ericson posit a model in which firms can invest in research and
development. Successful firms in the process grow and obtain a large size,
while the others exit. Hence small firms that just start the discovery process
are the most likely to grow.

Cabral (1995) adopted Jovanovic’s learning model, but added sunk
costs to more thoroughly flesh out the negative relationship between firm size
and growth. Cabral assumes that each firm is endowed with a variable cost

function

5,6161, / S.)

where q is quantity and s is the firm’s efficiency type at age t. As in
Jovanovic’s model, firms are price takers and price is constant in all periods.
By making two assumptions, however, Cabral proves that this formulation

can be consistent with Gibrat’s Law, but only for a special case:

Assumption 1: L < F < H

Assumption 2: E(s.2 I 51) = 81

63

Regarding assumption 1, if 81 = L, then firms will exit the market as the
robability of survival is zero. Alternatively, if 81 = H, firms have a 100 %
probability of staying. Firms with S1 = F have a probability of survival equal
to 0 < a + B < 1. Profit maximization, given the above cost function, shows
that q1 and 51 are proportional. In a proof, Cabral show that for firms with L
or H probabilities, Gibrat’s law holds, while for firms of efficiency type F
Gibrat’s law can be violated.23

Cabral adds, however, that firms must build capacity at a cost of k per
unit, and that this cost is totally sunk. This results in the following: in each
period, before choosing qt, firms much choose capacity K; and pay k(Kt — K“),
where K is the cost of capacity. If this is true, Cabral argues, then Gibrat’s
law will not hold.

In a stylized framework like this, we can imagine large firms with a
zero probability of failure. This zero probability would encourage larger firms
to invest in their optimal capacity from the start. In contrast, small firms
face a positive probability of exit, encouraging them to invest in small
increments up to their long range capability. If they invest a lot in their first
period with a positive probability of failure, then they face a high potential

loss if their level of s1 is quite low. Firms that succeed stay in the market and

64

continue to invest at a proportional cost of k to their Kt — Km capacity
adjustments; hence, these small firms experience a “supra-normal” growth
rate.

The theoretical constructs of both Jovanovic and Cabral offer
reasonable explanations for the failure of Gibrat’s law. Obtaining data on
firm “sunk costs” or a proxy for such poses difficulties for empirical testing,
however, especially in lesser developed countries where data is difficult to
obtain.

Cressy (1996) proposes another dimension pertinent to the growth of
new entrants. Citing the work of Frank (1986) and Hudson (1990), Cressy
suggests that pre-entrepreneurial income drives new firm growth and
survival rates. Dropping the “learning” model paradigm, he develops a model
with firm growth as a function of pre-entrepreneurial income, age of the
proprietor at startup (to proxy human capital), other business characteristics,

and target income growth. More specifically:

30 =(1+k)f(%aao,xo)68‘

k

23 Cabral’s argument parallels that of Mansfield (1962).

65

where k is income growth, w is pre-entrepreneurial income, a is age of
proprietor at startup, x is other firm characteristics, and c is initial cash flow
of sales. Taking a second order log log expansion, Cressy tests this
framework in a manner comparable to Evans (1987a). Cressy finds pre-
entrepreneurial income to play a significant role in firm growth, but is unable
to validate the relationship with the age of the proprietor.

One argument against Cressy’s model might be that the “pre-
entrepreneurial” income variable is just a fancy proxy for managerial ability
or proprietor learning. Further, in discarding Jovanovic’s learning model, a
very strong intuitive explanation of firm growth is lost.

Another dimension of firm growth not touched upon by any of these
authors is the relationship between firm growth and profitability. Dobson
and Gerrard (1989) modeled the growth of firms in the Leeds engineering
sector with profitability as a dependent variable. Building on what is termed
the Penrose effect (Penrose, 1959), the theory behind the model posits that
there is a trade off between firm growth and firm profitability. If you choose
one, you have to trade away some of the other. It is easy to see the potential
effect firm age might play on this relationship as well. Older firms, for
example, may choose growth in their early career, and profitability later as
long as they survive. Dobson and Gerrard found a negative relationship
between growth and profitability with profitability defined as the ratio

between operating profits and sales, but a positive relationship between the

66

two with profitability defined as the ratio of operating profits to net total
assets. This latter result is consistent with other findings (Cubbin and
Leech, 1986, and Grinyer and McKiernan, 1983). Their framework is broadly
consistent with the Marris growth model and Mueller’s life cycle theory of
firm growth.

Arrighetti (1994) proposed a two firm typology to explain differences in
firm growth. Briefly, O-D firms are defined as low efficiency, low cost firms
or industries where little or no growth is required after startup to turn a
normal profit. For O-D firms, for example, initial startup capital is typically
not a problem. S-O-D firms enter at a sub-optimal size and need to grow in
order to survive. These firms are typically in more technologically advanced
industries, for example. Through the use of cluster analysis and some other
descriptive statistical techniques, Arrighetti showed this typology can be
useful in explaining differences in growth rates. For the Jamaican sample of
microenterprises, the majority (if not all) of the firms fit the OD typology as
described by Arrighetti. In fact in most LDC’s, microenterprises would fit
this criterion. This dichotomy doesn’t add much to the debate in LDC’s.24

There are several other significant factors which affect growth that
have not been specifically included in theoretical models. These include the
role of credit in firm growth, educational background of the proprietor,

gender, location of the firm, macro economic factors, and customer types.

 

24 See Davies, et a1 (1992) for a review of several other potential firm
typologies.

67

Several of these will be touched on briefly below, followed by hypotheses
testable with the Jamaican data.

Access to appropriate credit and financial resources has been a hot
topic for microenterprise programs in recent years. Although the financial
needs of microenterprise in most instances are small in absolute numbers
(Liedholm and Mead, 1987), the significance of the role of credit cannot be
understated. The need can be identified at multiple junctures, from startup
to operating capital (Rhyne and Otero, 1994), but since the entrepreneurs of
these enterprises typically lack assets as collateral to a traditional bank loan,
special programs have been designed to provide credit. Reed and Befus
(1994) also point out the significance of what they call “transformation
lending,” or obtaining the necessary capital to grow the microenterprise into
a small business.

For Jamaican microenterprises, only small percentages of the firms
reported accessing formal credit as a source of startup capital. Anderson
(1994) reports that between 0.7% to 16.7 % of the microenterprises accessed
funds through commercial banks or other programs, depending on sector (see
Table 3.1), with the majority of funds coming from personal savings or the
sale of assets. Another source of funds for micro-entrepreneurs not reported
in Anderson’s study is through the Jamaican black market (Grosse, 1994).
Access to some form of credit appears to be an issue for many Jamaican

microenterprises.

68

Table 3.1

Main Sources of Starting Capital for Jamaican Microenterprises

 

 

   

      
 

   

    

 

 

 

 

 

 

 

 

 

 

 

 

 

Industry Local Sale of Commerical Solidarity Credit
Savings Assets Bank Loan Program Union

Manufacturing 86.6
Construction 60.9 13.0 4.3 8.7 -
Wholesale 73.3 16.7 16.7 10.0 -
Trade
Retail Trade 88.6 14.9 6.2 5.5 1.3
Petty Trade 86.2 9.0 0.7 4.1 1.4
Restaurants, 94.3 14.6 5.7 5.1 3.8
Hotels
Transport 69.4 8.3 31.9 4.2 5.6
Professonal 79.2 1 1.3 13.2 9.4 1.9
Services
Social Services 100.0 23.6 23.6 7.3 1.8
Personal 82.3 26.8 2.6 2.9 2.9
Services
Motorcar 95.6 17.3 2.9 6.4 2.3
Repairs
Other Repairs 82.7 5.8 9.6 9.6 -
Total 86.6 16.6 6.1 5.1 2.1
N=2017

 

 

 

 

 

 

 

SOURCE: I992 National Survey

3. Table adopted from Anderson, 1994, pg 98.

b. All numbers are percentages and each cell is not mutually exclusive. If a firm accessed funds from both savings and a
commercial lender, then they are double counted in relation to a constant base.

The gender of the proprietor is another significant dimension of the

microenterprise landscape. Liedholm and Mead (1987), pooling information

across several African countries, highlight the large percentage of enterprises

headed by females.

69

Downing (1990) points out that the constraints and

responsibilities of female proprietors is different than their male
counterparts, potentially influencing firm performance and survival. In
Jamaica, Hotchkiss and Moore (1996) report that gender compensation
differentials in the formal sector approach 20%, with females receiving
significantly lower mean salaries than their male counter parts. Other
research in Jamaica has measured this difi'erential closer to 50%. Anderson
(1994) finds that female proprietors are more likely to have multiple
businesses and are less likely to obtain formal credit than male run
microenterprises. Gender should be examined as a dimension of
microenterprise growth.

The location of the firm is another relevant dimension. Different
market conditions between regions, or between urban and rural areas can
affect firm performance. Further, the physical location of the firm, whether
it’s a home business, roving, or in a bustling shopping complex may
determine the firms growth potential. Strassman (1987) reports, for
example, that home based enterprises in commercial areas generate more
income than their counterparts in rural areas. Schmitz (1995) found that
firms that grow in clusters have a greater potential for growth, suggesting
that commercial versus home business or urban versus rural might play a
role in the growth equation.

This leads to several testable hypotheses on the factors which influence

firm growth:

70

1. Firm growth is inversely related to firm age (Jovanovic, 1982).

2. Firm growth is inversely related to firm size (startup size,
Jovanovic, 1992).

3. The location of the firm will either hinder or contribute firm
growth. This applies to rural/urban location or the physical
characteristics of the firm.

4. The age of the proprietor is inversely related to firm growth, as a
measure of human capital formation (Cressy, 1996).

5. The firms historical access to credit, either for startup or operating
funds, should positively influence firm growth.

6. Beyond the individuals human capital, the gender of the proprietor
should affect firm growth.

7. Further defining the human capital of the firm, the educational

level of the proprietor should determine firm growth.

In the next section, the econometric issues involved in building a model of

this sort will be raised. Following this, the variables used for the analysis

will be described in detail.

71

3.3 The Econometrics of Panel Data

A huge volume of empirical research exists which examines the growth of
firms as functions of startup size, firm age, and other firm or market
characteristics. In most of these studies, the data have been cross sectional
and OLS has been the choice for analysis. Panel data, however, present
distinct advantages over cross sectional data in the examination of firm
dynamics and opens the door to additional econometric techniques and data
manipulations to address some estimation problems.

To begin with, panel data provide some unique analysis advantages to
either cross section or time series data. A brief list includes a reduction in
the collinearity among explanatory variables (Baltagi, et al, 1992), a better
control of the effects of missing variables (Hsiao, 1986), but more significantly
for this research, a better means of discerning the processes governing
growth and understanding the dynamics of change (Ashenfelter, et al, 1985,
and Hsiao, 1986). Ben Porath (1973) uses the following example: cross
sectional data tells the story that on average, fifty percent of women in the
United States are in the job market. What the cross sectional data does not
reveal, however, is if all women transition in and out of the work force
roughly half the time, or if one group of women is consistently employed with

the other not participating. Regarding firm growth and dynamics, the

72

application is clear. Examining the pattern of change for an individual or
firm over time results in a better understanding of the dynamic process.

The advantages alluded to above can be obtained by leveraging several
econometric techniques, data manipulations, or variable definitions. The
results are more efficient estimates than cross sectional OLS can deliver.
One of the main short comings of OLS can been seen in the following panel

model specification:

Equation 3.1

' I
yr: 2a + Bx” +721 +1111

where x11 and zit are vectors of exogenous variables. The OLS regression of yit
on Xit and 2: yields unbiased and consistent estimates of or, B and y. However,
if zi are unobservable and the covariances between xit and z: are not zero, the
coefficients on X“ will be biased. This is a heterogeneity bias, common, for
example, in the estimation of male-female wage gaps (Polachek and Kim,
1994) or any data which might contain unobservable individual
characteristics. With cross sectional data, this bias cannot be corrected
perfectly. With panel data, however, firm specific variables could be
introduced, the variables could be first differenced or estimated with mean

deviations to eliminate the source of the bias to produce an unbiased and

73

consistent estimate of B. An example can be visualized in the following

covariance transformation”):

Equation 3.2

(ya—.131): 13'(xir -f1)+(i‘lir _p’_r)
Sweeping out the means removes the bias introduced by the unobservable
variables which vary through time. Transformations such as this are not
possible with only one cross section of data or with a single time seriesZG.

Panel data hold out the hope at least of producing consistent unbiased

estimates.

Two common specifications used in the panel data literature are the

fixed and random effects models.

Equation 3.3

yr: :ai + 13x7: +72: +0411" “(I

 

25 First differencing, in this context, could remove the effect of the unobserved
variable z which is constant through time but varying across individuals.
Mean deviation, on the other hand, removes the effect of unobserved
variables.

2“ Deaton (1985) and others have developed techniques to mock panel data
with a series of cross sections.

74

In this general panel specification, all three possible types of panel variables
are depicted: X11 varies over time and firms (output, employment, raw
material costs), 2: varies only over time (interest rates, exchange rages,
GNP), and q: varies only over firms or cross sections (demographics). When
on is treated as a fixed constant, the model is referred to as a fixed effect

model. Alternatively, a random effect model treats or: as a random variable.

Although both assumptions have their own appeal in certain contexts, the
assumptions made regarding the orthogonality of the effect to the exogenous
variable(s) drives much empirical research into the fixed effect camp
(Ashenfelter, Deaton, and Solon, 1984). If the unobserved effect is indeed
correlated with any exogenous variable, the fixed effect formulation still
results in unbiased and consistent results, whereas the random effects
estimator will be biased. As seen in equation 3.2 above, the fixed effect model
“sweeps” the bias out. The problem with either first differencing or mean
deviation for the fixed effect model, however, is the eradication of the time
invariant variables from the model, which are often some of the key variables
of interest. The choice between the fixed and random effects model becomes
in part one of the efficiency of the estimates balanced against the objectives of
the research.

For all the benefits of panel data, they also come laden with several
problems, and in the words of Ashenfelter, Deaton, and Solon (1984), there is

little “justification for the mindless enthusiasm that is sometimes expressed

75

by econometricians” for the collection and analysis of panel data. Some of the
most serious issues affecting panel data include attrition bias, measurement
error, serial correlation, data collection cost, non-reporting, and border, seam
and rotation bias (Ashenfelter, et al, 1984; Hsiao, 1986; Lillard and Smith,
1986). Some of these problems can introduce significant bias into the
analysis. The first two of these are the most relevant for these data, and will
be dealt with in turn.

The intuitive effect of attrition bias is obvious. If systematic patterns
of individuals or firms drop out of the panel sample or decline interviews, the
panel can become lopsided and reveal results that pertain not to the
population or intended sample, but to a unique sub-sample of firms. The
existence of attrition bias is difficult to ascertain and even more difficult to
control for. For these data, several potential sources of attrition bias exist.
Firm deaths, moving firms, the remote location of some firms, and logistical
problems with the enumerators, who were depending on public
transportation much of the time. Several recent studies of firm dynamics
have tested for the effect of attrition bias on estimates and found little or no
effect (Hall, 1987; Evans, 1987; Dunne, Roberts and Samuelson, 1989;
Wagner, 1992). In section 4.5 below, the role of attrition bias in these data is
addressed following Hall(1987).

Measurement error, if it exists, is a problem for any type of data or

estimation technique. For panel data, however, measurement error can

76

result in extreme coefficient bias. Ashenfelter (1984) uses the worst possible
scenario to describe the potential bias. For a sensitive variable, say
household income, the data collected for each individual might be reported at
random (or laden with error). The calculation of the change between two
random numbers will grossly exaggerate the dynamic process. Pischke
(1995), for example, reports excess variability in individual household data
reporting income; a large portion of this “excess variability” is. due to
measurement error. Hence, the ratio between the measurement error
variance to the total variance will be quite large. For an OLS estimate, by
contrast, this ratio will be quite small. Given other less extreme scenarios,
panel data may introduce lesser amounts of bias into the estimates.

A solution to the measurement error problem is the use of
instrumental variables (IV). IV estimation has been suggested in many
difierent ways (Hausman, 1984; Hsiao 1986, review many different
techniques; Chamberlin, 1981). By definition, an IV variable, Z, is a new

variable such that

Equation 3.4

plim(Zz,.8,)/n = 0

77

This criterion is met by a second lag on the dependent variable(y 11-2) or by (y
1,82 — y 1,93) (Hall, 1987; Hsiao, 1986).27

In summary, a cross sectional data set analyzed with OLS techniques,
followed by a large body of past empirical research on firm growth, contains
some potential estimation problems. Several of these problems can be
addressed through the use of panel data. In the analysis which follows, cross
sectional models of firm growth are considered first. This will enable more
direct comparisons to past research and serve as a baseline to compare to
panel data models. OLS will then be utilized in a panel data model,
including an instrumental variable approach to correct for measurement
error. This will serve as good measure of the bias introduced by cross
sectional data, OLS, and measurement error. To correct for
heteroskedasticity, the Huber correction is applied, and firm age is
incorporated into most of the models (firm age serves as a firm level variable
which can capture heteroskedastic effects in OLS). A fixed effect and first
differenced OLS model are also considered, but the results from these models
are untenable. An unbalanced panel was used for all estimations, mostly for

the purpose of maximizing the analyzable sample size.

 

27 For these data, y 1,1.2 was used, as the other specification erodes the sample
size available for analysis.

78

3.4 Data Definitions

The dependent variable for the employment models that follow were either
the natural log of employment, or the natural log of growth in employment
between two years (see explicit definition below). All the models are in log
log form, to minimize correlation between variables, capture appropriately
non-linear multiplicative relationships, and ease interpretation of the
variables. The exogenous variables were defined as follows (all continuous

variables are in natural logs):

Time hrying Variables

Firm Startup Size: the number of employees at firm startup.

Lag on employment or lag of employment growth.

Firm age/Firmage 2: the age of the firm and the firm age squared, in years.
Firm Age" firm size: interaction term between firm size and age (note the two
definitions of firm size).

Seasonal Variables: controlling for seasonal or macro economic effects

Time Invariant Variables
Gender: dummy variable for male or female proprietor
Age of Proprietor: nine dummy variables for the proprietors age.

Pre Education: dummy variable for level of education completed.

79

Post Education: dummy variable for business training obtained by the firm
after startup.

Startup Credit: dummy variable for firms which obtained startup capital
fiom a lender.

New Credit: dummy variable for firms which obtained a loan from a lender
after startup.

Parish: fourteen dummy variables for the parish the firm operates in.
Location: nine dummy variables to describe the physical location of the firm.
Rural or Urban: dummy variable to indicate the urban or rural location of the
firm. The definition for rural or urban follows STATIN’s classification, which

follows the UN. classification of localities of less than 20,000 for rural.

Not all of these variables are used in each analysis, but the definition
of the variables is consistent across model runs. The time varying variables
capture the important dimension between firm size, age and growth.

Several other variables capture other human capital characteristics of
the firm. The education and training variable, as well as the variable on
proprietor age fall into this category. Next are two variables that detail the
firms success in obtaining credit, either at startup or post-startup for
operational or expansion purposes.

Finally, several variables characterize the firms physical location. The

country of Jamaica is divided into 14 parishes. Each parish is a different mix

80

of rural and urban areas, and this dimension of firm location is captured in
the rural/urban dummy. The firms actual physical location is characterized
by another series of dummies. These variables describe if a firm is located in

a commercial building, a yard, a private house, in the market, etc.

3.5 Employment Growth: A Cross Sectional Analysis of Jamaican
Microenterprise between 1990 and 1994
3.5.1 Introduction

In this section, a formal review of the relationship between firm
growth, startup size, and firm age in the Jamaican context will be examined.
The analysis follows the majority of the earlier research from other countries
on this topic, utilizing a cross sectional analysis with OLS as the principal
econometric technique. As the focus in this section is on a five year span of
data, the issues addressed here should be considered long run effects.

As mentioned above, several important questions can be addressed in
this section. At the most basic level, what is the relationship between firm
growth and firm age and size, and how do these estimates compare with
results from other countries? Does the inverse relationship between firm
growth, and age and size hold? Secondly, how do human capital and other
firm characteristics such as location or access to credit influence firm growth?
Finally, this cross sectional analysis spans a five year period, yielding

insights into the robustness of the estimates, and their consistency or change

8|

over time. What are the long term dynamics of the MSE sector? The firms
examined here consist strictly of survivors.

An additional econometric issue will be addressed as well. In the
empirical literature, the dependent variable in the firm growth equation is
modeled in one of two ways, in levels (the log of employment) or as a
transformation into annual growth (the natural log of annual employment
growth). Both definitions characterize employment growth, but the latter
suffers from a negative correlation of the dependent variable with firm
startup size and firm age, two of the exogenous variables. By comparing
both approaches, the magnitude of this bias can be ascertained.

In section 3.5.2, a cross sectional model will examine firm growth with
log employment used as the dependent variable. In section 3.5.3, roughly the
same model specification will be used, changing the dependent variable to the
natural log of firm growth. Both sections review results for each year

between 1990 and 1994. Conclusions follow in section 3.6.

3.5.2 Five Year Cross Sectional Results: Log Employment

To begin with, these data will be analyzed in a manner consistent with
the majority of earlier research on firm growth, through an OLS cross
sectional analysis with the dependent variable defined as the natural log of
employment. Defined in this way, firm growth is growth in employment

since firm startup. Some of the research addressing firm growth in this way

82

include Prais (1956), Wagner (1992), Lever (1995) and Reid (1995) to name a
few (see Table 3.4 for more details).28 Each of the years is modeled
separately.

This first cross sectional approach adopts two different model
specifications. The first (equation 3.5) is a simple model and corresponds the
most closely with previous studies testing for the validity of Gibrat’s law.
The second specification (equation 3.6) incorporates other time invariant

dimensions, including demographics of the proprietor and firm.

Equation 3.5

In employment, =01 + firmstartupsize, + firmage,

Equation 3.6

In employment =0t + In startupsize, + 1n firmage, +
+ gender; + agepropridor, + preeducatbn, + posttrainiig, + precrediti

+ postcredit, + location, + parish + ruralurban

 

28 Many of the earlier studies model firm growth based upon net assets or
capital employed. For microenterprises in LDC's, this type of data is almost
nonexistent. Most firms do not practice any type of record or book keeping.

83

Table 3.2
Cross Sectional Model of Employment:
Jamaican Microenterprise, 1990 thru 1994

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variables Year of Analysis
1990 1991 1992 1993 1994
n=354 n=360 n=368 n=318 n=284
Dependent In Employment
Variable
Ln Startup .532803] .8845678 .8320861 .597368 .5910784
Size (10.590) (24.668) (19.687) (11.567) (11.200)
Ln Firm -.0640387 0061898 -.005256 .0464051 -.0072965
Age (-1705) (O.I86) (0.144) (0.951) (0.130)
Constant .5] 10044 .2432232 .2959694 .1663 I47 .2904917
(5.169) (2.728) (2.947) (l.2l0) (1.802)
R Squared 0.2574 0.57l6 0.5217 0.3050 0.3023
F Value 59 . 77 307.84 206.23 67.27 62.91
(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) 7

 

7 Source: 1990 Census, I992 National Survey, Jamaican Quarterly Panel Survey

Values in parentheses are t values.

The Huber correction of variance is used in all years.

84

Table 3.3
Complete Cross Sectional Model of Employment: Jamaican
Microenterprise, 1990 thru 1994

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variables Year of Analysis
1990 1991 1992 1993 1994
n=345 n=360 n=359 n=3ll n=277
Dependent In Employment
Variable
Ln Startup .439889 0.803507 0.733143 0.523619 0.541599
Size (7.905) (17.013) (13.892) (3.848) (9.642)
Ln Firm Age .0058346 0.06174 0.080363 0.143725 0.09653
(0135) (1.559) (1.899) (2.352) (1.404)
Gender of -.1829679 014925 0.17412 0.12476 0.18744
proprietor (2.861) (2.732) (3.038) (1.884) (2631)
Age of
Proprietor
<20 * i a e 9
2024 .5213594 (dropped) 019501 0.830963 01691
(2.578) (-1.382) (4.438) (0850)
25-29 .552959 0.248257 0.03925 0.580294 ~0-23694
(3.166) (1.828) (0.278) (3.475) (-1 .332)
3034 .8024992 0.284105 0.073812 0.777878 43-01019
(4.501) (2.241) (0.532) (4.509) (0055)
35-39 .5970606 0.24781 1 0.016924 0.592237 -0-04172
(3.421) (2.127) (0.133) (3.991) (-0-251)
4049 .5392764 0.1 13017 0.16567 0.543858 -0-24466
(3.062) (1.016) (1.223) (3.572) (4.522)
5059 .5403868 0.088312 0.17225 0.453122 -0-27041
(2.908) (0.699) (1.251) (2.471) (4353)
60-69 .4216193 0.125915 0.21334 0.441643 0.26696
(2.212) (1.006) (1.525) (2.382) (4354)
70+ .4712388 0.05212 -0.32426 0.463959 -0-32974
(2.060) (-O.328) (1.710) (2.136) (-1 .453)
Education .01842490 -0.00818 -0.00854 -0.07463 0.03977
(0.300) (0.151) (0153) (1.022) (1.216)
Business .0072192 0.059362 0.177403 0.123436 0.047725
Training (0.075) (0.609) (1.901) (1.009) (0.377)
Startup -.0128757 005165 002768 0.104465 0.06264
Credit (0.107) (0.512) (0.268) (0.788) (0.486)
New Credit 0971723 0.112713 0.144289 0.151908 0.048692
(0.948) (1.170) (1.494) (1.469) (0.492)

 

 

 

 

 

85

 

 

—Parish

i

i

t

i

*

 

 

 

 

  

 

     

    
 

Kingston
St, Andrew .097146 0.00441 0.102681 0.074805 0.195895
(0.645) (0031) (0.748) (0.438) (1.160)
St, Thomas 0324596 0.132176 0.138636 0.15606 0.17912
(0.143) (0.726) (0.692) (0.509) (0.590)
Fenland -.050654 0.01 1744 -0.07967 0.212306 0.935016
(0.176) (0.054) (0.516) (0.674) (4.101)
St Mary -.1567723 001162 002436 016586 013256
(0.892) (0066) (0.136) (0792) (0.792)
St, Ann -.181158 0.062301 0.080959 0.11219 0.033411
(1.203) (0.415) (0.538) (0.644) (0.179)
Trelawny .2997303 0.15999 0.18897 0.65919 0.23129
(0.747) (0.773) (0.886) (3.023) (0.885)
St, James .0151643 0.004446 0.00392 0.045876 -0.0361
(0.108) (0.035) (0.031) (0.286) (0.184)
Hanover -.1604284 0.19086 0.1578 0.21862 0.19152
(0.885) (1.489) (1.205) (1.297) (1.104)
Westmorerand -.1057035 0.025464 0.007379 0.25581 0.32387
(0.712) (0.169) (0.050) (1.407) (1.655)
sr_ 51(23me -.312715 0.14696 0.12573 0.27179 0.33426
(2.343) (1.204) (1.019) (1.769) (2.167)
Manchester -.0506963 0.0234 0.03481 1 0.0991 1 007658
(0352) (0.174 ) (0.260) (0.573) (0.429)
Clarendon -.0693708 0.0052 0.02667 0.03599 0.05077
(0.398) (0.036) (0.173) (0.21 1) (0.272)
St, Catherine -.1852519 0.048727 0.058552 0.17479 0.13128
(1.388) (0.341) (0.420) (0979) (0.694)
Location
Private Home " "’ * * *
Private Yard .102782 0.094743 0.200044 0.264695 0.160563
(1.083) (1.258) (2.333) (2.714) (1.647)
Leased Land .4762091 0.356134 0.415528 0.67353 -0-03085
(3.389) (2.065) (2.337) (4.033) (0195)
Open Land .0678048 0.564813 0.312048 0.421895 0.371207
(0.325) (1.858) (2.434) (2.175) (1.978)
Roadside , .1214862 000063 0.142309 0.319297 0.134012
(0.863) (0.006) (1.310) (2.060) (0877)
Commercial .1845714 0.156006 0.210638 0.272745 0.102472
Bldg. (2.526) (2.828) (3.948) (3.494) (1275)
Market -.076066 0.041048 0.099942 0.215489 0.045759
(0.706) (0.310) (0.752) (1.424) (0313)
School Gate .4272104 0.216045 0.214667 0.167754 0-065947
(1.609) (1.040) (1.451) (1.076) (0469)
Other .4012704 0.113018 0.121026 0.289116 0205623
(2.516) (1.203) (1.279) (1.446) (0.987)
Rural/Urban -.1704202 0.07376 0.09383 0.0339 0.05177
(2.650) (1.321) (1.625) (0.454) (0.689)
Constant -.0707647 002473 0.157551 0.67235 0.332288
(0.315) (0.150) (0.794) (2.943) (1.298)

 

 

 

 

 

 

R Squared 0.4229 0.6387 0.6231 0.4344 0.4497
E Value 8.97 32.89 26.27 15.67 24.30
(0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

 

Source: 1990 Census, 1992 National Survey, Jamaican Quarterly Panel Survey
Values in parentheses are t values; Sandwich estimator of variance applied in all years.

86

 

Tables 3.4 — 3.5 report the cross sectional results for the models in
natural log levels. The coefficient for each variable is listed, followed by the t
value in parentheses. For these cross sectional results, the R squares vary
between .26 and .64, and in each model the F statistic indicates the
hypothesis of no difference between coefficients should be rejected. The
Huber correction for heteroskedasticity is applied in all of these models”.
For these estimations, the interaction term for firm age has been dropped as
they were highly insignificant”. This interaction was adopted by Evans
(1987) to capture the non-linear learning relationship posited in Jovanovic’s
(1982) learning model. Appendix I contains some of the expanded models.

First, the relationship between firm size and growth follows the
inverse relationship posited by Jovanovic's learning model. A coefficient for B
< 1 indicates that smaller firms grow faster than larger firms, (3 = 1 supports

proportional growth, and B > 1 indicates larger firms grow faster (an

explosive system). The coefficient for firm startup size is less than one and
significant in all years, which means that there is a negative relationship
between total firm growth since startup and startup size. The coefficient for
startup size varies between .532 and .884 for the first model and between

0.440 and 0.804 in the extended model. This evidence clearly refutes Gibrat’s

 

29 Referred to as the sandwich estimator of variance, this technique has been
credited to both Huber (1967) and White (1980, 1982).

87

law. The coefficients on size for the extended model are consistent and
slightly deflated, reflecting the additional variance explained by the added
variables. The consistency in the results underscores the stability of the
model and the probable absence of multicollinearity in the data.

Comparing this result to other studies with a similar definition of the
dependent variable, the result is consistent (Reid, 1995; Lever, 1995), both in
sign and magnitude of the coefficients. Significantly, both of these studies
modeled firm growth as employment and not net assets. Wagner (1992)
reports a coefficient much closer to 1, but this work focused on firms of a
much larger scale. Further, estimates based upon net asset growth report
coefficients much closer to 1 as well.

An additional important aspect concerns growth dynamics across
years. Specifically, the coefficients on firm size fluctuate on a year to year
basis, but remain consistent in magnitude (always an inverse relationship)
and sign. For 1991 and 1992 in particular, firm startup size obtains a much
weaker negative relationship with firm growth. One plausible explanation
for this deviation can be found in the macro economy. In 1991, the Jamaican
dollar was liberalised. Consequently, the exchange rate jumped from 8:1 to
32:1, with the inflation rate soaring from 25% to 81 % and a corresponding

slide into negative GNP growth (see table 2.1).31 If, for instance, access to

 

30 One exception to this was the result for 1990. See Appendix I for the
details.

31 Hay and Louri (1994) suggest that inflation is the most detrimental macro
shock for mircoenterprises.

88

credit became tenuous in this period of contraction and this dimension is
correlated with the startup size/growth relationship, then an argument for
the shifting coefficients can be obtained. There is no hard evidence to explain
the fluctuation, but these extraordinary macroeconomic shocks in this period
provide a plausible explanation.

A second key result relates to firm age. For this sample, firm age in
the short model is statistically insignificant for each year estimated
(additionally see footnote 13), and significant only in 1993 for the extended
model. Further, the sign on the coefficient suggests a positive relationship
with firm growth. This conflicts with several other empirical studies (Reid,
1995; McPherson, 1992; Evans, 1987), and does not support Jovanovic’s
theoretical model. Looking across years, the sign in the short model
fluctuates between a negative and positive sign, but regardless, the variable
is insignificant throughout. This result, if it holds, contains very important
implications for the life cycle theory of the firm.

Very significantly, however, the relationship between the age of the
proprietor and firm growth is significant, although the signs on the
coefficients vary somewhat across the years. First of all, the coefficients are
jointly significant in some of the years, with several individual ages
significant as well.32 This result supports a firm life cycle effect, and if the

age of the proprietor is posited as a proxy for learning, then this provides a

 

32 The F statistic for the joint significance of Age of Proprietor variable is
3.36, 1.58, 2.04, 4.20, and 1.99 for the years 1990 through 1994, respectively.

89

link to Jovanovic's theory. Jovanovic's posited negative relationship,
however, finds support only in results for 1992 and 1994 (negative
coefficients). Second, the results across the years fluctuates substantially,
with the sign flipping from positive to negative in 1992 and 1994. In a broad
sense, there is a consistent trend in the coefficients, however. Younger
proprietors affect firm growth more positively (or less negatively for 1992 and
1994) than the oldest proprietors. This has intuitive appeal, and evidence in
other countries supports this trend (McPherson, 1992, for Botswana).
Overall, age of the proprietor plays an important role in explaining firm

growth, and for these data supercedes firm age.

90

680228 225 8983:... an... .8 o. 55 mm... .23 want 92:. .23— 8“. .n

68922.8 963 cm 55 4.8. 5.3 8:... 22:. 2:0 .226. 32 E S: .88.. E E E 8: m. .39: Macaoioi .o

2858...... Eu 5 E: :5 20?...

_ E 3:32: 2 .32: 05 .33.: .261. =— d

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

E..- :8 2.: 9.5.8352 22.5.58: 82 122 .52. 82m.

.2... 2.2.3. .8236: 25.38.55: 2.8.5.9.: $2.32 .82. 9...).

:23 o. 8.- 83 1 8a. 8.5.58 58...... 29.32 8.53.... 5.9.32 5253...: 82-82 .82. confides.

no..- 2.: .m: 2.5.835: 2.8.6.3.. $2.22 5.2. :8:

822.8%

82.- m2 2.: 8.32.... 22.32 5886.3: 2.2-92 5.... 9.2.3. 85...:
539.0 2.3223 .525qu .8 cos—.50

2.... m... 2.: 3.6.8988 .85 3.8%... .568 $2 38.. .2..er

8... 2 2.2.82 8%. _.< .5533: 2.2-2.2 .32. 23.

:25... .85. 2.2.2...

one... 859...... 8 85.8562 3.3.8.582 26863:... £2.52 .32. .38..

.32.

no... eoe. 2.: 8.5.2.... 8:82 98:. 82 £2.22 82.: as 2.5.:

mo. .3. 255.60 22 3.58.5.2... 2.2-23. .82. .883

.2. mom 2.: m.3. .5... 8.2.5. 983 .62 £2- . 8. .52. .e .o 5.92

86.55.". 52.

8. m... 2.: 8.52.58 ..< .288 3.82 .2685 2.... 22.88

eo. 2.: 2.: .23.... .62 988... 62 $82 :22. 8:52

52.

2.. RN 2.: 365.5282 88.3 .oz $-32 .5252 2... 22.5.5...

8.. 22 2.: 8.52.8 22 38...... 62 82.28. .82. 5:

2... 2.2. 2.: 2.? 5.... as... 22 235.85.. .22. _ W2 32. 8......

8.. 2m 2.: 8.? 5.3 9...... 22 982 .oz $2.32 .32. 8.35am

#2

2... R2 2.: 8.5858 365....-.52 982.. 82 82.22 8.2.8.23 2... 22.5

n 35 yin—am 2.5550 2.»... 32.3.5 «Eat—5 35 3...».— .353.

 

 

 

 

 

 

 

 

£964 5 033.34» 2.352.»: he nee—Eon

 

 

539.0 E...“— caa 35 amazon Eamnezfiem 2; Wm 033.

91

A third set of results correspond to proprietor characteristics. Certain
demographic dimensions of the proprietor provide important explanatory
insights to firm growth as well, although some of these results are not as
consistent across the years. For starters, the results for gender reveal that
female headed firms negatively effect firm growth. This is particularly
problematic in the micro sector, given the number of firms headed by females
and the predominance of female headed households in Jamaica. This builds
on the finding in Chapter II, which suggested that female headed firms
experienced slower employment growth than their male counterparts. Those
results also suggested, however, that these firms also experienced important
advantages over their male counterparts, which needs to be contextualized
when interpreting the importance of this finding.18

With the marginal exception of 1992, education, either formal school or
specific business training, had no effect on firm growth. In 1992, business
training was positively related to firm growth and marginally significant.
This is an important result given the large emphasis placed upon
microenterprise training programs in Jamaica. Actual access to credit played

no role in firm growth.

 

13 Gustafson and Liedholm (1995) detail a strong sales and output

performance of female headed firms, in contrast to languishing male headed
firms.

92

The last set of results relates to firm location. Both parish and
location are jointly significant or marginally significant in all of the years.19
With a few exceptions, however, only a few of the individual locations yield
statistically significant results. First of all, the majority of the parishes
exhibited a negative relationship with firm growth, but only the results for
Saint Elizabeth were statistically significant, with a strong negative effect on
firm growth. Saint Andrew, which incorporates some of the urban area of
Kingston, is the only parish with consistently positive coefficients. Regarding
the physical location of the firm, individual significant coefficients obtained
for firms on either leased land or in commercial buildings. The coefficients on
these variables varied considerably fiom year to year, however, with no
apparent consistent directional inspiration. With the exception of 1990, the
rural or urban location of the firm does not discriminate on firm growth. In
1990, however, rural firms were negatively related to firm growth. In sum,
location plays a significant although marginal role in firm growth, although
the lack of consistency in the results across time make drawing definitive
conclusions on dynamics difficult.

In sum, four key conclusions characterize the cross sectional model in
natural log levels: one, startup size is inversely related to firm growth (B<1).
Secondly, firm age does not explain firm growth, although age of the

proprietor does. Thirdly, gender matters in firm employment growth, with a

 

19 The F statistics for firm location are 2.66, 1.66, 2.77, 3.43 , and 1.93. The F
statistics for parish are 1.47, 1.83, 1.45, 1.67 and 4.16 for 1990 through 1994

93

negative relationship revealed for female headed firms. Finally, the location
of the firm is important, but the lack of consistency in the results across time
render definitive implications impossible. These findings have crucial policy
and empirical implications. But how robust are they when the dependent

variable is defined in a different manner?

3.5.3 Five Year Cross Sectional Results: Growth in Employment

A slight deviation from the above approach parallels several other
recent studies on firm growth (Hall, 1987; Evans, 1987; Dunne, Roberts, and
Samuelson, 1989; Arrighetti, 1994; Cressy, 1995), which have defined the
dependent variable as the natural log of employment growth. Defined in this
way, the dependent variable can be understood as the annual growth in
employment since startup. As alluded to above, however, this specification is
subject by definition to a negative correlation with the dependent variable
resulting in a biased coefficient on firm startup size and age. Both are found
on the right and left hand size of the growth equation (see definitions below).

By examining this specification, two crucial insights are gained. One,
a more direct comparison with these recent studies is achieved. Two, the
effect of defining the dependent variable in this way can be gauged. One
very notable difference to the above results: the interaction term for firm age

is left in as some of these coefficients are significant with this specification.

 

respectively.

94

The partial derivatives of growth with respect to firm startup size and firm

age are consequently reported.

The definition of firm growth used here is consistent with Evans (1987)

and Dunne, Roberts, and Samuelson (1989).

Equation 3.7

In employment = a + [3X +yZ + 8

with

In F irmGrowth = (ln employment” — ln startupsize 1.: 0 ) / firmage

where 01 is a constant, X is a vector of time and cross sectional varying
variables, Z is a vector of time invariant variables, and s is the error term

(see specification 3.4 and 3.5 for details on the exogenous variables).

95

Table 3.5

1990 thru 1994

Cross Sectional Model of Employment Growth:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variables Ending Year for Growth Model
1990 1991 1992 1993 1994
n=354 n=360 n=368 n=318 n=284
Dependent ln Annual Growth from Startup
Variable
Ln Startup -.2960843 -.0422646 -.0465315 -.1 177187 -.1201862
Size (-5.329) (-2.232) {-2.698) (-4.292) (-4.427)
Ln Firm -.206484 -.0333079 -.026l36 .0000492 -.0272202
Age (2.710) (-l.062) (.955) (.001) (.545)
Ln Firm .0274899 .0023201 .0010431 -.0030934 .0012247
Age (1.943) (.450) (.815) (.409) (.153)
Squared
Ln Firm .0967071 .011851 .0120388 .0315852 .0322358
Age * (4.693) (1.888) (.034) (3.578) (3.782)
Startup Size
Constant .3727841 .0939525 .0859612 .0459734 .086056
(3.804) (2.031) (2.085) (.669) (1.132)
R Squared 0.3256 .0656 .04722 .1512 .1727
F Value 15.45 6.94 8.09 14.10 13.59
(0.0000) (0.0000) (0.0000) (0.0000) (0.0000)
Partial Derivatives
Firm Startup -0.07538 -0.01342 -0.01631 -0.03477 -0.03287
Size at mean
Firm Age .. -0.03233 0.085812 -0.01465 -2.4E-06 -0.00428

 

 

 

 

 

Source: Survey

 

Data
Values in parentheses are t values.

Derivatives for startup size and firmage coefficients were evaluated at the mean.

96

 

Table 3.6
Complete Cross Sectional Model of Employment Growth:
1990 thru 1994

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variables Ending Year for Growth Model
1990 1991 1992 1993 1994
n=345 n=351 n=359 n=31 l n=277
Dependent In of Annual Growth from Startup
Variable
Ln Startup -0.31172 -0.06311 -0.062 -0.12819 -0.l4148
Size (-5.191) (-2.983) (-3.387) (-4.421) (-5. 126)
Ln Firm Age -0.24244 -0.06222 -0.04878 -0.02509 ~0.03858
(3.218) (1.743) (1.577) (.507) (.531)
Ln Finn Age 0.037031 0.008415 0.006207 0.002724 0.003911
Squared (2.512) (1.383) (1.177) (.325) (.385)
Ln Finn Age 0.097682 0.016347 0.014457 0.03244 0.037491
.1 Startup (4.400) (2.317) (2.391) (3.455) (4.371)
Size
Gender -0.03036 -0.0175 -0.01533 -0.01571 -0.0202
(-2.530) (-2.829) (~3.019) (2.343) (-3.019)
Age of
Proprietor
<20 1 t l t #
20-24 0.390534 (dropped) 0.004839 0.15 161 1 0.000896
(7.174) 0.0293 (.312) (6.561) (.040)
25-29 0.42787 (1.493) 0.026735 0.128983 -0.02685
(7.572) 0.035716 (1.560) (6.354) (-1.370)
30—34 0.438975 (2.073) 0.03 8704 0.150487 -0.00803
(7.997) 0.031663 (2.295) (7.171) (-.43 7)
35-39 0.415571 (1.979) 0.031633 0.132264 -0.00271
(7.134) 0.015525 (2.177) (7.142) (-. 146)
40-49 0.41 1261 (1.041) 0.014291 0.126823 -0.02228
(7.954) 0.01337 (1.004) (7.044) (1.307)
50-59 0.404875 (.835) 0.012685 0.1 19133 -0.02613
(7.609) 0.01268 (.849) (5.975) (-1.365)
60—69 0.376242 (.847) 0.008421 0.1 14148 -0.03099
(7.1 15) 0.00393 (.587) (5.859) (-1.658)
70+ 0.377791 (.259) 0.00509 0.1 16791 -0.02934
(7.920) (.328) (6.056) (-1.633)
Education 0.007239 0.002529 0.002266 -0.00384 -0.00612
(.647) (.395) (.399) (-.560) (-.954)
Business 0.002216 0.007504 0.017927 0.0101 13 0.004581
Training (.158) (.550) (l.825) (.912) (.427)
Startup 0.015664 -0.00304 -0.00345 0.014812 0.009064
Credit (.890) (.287) (.353) (1.181) (.805)
New Credit 0.008378 0.01217 0.014562 0.012158 -8.9E-05
(.595) (1.144) (1.514) (1.377) (.011)

 

 

 

 

 

 

97

 

 

Parish

 

 

 

 

 

 

Kingston "' * "' "‘ “
St. Andrew 000379 0.009142 0.018578 0.009545 0.012109
(.150) (.588) (1.558) (.511) (.831)
St. Thomas 000123 0.012234 0.010593 001201 001903
(.035) (.874) (.783) (.572) (1.093)
Portland 0.024051 0.003441 0.00838 0.017458 0.093475
(.339) (.1 10) (.458) (.452) (4.983)
St. Mary 002974 000323 000235 000458 000984
(1.055) (.221) (.175) (.282) (.598)
St. Ann 0.000501 0.02154 0.023528 0.001415 0.011478
(.028) (1.304) (1.583) (.087) (.590)
Trelawny 0.149578 0.01857 -0.0l68 00597 001928
(1.503) (.853) (.850) (2.577) (.975)
St. James 0.024572 0.00977 0.005722 0.012985 0.003477
(.955) (.780) (.595) (.925) (.188)
Hanover 001455 000957 000529 001 155 001091
(.541) (.799) (.552) (.735) (.738)
Westmoreland 000085 0.007547 0.005927 0.01 153 002455
(.039) (.535) (.451) (.715) (-1.511)
St. Elizabeth 003701 000559 000475 001443 002121
(1.594) (.582) (.454) (1.079) (1.531)
Manchester 0.022508 0.01014 0.013504 0.00557 -3.ZE-05
(.835) (.521) (.955) (.324) (.002)
Clarendon 0.002229 0.013303 0.011387 0.00551 1 00024
(.088) (.792) (.824) (.445) (.153)
St. Catherine 000514 0.015487 0.01 1837 000295 000534
(.288) (1.130) (.977) (.195) (.401)
Location
Private Home "' ' 5 "' "'
Private Yard 0.018178 0.006446 0.015161 0.020885 0.009735
(1.200) (.756) (1.720) (2.212) (1.128)
Leased Land 0.05957 002261 1 0028751 0.052595 -0.01228
(2.589) (1.383) (1.959) (3.405) (.959)
Open Land 0.002987 0034634 0019369 0.025205 0.024489
(.113) (1.554) (1.690) (1.922) (1.521)
Roadside 0.00339 -0-0071 I 000461 0.025794 000059
(.121) (--470) (.354) (1.435) (.038)
Commercial 0.023209 001389 0015585 0.025455 0.005213
Bldg. (1.588) (2-105) (2.919) (3.203) (.578)
Market 0.015813 0007214 (1010631 0.022714 0.005352
(.574) (.747) (1.102) (2.047) (.534)
School Gate 0.1155 0034992 0023744 0.015718 0.001317
(1.990) (1243) (1.532) (1.201) (.112)
Other 0.035173 0-01 1 ‘32 0-008025 0.022345 0.01 1234
(1.350) (.938) (.775) (1.102) (.555)
Rural/Urban 001795 000485 000452 000035 000345
(1.394) (.585) (.733) (.045) (.450)
Constant 0.0043 0.099972 0.077748 0.05854 0.128909
(.053) (2.120) (1.959) (1.040) (1.497)
R Squared .4580 .1980 .2549 .3197 .3458
F Value 145.80 1.57 2.32 59.08 5.52
(0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

 

 

 

 

 

 

98

 

 

Partial Derivatives

 

 

 

Firm -0.08879 -0.02332 -0.02571 -0.04089 -0.03994
Startup Size

It IIICII‘I

Firm Age -0.02424 -0.01272 -0.01011 0.006262 0.001575
It MCI“

 

 

 

 

 

 

Source: Survey Data
Values in parentheses are t values; Parial derivatives are calculated at mean values.
All equations estimated using the sandwich estimator of variance.

Tables 3.5 and 3.6 report the results for both the short and extended
models. The coefficient for each variable is listed, followed by the t value in
parentheses. The model is significant for all years, and the R squares range
from .05 to .47. The Huber correction is applied in all instances. In general,
these models exhibit lower R squares and F statistics compared tO the model
in log levels. In reviewing the results, comments will first focus on the
interpretation for Jamaican microenterprise, followed by a discussion of
model performance compared tO the specification in the prior section. In
general, the results are fairly consistent across the models, with the caveat
that firm age Obtains strikingly different results.

First, these results indicate again that firm startup size is inversely
related to firm growth. As the interaction terms remain in this model, the
partial derivative between startup size and firm growth reveals the relevant
relationship. These values range between -.01 to -.08 in the different years.
The pattern Of the fluctuation in the coefficient is similar to the pattern for
the model in levels, and as discussed above, this may reflect the shifts in the

macro economy over the time period. These results match with any number

99

 

Of earlier studies on the relationship between firm startup size and growth
(see Table 3.4). Specifically, Mata (1994) and Evans (1987) report coefficients
Of -.021 and -.072 respectively. Dunne, et a1 (1989) reports a coeficient Of -
.29. The sign on this coefficient, however, indicates a slightly smaller
negative relationship with firm growth than the previous model, reflective Of
the negative bias introduced into the estimation with this definition of the
dependent variable.

Second, the most striking deviation from the previous findings is with
regard to firm age. TO begin with, the coefficient on firm age is at least
marginally significant in all Of the years (in 1990, F=12.55; 1991, F=3.21;
1992, F=4.08; 1993, F=4.08; and 1994, F=9.00). This stands in direct contrast
tO the previous model results, where the hypothesis Of no relationship
between firm growth and firm age could not be rejected. Further, the
coefficient on firm age is negative (1990 - 1992) or positive and close to zero
(1993-1994), which again contrasts with the previous results but is consistent
with some Of the most recent findings in the literature confirming a negative
relationship with firm growth (Evans, 1987 or Mata, 1994, for example).
Similar to the context Of firm size and growth, the most plausible reason for
the result in this research is the negative correlation between the dependent
variable and the firm age variable. This draws into question, however, the
results from many Of the previous studies which have landmarked the result

on firm age as support for several very important theoretical models. For

100

these Jamaican data, the bias effects the sign and significance on the
coefficient.20

The other results from this model are broadly consistent with the
findings and discussion from the previous section, and hence will not be
reviewed here. However, two elements surface that do bear mention, both
relate to 1992: the dummy variable business training and access to new
credit become weakly significant. A surprising finding at this point has been
the lack of importance of both education and access to credit to firm growth.
Education is an important measure Of human capital accumulation, as well
as innate ability, desire for success, and other dimensions. In this analysis,
firm age and age Of the proprietor also measure some aspects Of these
dimensions. Access to credit, on the other hand, may simply suffer from
under reporting of information on this sensitive topic. In either case, the

results for 1992 can be at least recognized as pointing toward the possibility

 

2° TO provide clarity to this issue, a fixed effect model was estimated to yield
an unbiased estimate Of firm age. In a fixed effect model, if all of the
explanatory variables are exogenous, the covariance estimate is best linear
unbiased. Even in the case where there are omitted individual attributes,
which is likely the case with the Jamaican data, the fixed effects estimates do
not suffer. For this estimation, however, only the firm age variable is
modeled. This is because all of the time invariant variables drop out in a
fixed effect model, and a lagged variable (and interaction) on firm size is
dropped to avoid the negative correlation in the model. The results Of this
examination, however, yielded a model with a very low R squared and F
value, and coefficients on firm age and firm age squared which were
insignificant (the results can be found in Appendix B). This result is
consistent with the outcome Of the first model specification, where the null
hypothesis for firm age cannot be rejected (equation 3.1, table 3.2). Although
this result cannot refute the outcome Of the second model, it builds support

101

Of relevance. More focused research on both of these issues is needed to
unravel these questions.

Finally, the model in log levels from the previous section capitalizes on
several advantages. First, the negative bias resulting from the inclusion Of
startup size and firm age on both sides Of the equation is manifested in
biased estimates on both Of the variables. Second, the predictive power of the
second set Of models clearly suffers, with falling R squares and F values for
model significance. Although these measures don't represent the definitive
measure Of the best model, the inconsistent results in the age and size

variable do point to the strengths Of the previous specification.

3.6 Conclusions

This chapter examined long term growth dynamics through the lens Of a
cross sectional analysis. In so doing, classic questions were addressed on the
drivers Of firm growth, such as the relationship between growth and firm size
or age. Two different approaches were adopted, enabling a comparison to the
bulk Of previous research, as well as providing insight into how the two
approaches affect the results. Further, as the data covered a five year time

span, a unique cross year viewpoint was achieved.

 

for rejecting firm age by itself as a key defining dimension Of growth for
Jamaican microenterprise.

102

Three important findings sum up the results. First, the Jamaican data
clearly refute Gibrats law for this sample Of firms, although the magnitude Of
the inverse relationship was small in magnitude in most Of the years.
Secondly, The model in levels reveals that firm age is not negatively related
tO firm growth as posited by Jovanovic (1982), although the age of the
proprietor is related tO firm growth. Finally, the results on firm age appear
sensitive to the definition Of the dependent variable leading to some
contradictory results. In the remaining sections, this potential bias will be
accounted for.

The next chapter extends this analysis by again tapping into the
richness Of panel data. In the following chapter, long and short run dynamics
will be addressed through some common panel data econometric techniques,
hopefully to shed some additional light on these complex issues. Further,
some Of the potential bias and estimation difficulties inherent in these type Of

data will be addressed.

103

CHAPTER IV

A PANEL DATA MODEL OF LONG AND SHORT RUN
DYNAMICS FOR JAMAICAN MICROENTERPRISE

4.1 Introduction

The last chapter related some key insights Of firm growth based upon a
cross sectional OLS analysis. By reviewing the performance of these micro
firms over time, some light was shed on aspects Of firm dynamics. The
models presented thus far, however, have been linear static and not truly
dynamic. A genuine dynamic model by definition incorporates elements of a
firms past performance into the specification. This type Of analysis is not
possible in a cross sectional approach. Panel data presents the potential for
this type Of analysis, and provide several additional analytical benefits, only
a few Of which have been fully exploited thus far.

The Objectives Of this chapter are threefold. The first Objective is to re-
examine the firm startup size and growth conundrum utilizing a panel data
model. The analysis adopted here fundamentally extends the cross sectional
analysis of Chapter III by incorporating all the years into an OLS regression.
As this analysis follows the same specification as the model in employment
levels, the results here should highlight the added insight Of incorporating

the time dimension into the regression framework.

104

The second Objective, an extension Of this analysis, forges new ground
as panel data econometric techniques will be utilized to explore firm growth,
both in the long and short run. Significantly, a dynamic model of firm growth
will be introduced, incorporating a lagged dependent variable into the model
specification.36 Further, the models thus far have assumed that the variables
are measured without errors. This assumption is somewhat suspect, due to
recall error on the part Of the proprietor and other sources of error. For this
analysis, an Instrumental Variable (IV) approach will be introduced to
account for this type Of error.37 Additionally, very few theories of long run
growth for micro firms have been developed (McPherson, 1992; Evans, 1987),
and even less discussion has focused on short run growth issues. The
discussion here hopes tO shed some light on the connection between the two.

The final Objective addresses the effects Of attrition bias on this data.38

Panel data does not come without some special restrictions, and perhaps one

 

36 A lagged dependent variable model is in essence equivalent to a Koyck
distributed lag after appropriate transformations. However, in a lagged
dependent variable model, the principal estimation Challenge appears when
the disturbance is generated by white noise. This results in a disturbance
term that is essentially an MA (1) process. TO yield consistent estimates Of
the parameters, an IV approach is required.

37 Additionally, an OLS lagged dependent variable model will be inconsistent.
Hsiao (1986) shows, however, that an appropriate choice Of an IV variable
will result in consistent estimates.

38 Another issue is firm heterogeniety (firm fixed effects), which theoretically
play a significant role in this data. A few comments in this regard are in line.
A fixed effect model was fitted to the above data, with very poor model
results. The model is not reported here, in part because a fixed effect model
with a lagged dependent variable does not produce consistent estimates.
Even with the poor model fit, however, the result on lagged employment was
in line with the one other study leveraging this technique to examine firm

105

Of the most troublesome data problems introduced by the use Of panel data is
the issue Of attrition bias. In the final section, the general approach adopted
by Hall (1987) will be used to address the issue.

In section 4.2, the firm startup size panel model is reviewed. Sections
4.3 and 4.4 examine long and short run firm dynamics, respectively. Panel

data estimation issues are dealt with in section 4.5, followed by conclusions.

4.2 A Panel Data Econometric Approach to Firm Startup Size and
Firm Growth

This section extends the analysis Of the previous chapter by re-
examining the relationship Of firm growth tO firm size and age in a panel
data model using OLS. The same model specifications are used, with the
added advantage Of multiple Observations on each firm over time. Seasonal

dummies are also incorporated. As the variables in this regression are all

 

growth. Mata (1994) found that the coefficient on firm size Obtained a
negative relationship to firm growth, but with a sizeable adjustment in
coefficients from the OLS estimate. The coefficient on the Jamaican data was
-.39; Mata’s estimate was -.54. The firm fixed effects were highly significant.
An IV fixed effect model was also fitted. Here again, very poor model results
Obtained, and additionally, the coefficients were very sensitive to the Choice Of
instrumental variable. Polachek and Kim (1994), in modeling household
income in a fixed effect framework, found their results very sensitive to the
choice of the IV variable. The Jamaican data were tOO limited to leverage
these techniques. Great promise is held out for this type Of analysis,
however, as the method deals with several important estimation issues that
arise in panel datasets.

106

time invariant,39 the typical transformations available to correct for some
data problems are not available. As such, the contribution Of this section is tO
search for the added insights Of a panel data OLS model. Equation 4.1

details the specification.

Equation 4.1

lnemployment” =01 + 1n startupsizu + 1n firmageu +
4seasonals” + gender; + agepropridor, + preeducatbn, + posttrainhg, + precredit’

+ postcredit, + location, + parish, + ruralurbar;

The model results are in Table 4.1 below. As above, two models are
presented, a short model which only includes firm size, firm age and seasonal
variables as explanatory variables, and an extended model taking into
account firm and proprietor characteristics. A logarithmic expansion of the
growth function was estimated for both models. The extended model
improves the model fit slightly, with an R squared Of .48 compared to .40.
For both models, the Huber correction for heteroskedasticity accounting for

clustering of firms is applied.

 

39 Firm age is the one exception, and this is dealt with elsewhere in the text.

107

Table 4.1

Five Year OLS Employment Growth Model

 

Model 1
=1684
379 firms

Model 2
N=1643
370 firms

 

   

 

Coefficient

   

P value

Coefficient

 

        

P value

  

 

 

 

 

 

 

 

 

 

 

 

 

 

Depndenet _ F In emplmoyent Ln emplomntye “I
Variable

Startup size .6962719 19.710 .6170117 15.911
Firm age -.0128839 -0.424 .0661368 1.969
Gender of - - -.1574036 -3.697
Proprietor

Age of - -

Proprietor

<20 * It
20-24 .268101 2450
25-29 .2608241 2.386
30-34 .4158768 3.722
35-39 .3105838 3.072
40.49 .18921 12 1.759
50-59 .16435 1.352
60-69 .1355531 1.143
70+ .0790258 0547
Educational - - -.0263523 0603
Level

Post Ed Busin. - - .0750184 0.997
Training

Startup Credit - - .0133315 0.155
Post Startup - - .1135461 1.467
Credit

Parish - -

Kingston * "
St. Andrew .0866349 0.807
St. Thomas .0163732 0.085
Portland .0857377 0.461
St. Mary -.1034705 -0.806
St. Ann -.0224107 -O.193
Trelawny -. 1 747776 -1 .045
St. James .003377 0.033
Hanover -.187618 -1.822
Westmoreland -. 1207025 -1 .064
St. Elizabeth -.2415492 -2.513
Manchester -.0462043 -0.416
Clarendon -.0403963 -0.346
St.Catherine -.0723765 -0.664

 

 

 

108

 

 

 

 

 

 

 

 

 

 

Firm Location - -
Private Home "' "'
Private Yard .162951 2.577
Leased Land .3 825079 3.323
Open Land .3622509 3.220
Roadside .1441 124 1.921
Commercial Bld .1913614 4.186
Market .071 1994 0.812
School Gate .286635 1.423
Other .2335346 2.203
Rural/Urban - - -.085559l -1.869
Seasonals .0760083 2.405 .0590485 1.831
.0736933 2.363 .0511123 1.589
-.040665 -1 . 195 -.0692273 -2.000
-.0595526 —1.588 -.0888224 -2.3 19
Constant -.0529923
| R squared 0.3970 0.4812 1
F value 81.07 24.95
I (0.0000) (0.0000) I

 

 

 

7 Source: STATIN Survey Data
a. Figures in parentheses are the t probabilities
Overall, these results are congruent with the cross sectional analysis.
First of all, the inverse relationship between firm growth and firm size is
confirmed ([3<1). Similarly, the coefficients on the age of the proprietor are
jointly significant and positive, with individually significant coefficients for
the younger cohorts. The results for gender, firm location, and parish are all
significant, and the education of the proprietor and access to firm credit
continue to lack explanatory power in the model. In the broad stroke, this
model confirms the majority of the findings from the log employment in
Chapter III.
Three differences appear in these results, however. First, in the short

model firm age is insignificant, but in the extended model, the relationship

109

with firm growth is "weakly significant"40 and marginally positive. The
result on firm age has been inconsistent across the models, ranging from
highly insignificant and negative to significant and positive. The relationship
between firm age and growth for Jamaica remains somewhat unclear.

Second, the seasonal dummies are significant across the five year
period. This reflects the significance for microenterprises of the deviations
taking place from year to year in the Jamaican economy, and confirms the
importance of the fluctuations observed in the year to year analysis of
Chapter II. In themselves, the dummies don’t offer any actionable insights,
but does point to the sensitivity of these micro firms to seasonal effects.

Finally, the rural or urban location of the firm obtains weak
significance as well, with rural firms negatively related to firm growth. In
contrast, the cross sectional analysis revealed a rural/urban result
consistently and unequivocally insignificant. Again, by incorporating the
entire time frame into the analysis this subtle effect on firm growth becomes
apparent.

In summary, this panel model did not provide inherently different
results from the cross sectional analysis. The additional "weakly significant"
results hardly justify in itself the added expense and effort involved in
collecting this type of data. But the data specification adopted in this chapter
mirrors the cross sectional analysis (with the exception of the seasonal term).

As such, it does not tap into some of the important benefits that panel data

 

40 This phrase is used by Arrighetti (1995)

110

have to offer. In the next section, the model specification will be changed to
include a lagged dependent variable, and introduce new estimation

techniques to provide better insights into dynamic issues.

4.3 Long Term Employment Growth Revisited: A Panel Data
Lagged Dependent Variable Approach

4.3.1 Introduction

Firm dynamics have been considered extensively above, mostly from
the vantage point of how a firm changes and reacts to change over time. In
this section, firm dynamics will be modeled explicitly, by the inclusion of a
lagged dependent variable. Further, the models in the next two sections use
Instrumental Variable techniques in estimation, to control for measurement
error in the regressor and thereby obtain consistent and efficient estimates of
the OLS coefficients.

The use of lagged employment represents a unique approach in the
investigation of firm growth. Section 4.3.2 reviews the results, followed by
conclusions in section 4.3.4. In Section 4.4, a similar approach will be used to

unpack short term dynamics.

111

4.3.2 Five Year Dynamic Employment Model: Results

These data are unique as they span a period of five years, and further,
they provide detailed insights into firm dynamics over the period as they
track employment on the same group of firms. The model specifications
adopted here are similar to the those above, with two exceptions: firm
startup size is dropped and a lagged dependent variable is added to capture
the year to year dynamic effect. Equation 4.2 details the specification, and

Table 4.2 reviews the results:

Equation 4.2

In employment” =0: +1nemployment,’,_, + 1n firmage” +
Sseasonals, + gender; + agepropriaor, + preeducatbn, +

positrainhg, + precrediti + postcredit, + location, +

parish, + ruralurban

Two models are presented below. The first column summarizes an OLS
lagged dependent variable model. The second column review an OLS IV
model, using an additional lag on employment as an instrument (Hsiao, 1986;
Hall, 1987). Based upon the choice of instrument, the analyzable sample size
shrinks to 869 observations and 339 firms. Both models are constrained to
this sample size so the results can be more appropriately compared. The
findings are not affected by the decreased sample size, however. This is

demonstrated in an alternative OLS model which does not constrain the

112

sample (see results in Appendix B). Finally, findings are only presented for a
model excluding the firm age nonlinear and interaction terms. These
variables were insignificant when added to the model, and hence dropped

from consideration (see Appendix B).

113

Table 4.2

Five Year ln ln Dynamic Employment Model:
Jamaican Microenterprise, 1990-1994

 

 

 

 

  

 

 

 

 

 

 

 

 

 

 

Coefficients OLS OLS 1V
Model 1 Model 2
=869 n=869
338 firms 338 firms
Dependent In employment In employment
Variable
__—d — flan?“ 804T _ —
employment (35.638) (27.899)
(t-l)
Ln Firm Age .0245622 .0219542
(1.145) (1.048)
Gender of -.043786 -.0369045
Prop ('1-830) (-1.551)
Age of
Proprietor
<20 * #
20-24 -.4537308 -.4556488
(-6.963) (-7.094)
-.6665265 -.6732817
25 29 (-9542) (-9545)
-.5352847 -.5449269
30 34 (-7.946) (-7.932)
-.5706178 -.5787424
- 9
35 3 (-9.308) (-9.306)
-.6191627 - 624691
-49 ‘
40 (-10.737) (-10.883)
-.6587063 -.6616356
50-59
(-9.687) (~9.869)
-.6600146 -.6613702
60 69 (-9723) (-9.920)
70+ -.6398738 -.639256
(-8.384) (-8.504)
Level of -.009941 -.01 10496
Education (-O.383) (0435)
Post _ .0004343 -.0015957
Education (0.013) (-0-043)
Business
Training
Startup .0498556 .0450554
Credit (1.294) (1.214)
Post-Startup .0471863 .0428418
Credit (1.378) (1.273)

 

114

 

 

 

Parish

a:

*

 

 

 

Kingston
.0770678 .0728824
St. Andrew (1347) (1.287)
-.0621 149 -.0666472
St. Thomas ('0-706) (-0.783)
.0755952 .0740498
Portland (0-625) (0.616)
-.1041558 -.1084765
St. Mary (-1 -.605) (- l .699)
-.0206224 -.0246243
St. Ann (-0.341) (-0.41 1)
-.2013154 -.196786
Trelawny ('2-405) (-2.37 I)
.003109 .0033352
St. James (0051) (0.056)
-.0588212 -.0564228
Hanover ('0-971) (-0.956)
-.0973743 -.0970599
Westmoreland ('1 ~44 1 ) (-1 .463)
-.079697 -.0792561
St. Elizabeth (-I .547) (-1.573)
-.0006133 .0000972
Manchester ('0-010) (0.002)
.0023012 .0009705
Clarendon (003 7) (0.016)
-.0422042 -.046546
St.Catherine ('0-744) (-0.830)
Location of
Business .. *
:3
(2.116) (2.009)
-.0008855 -.0107978
Leased Land 00.015) (-0194)
-. 1308237 -. 1365694
Open Land (4018) (-1 .020)
. .1321281 .131509
Roadsrde (2.222) (2.243)
- .0698569 .0622446
Commerc1a1 Bld (2.485) (2220)
.0655286 .062072
Mark“ (1.487) (1.439)
.064718 .0570277
School Gate (1.229) (1.160)
.1134357 .1099734
Other (1.443) (1.426)
Rural or .0076436 .01 18417
Urban (0290) (0.469)

 

 

 

115

 

 

 

 

 

 

 

Seasonals - -
.1 147208 . 1059489
(1.580) (1.443)
-.0007968 -.0094401
(-0.010) (-0.121)
.0455494 .0392137
(0603) (0.517)
Constant .5675724 .5764395
(5.227) (5.304)

[V Variabe Used

    

Employment
(t-Z)

 

 

 

F Value 129.88 215.57
(0.0000) (0.0000)

 

 

 

 

Source; STATIN SURVEY DATA
Note: All values in parentheses are 1 values.

Overall, the model performed well, with an R squared of .72 for both
models, and an F ranging from 129.88 to 215.57, significant at the 99% level.
This is an improved fit compared to the cross sectional model, and the IV
result reflects a slightly improved fit over the OLS model. The Huber
correction adjusting for clusters of firms was used in both models. The model
results were consistent and stable across different sample schemes and
specifications. This contrasts with the cross sectional model which was
sensitive to changing specification, time and variable definitions.

The IV estimation was introduced to correct for the presence of
measurement error, but the results reveal little evidence of bias due to

measurement error or the deleterious effects of introducing a lagged

116

dependent variable. Between the models, the shift in the magnitude of the
coefficients is minimal, and in only one case does the sign change (education).
In that case, however, the coefficient is insignificant and close to zero.
Measurement error plays a very small role in these results. Given that
measurement error poses so many problems in panel data (Ashenfelter, 1986;
Hsiao, 1986), this result validates the data and enumeration methods used.

The overall findings reveal some common ground and important
differences to the cross sectional OLS model. Similar to the models of section
4.2, significant results obtain for firm size (lagged employment), gender of the
proprietor, age of the proprietor and location. However, these five year data
also reveal several differences with the earlier models, both in the magnitude
of some of the coefficients as well as in the significance of some additional
variables.

To begin with, the dynamic relationship between lagged employment
and employment follows the inverse pattern observed in the previous model
between startup size and employment. The coefficient on lagged employment
is less than one and significant in both models, indicating a clear negative
relationship between lagged employment and growth. The IV specification
does not change the coefficient appreciably. Measurement error, therefore,
does not play a large role in coefficient bias for startup size. The negative
relationship between prior year firm size and growth can be interpreted as an

additional refutation of Gibrat’s law. Even over a short period of time, firm

117

size and growth follow the negative relationship posited by Jovanovic and
others.

Second, the coefficient on firm age, similar to the cross sectional model,
is not significant and is positive. As discussed above, this is contrary to the
evidence from several other countries and the theoretical model of Jovanovic.
Further the non-linear relationship between firm age and growth is refuted
in these results (Appendix 11). However, the categorical variable for the age
of the proprietor is jointly significant (F=16.58, P>F = .0000).41 In both
models, the coefficients for age of the proprietor have a negative sign with a
trend toward an increasing negative magnitude. In other words, firms with
older proprietors grow slower, but older firms don't necessarily grow slower.42
Although the trend in the coefficients between the youngest to oldest
proprietors doesn't follow a linear or even consistent trend, the youngest
proprietors do consistently grow faster than the oldest (-.45 to -.63, youngest
to oldest respectively).

A third result regards firm credit, and is laden with caveats. The
result is found in Appendix II, where the full sample OLS models are

reviewed. In these models, there is a positive relationship between

 

41 This F value is for the OLS IV model.

42 In another approach to unpack the firm age finding, a fixed effect model
was run with just firm age as the explanatory variable. In a fixed effect
model, firm heterogeneity is accounted for in the firm fixed effects, implying
these effects will not be confounded with firm age. The model results
revealed a negative but insignificant coefficient on firm age and firm age
squared, highly significant fixed effects, but a very poor model fit (low R
squared and F statistic). This model did not bring resolution to this issue.

118

employment growth and credit, both for startup and post-startup credit.
Importantly, this model reports a significant relationship between startup
credit and firm growth (t value = 2.205), and a nearly significant relationship
between post-startup credit and growth (t value = 1.767). These results,
however, become insignificant in the IV model. Disappointingly, the driver of
the insignificant IV result cannot be fully understood, as it can be due to
declining sample size or to measurement error corrected by the IV
estimation. This finding, however weak, has significant implications for GO
or NGO programs which desire to influence the growth of microenterprises.
Fourth, the location of the firm is an important determinant of firm
growth. Both the physical location of the place of business and the parish the
business operated in is significantly related to firm growth. The parish
variable is jointly significant ( F = 1.87, P>F = 0.0403), and only two
coefficients on parish obtain individual significance. This might reflect the
differing levels of macroeconomic growth occurring in the different localities.
Regarding the physical location of the firm, firms in more established
locations (commercial buildings and markets) have a greater positive effect
on firm growth.43 Very significantly, the coefficient is positive and significant
for commercial buildings. In the IV model, a negative relationship obtains for
firms located on open or leased land. The location of a firm between a rural

or urban area has no effect on growth. Fisseha’s (1993) 12 year retrospective

 

43 The F statistic for firm location is F=1.89 for the OLS IV model.

119

dynamic study reveals little difference between rural and urban firm growth
rates as well, supporting these results.

As a fifth finding, the gender of the proprietor matters in this
representation as well. The coefficient on gender is highly significant, and
the negative sign on the coefiicient confirms the negative relationship
between female proprietors and firm growth. This result finds support in
Hotchkiss and Moore (1996), who report on earning differentials in the formal
sector in Jamaica. In particular, they report that compensation differentials
range between 49% to as little as 2% to that of their male counterparts,
depending on the occupation. They further report that these differentials are
not the result of different job market characteristics between men and
women, but that the market simply treats men and women differently when
it comes to compensation.

Sixth, a surprising result thus far is the lack of significance of
education or business training on firm growth. Polachek and Kim (1994)
explored this phenomena in a different context. For a fixed effect model,
these authors suggest that education or training might influence the slope
and not the intercept. Specifically, more motivated workers would have
steeper earning slopes than their counterparts. What they found, however,
was that a fixed effects model with time varying slopes did not do a better job
than a model with just shifting intercepts. This leads them believe that

motivation impacts the proprietor early in their life in their choice of

120

education and training, but the education does not impact them continuously
through time. This provides a plausible explanation for the lack of
significance on these variables.

A final issue not addressed explicitly in any of the models above is the
persistence of past growth on current growth. This is a corollary of Gibrat’s
law, and has been examined in several papers, most notably perhaps by
Chescher (1979). The significance of lagged employment as an explanatory
variable suggests but does not confirm the persistence of growth. To asses
this, an OLS model in first differences was run, which in effect redefines the
dependent variable explicitly as growth, and the exogenous variable of lagged
employment as lagged growth. Tested in this way, verification of the
persistence of growth also tests for presence of serial correlation in the data
(Dunne, et a1 1994). The OLS first difference model provides the additional
advantage of sweeping out measurement error and the effects of firm
heterogeneity.

Similar to the fixed effect model, however, the results of this model
were very weak (R squared of .03) and the F statistic was not significant.
This indicates the hypothesis that the coefficients were equal to zero could
not be rejected. No clear indication of the persistence of growth or of serial
correlation can be confirmed through this model. The conclusion supported

here is that past growth is not a good indicator of future growth.

121

4.4 Short Term Employment Growth Revisited: A Panel Data
Lagged Dependent Variable Approach

4.4.1 Introduction

The last section focused on long run dynamics. Do the relationships
identified in the long run, however, pertain in the short run? Do different
factors afi'ect the growth pattern? The analysis of this section focuses on
short run dynamics, examining firm growth on a quarter to quarter basis
over the two year time period between 1993 and 1994 . The treatment here is
unique in that roughly the same group of firms included in the long run
analysis are considered here in the short run.

The same model specification and estimation techniques used to
examine long run growth are adopted here. Section 4.4.2 examines results,

followed by conclusions in section 4.4.3.

4.4.2 Quarterly Employment Panel Data Model: Results

The model adopted in this section follows the lagged dependent
variable specification of the last chapter; the data definitions are exactly the
same as well. OLS is employed to estimate the model in equation 4.3. The
Huber correction adjusts for heteroskedasticity. The complete specification

follows:

122

Equation 4.3

lnemploymeng', = on + 1n employment,“l + 1n firmageu +
6seasonal; + gendet; + agepropriaor; + preeducatbn, +
positrainizg, + precredii + postcredi; + location +

parish + ruralurban

Each of the hypotheses dealt with in the above chapter will be treated in
turn. As above, the model excludes firm age interaction terms, but these
results are available in Appendix III. Instrumental variable techniques are
employed in both models to control for measurement error. Table 4.3

summarizes the results.

123

Table 4.3
Quarterly ln ln Dynamic Employment Model:
Jamaican Microenterprise, 1993-1994

 

 

       

 

 

 

 

 

 

 

Coefficients OLS OLS IV
(A) (B)
n=650 n=650
266 Firms 266 Firms
Dependent 1n employment In employment
Variable
Ln lagged if i _ — _ L ”
(t-l)
Ln Firm Age .0423203 .0210071
(2.085) (1.161)
Gender of -.0791643 -.046219
Prop 02.957) (-1.933)
Age of
Proprietor . .
<20
2024 .1250795 .0858123
(1.668) (1.286)
2549 .0629241 .0533443
(0912) (0.867)
3034 .0787337 .0346441
(1.158) (0.584)
_ .1092747 .07161
35 39 (1.484) (1.111)
_ .0163726 .0141279
40 49 (0.234) (0.220)
50.59 .0008491 .0029977
(0.012) (0.051)
.0090302 .0239949
- 9
6O 6 (0.116) (0.341)
70+ -.1307186 -.0978818
(-1.456) (-1257)
Level of .0297357 .0239982
Education (1 ~01 1) (0.919)
Post — .0055823 -.0200684
Education (0.125) (0560)
Business
Training
Startup .037407 .0159523
Credit (0.731) (0.402)
Post-Startup .0603361 .0179819
Credit (1.240) (0.458)

 

 

 

 

 

124

 

Parish

*

*

 

 

 

Kingston
.0354801 -.0195355
St. Andrew (0.450) (-O.281)
.0436889 .0400882
St. Thomas (0510) (0.562)
.1 123851 .0533679
Portland (0.660) (0.309)
.1467945 .1192002
St. Mary (1.131) (0.907)
-.0335053 -.0317658
St. Ann (-0.427) (0.435)
.1604453 .120242
Trelawny (1375) (1.081)
-.1586495 -.1419171
St. James (-1 290) (~1-089)
-.1751188 -.1341456
Hanover ('1 370) (—1 .449)
-. 1291804 -. 1029261
Westmoreland ('1 585) ('1-416)
-.0856099 -.0731962
St. Elizabeth (4211) (-1.145)
-.0453301 -.0630914
Manchester (-0. 585) (-0.934)
-.0024488 -.0068074
Clarendon (43.030) (-0.093)
-.0159559 -.0257596
St.Catherine ('0-213) (-0.364)
Location
Private Home l '
Private Yard .0615752 .0310576
(1.573) (0.900)
Leased Land -.1196253 -.1542618
(-1.090) (0.480)
Open Land -.0274377 -.0569805
(-0.383) (-0.914)
Roadside .0021048 -.0175273
(0.039) (-0.335)
Commercial Bld .0593655 .0415204
(2.011) (1.667)
Market .1727609 .1505956
(1.583) (1.815)
School Gate .0090017 .01281 19
(0.197) (0.381)
Other .0704684 .0637825
(1.445) (1.860)
Rural or -.0210042 -.0137706
Urban (-0.718) (-0.553)

 

 

 

125

 

 

 

 

 

Seasonals - -
.041549 0.338299
(0.957) (0.737)
.0542327 .0539587
(1.291) (1.236)
.0001543 -.0044954
(0.004) (0117)
.0731273 .0726132
(1.929) (1.803)
Constant -.0276314 -. 0231 142
(-0247) (-0220)

 

   

‘ Instrumental Variable

 

 

 

 

 

 

Lagged NO YES

Employment

(t-2)

R squared 0.7122 0.6965

F Value 76.18 129.23
(0.0000) (0.0000)

 

 

Source: STATIN Survey Data
All values in parentheses are t values.

The quarterly dynamic model exhibits a good fit, and many of the
factors driving long run dynamics appear to affect the short run as well. But
first, the R squared varied between .69 and .71, with F values between 76
and 129. The IV variable used here is y m2, which satisfies the requirements
of an IV (Hall, 1987; Hsiao, 1986). The IV specification improves the fit
slightly, and results in slight shifts in all of the coefficients. This suggests
that measurement error does play a role in coefficient bias. In other words,
for these short run results, correcting for measurement error is an important
step in obtaining consistent estimates. In the models from the previous
section, the IV specification improved the fit for the OLS model only slightly,

and resulted in just minor shifts in some of the coefficients.

l26

The first set of comments will focus on quarterly dynamics, followed by
a discussion of the relationship to the long run findings. The first result is
regarding the relationship between firm size and growth. The coefficient on
lagged employment is less than one, meaning firm size in the previous
quarter is inversely related to quarterly employment growth. The magnitude
of the coefficient is very much in line with the other results. Indicative of a
correction for measurement error, the coefficient for the IV model shifts closer
to 1, to .88. The IV variable makes a sizeable adjustment to the parameter
estimate on lagged employment, much more so than in the five year model.
Assuming the IV estimator is behaving properly, this suggests the presence
of more measurement error in the short run data.44 In summary, the inverse
relationship on lagged employment points to the importance of small firms in
affecting short term firm growth. The policy implication is clear: the smallest
firms deserve attention, if not special attention.

This result suggests that Gibrat's law is invalid, even for the short run,
and builds more support for Jovanovic's 1982 theory. Beyond the theoretical
explanation of this result provided by Jovanovic, however, sample attrition
may contribute to this result (although it cannot be confirmed). As this
sample included firms with between 1 — 10 total employees (including sole-
proprietorships as well), firms that quickly grew beyond the size of 10 are

excluded as well as firms that failed. Since the sample frame dates back to

 

44 Appendix III reviews the full sample model. In these results, the small
shifts in the coefficients between models 4a and 4b, and between 4d and 4e,

127

1990, the possibility exists that the firms left in the sample were the smallest
firms that are still growing and the larger microenterprises that have hit
their efficient size. As there is no data for firms larger than 10 employees,
there is no way to gauge the extent of this potential bias.

The second result regards firm age and age of the proprietor. First, the
OLS results reveal a positive weakly significant relationship between firm
age and growth. Correcting for measurement error, the IV results negate
this finding. The age of the proprietor, however, is once again significant,
with coefficients gravitating from slightly positive to negative for the oldest
proprietors.45 For microenterprises, this result has intuitive appeal, as the
proprietor takes a very visible and controlling role in the firm. This has an
important efficiency implication as well. Firm efficiencies are not necessarily
passed down fiom generation to generation or owner to owner, but must be
learned new by each proprietor. This supports Cressy (1996), who
incorporates the age of the proprietor (not the firm) in his model of firm
growth. Different from the five year results, however, the proprietors in the
middle of the age distribution grow faster than the youngest and oldest
proprietors. One reason for this difference is that proprietors with some
experience are able to better manage the short run fluctuations and shocks

which apparently heavily effect these micro firms (see Chapter II for details

 

indicates that the shifting sample plays little role on the estimation.
45 The F statistic for joint significance of the age of the proprietor is F=2.30.

128

A third and final finding comes from a result in Appendix III and
regards the firms access to credit. In a full sample OLS model (column 4d)
without firm age interaction, the coefficient on post-startup credit is positive ‘
and weakly significant (t value = 1.791). This finding suggests that in the
short run life cycle, access to operations credit positively influences firm
growth, which finds support in Otero and Rhyne (1994), who posit that short
run operating needs are the most important for microenterprises. The
results here are only weakly significant and do not distinguish between
operating loans and capital expenditures. The significance once again points
to the potential importance of access to credit for firm growth.

Regarding long and short run dynamics, the environment appears
remarkably consistent, particularly good news for policy makers! First of all,
the magnitude and sign on the coefficients were generally consistent across
the long and short run models. In the short run model, the coefficient on firm
size was slightly closer to one; the coefficient on firm age was insignificant;
and the other human capital and firm demographic variables were consistent
as well. One interesting difference relates to the positive relationship
between the age of the proprietor and short run growth; this contrasts with
the long run result where the relationship was negative. Although both sets
of coefficients trend the same direction, this difference may reflect the
different human capital dynamic between the short and long run. This

variable in the long run reflects firm learning; in the short run, the variable

129

may just reflect proprietor energy or enthusiasm. Other than this one
exception, the long and short run dynamics reveal remarkably similar

patterns overall.

4.5 Attrition Bias

Several forms of estimation bias have been dealt with in the text.
These include measurement error, inconsistent and inefficient estimates due
to a lagged dependent variable, hetereoskedasticity, and serial correlation.
Before discussing conclusions, however, the very important issue of attrition
bias and the validity of the sample needs to be considered.

Attrition is a common problem with panel data, and there are many
reasons why firms may drop out of the sample. With a panel, unlike a cross
sectional study, respondents might grow tired of answering the enumerators
question(s) on an ongoing basis, the household or firm might change location,
or if large distances need to he traveled to interview the respondent, only one
chance might be allowed for the interview.

For the Jamaican sample, several of these issues are salient. In
particular, respondent fatigue was an issue, as this particular sample had
been originally formed and interviewed for the 1990 microenterprise census.
In 1992, the in-depth qualitative National Microenterprise survey was

conducted, and then in 1993 the Quarterly Panel Survey began. Despite the

130

fact that these panelist were on their third survey, however, very few
interviews came back as outright refusals. This is consistent with
Ashenfelter’s (1986) comments that respondent fatigue is seldom an issue in
developing countries. Other more significant concerns for this enumeration
are missed interviews due to long distances traveled for interviews, bad
weather delays, firm closures, and heavy work loads on the part of
enumerators, forcing them to sacrifice some of their field visits.

To test whether these sources of attrition bias impacted results, a basic
model of employment growth was estimated using OLS and OLS IV.
Following Hall’s (1987) approach to attrition bias, the employment growth
model was run on the subset of firms that were in the sample either 6-7
quarters, 5-7 quarters, 4-7 quarters, and 3-7 quarters respectively. In other
words, the sample is allowed to shift based upon which firms met the sample
criteria. If the results are consistent across definitions, then the conclusion of
little estimation bias due to sample effect obtains. The model is evaluated in

levels as well as with a log log transformation. The model is as follows:

Equation 4.4

employment = f(constant, lagged employment)

A couple of issues to note on this model: one, why this model? This

model specification is consistent with much of the literature on small

131

business growth (Audretsch, et a1 (1994), Dunne, et al (1994), and Mata
(1994)) In the literature, firm growth, often measured‘in firm assets, is
modeled as a function of firm size, startup size, or as a lagged variable of
growth itself. Since the objective here is merely to examine attrition bias, the
simplest specification was sought. Secondly, firm age is dropped as it was
insignificant in most of the modeled results presented above. Finally, both
OLS and OLS IV results are presented to parallel the main analysis above.

Tables 4.5 and 4.6 summarize the results.

I32

Table 4.4
Evaluation of Attrition Bias in a Panel Data Model of
Employment Growth: in Levels for Jamaican Microenterprises

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variables Number of Quarters Firm in Sample
6-7 Quarters 5-7 Quarters 4-7 Quarters 3—7 Quarters
n=408 n=558 n=656 ¥ n=697
Coeff I T Coeff ] T Coeff l T Coeff I T
Model I: OLS
Intercept 0.35 6.41 0.35 7.87 0.33 8.08 0.086 6.28
EMPLAG 0.79 36.37 0.79 43.12 0.81 50.23 0.81 48.57
Model F 1322.58 1859.25 2523.37 2358.99
(0.0001) (0.0001) (0.0001) (0.0001)
R-sq 0.71 0.70 0.71 0.68
Model 2: OLS with Instrumental Variables
Intercept .20 1.826 .18 1.950 .14 1.843 .15 1.934
EMPLAG .86 12.939 .88 15.943 .90 20.380 .90 20.112
Model F 167.41 254.19 415.33 404.49
(0.0000) (0 . 0000) (0.0000) (0.0000)
R - sq 0.70 0.70 0.71 0.70

 

 

 

 

 

Source: STATIN Survey Data

133

 

Table 4.5

Evaluation of Attrition Bias in a Panel Model of Employment
Growth: ln ln form Jamaican Microenterprises

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variables Number of Quarters Firm in Sample
6-7 Quarters 5-7 Quarters 4-7 Quarters 3-7 Quarters
Coeff I T Coeff I T Coeff I T Coeff I T
Model I: OLS (In In form)
dep var: In of employ
Intercept 0.07 3.86 0.08 5.39 0.082 5.84 0.096 6.14
LEMPLAG 0.80 34.21 0.81 40.29 0.81 45.64 0.80 41.95
Model F 1 170.09 1623.42 2082.88 1759.84
(0.0001) (0.0001) (0.0001) (0.0001)
R—sq 0.69 0.67 0.68 0.61
Model 2: OLS with Instrumental Variables
Intercept .026 1.254 .025 1.431 .027 1.555 .035 2.045
LEMPLAG .90 24.129 .91 29.310 .92 34.341 .92 34.227
Model F 582.19 859.07 1179.30 1171.52
R-sq 0.68 0.68 0 . 6 8 0.68

 

 

 

 

 

 

Source: STATIN Survey Data

These results reveal that firms moving in and out of the sample for all
of the reasons discussed above are not appreciably affecting the value on the
coefficient. Specifically, the coefficient on the employment lag maintains a
consistent value as firms with fewer quarterly representations were added.
Further, both the OLS and OLS IV model results in the same conclusion. For

example, in Table 4.5, the coefficients on EMPLAG (lagged employment) are

.79, .79, .81 and .81 for firms in 6-7 quarters, 5-7 quarters, 4-7 quarters, and

134

 

3-7 quarters respectively. In the OLS IV model, the coefficients are .86, .88,
.90 and .90. Further the model in In In form yields the most consistent
results, as one would expect (.80 - .81). If consistency across each model
dissolves the specter of attrition bias, then these results appear to bear that
out. Interestingly, several articles on small business panel data found little
evidence of attrition bias impacting their data (Hall, 1987; Evans, 1987;
Dunne and Hughes, 1994; Mata, 1994; and Wagner, 1994). Dunne and
Hughes (1994), for example, used a MLE and a probit model to account for
firms that died. Attrition bias still did not influence the OLS regression

results.

4.6 Conclusions

This chapter extended the analysis of the previous sections by
developing a dynamic panel data model of microenterprise firm growth. As
such, this chapter represents new and unique perspectives on
microenterprise dynamics. The dynamic models of sections 4.3 and 4.4
proved to be stable model specifications with good predictive power. Further,
the IV estimation introduced here revealed little evidence of measurement
error in the five year results, but contributed to an appreciable adjustment in

the short run model. This representation of microenterprise growth

135

definitely enhanced the simpler cross sectional specification presented
earlier.

The five year findings are consistent with the previous results. First,
firm size has an inverse relationship with firm growth, meaning the smallest
firms have the fastest growth. Firm size, in this context is defined as prior
year firm size. Further, female proprietors have a negative relationship with
firm growth, and firms in business districts or formal commercial buildings
effect firm growth positively. Regarding firm age, the variable is
insignificant across all of the OLS models, OLS IV, and the fixed effect model
as well. Age of the proprietor, however, is jointly significant, with several
individually significant age coefficients and a trend in the coefficient towards
a stronger negative relationship with greater age.

The quarterly analysis of firm growth reveals several things: One,
long and short term dynamic effects are fairly consistent. This is great news
for policy makers. Designing programs to address specific short term issues
along the lines discussed above should have a consistent carry through to
long term effects. Secondly, the inverse relationship between firm size and
quarterly firm growth holds. Third, the age of proprietor is inversely related
to firm growth, whereas age of the firm is not. Finally, measurement error
plays a more significant role in the short run results than the long run. This
may reflect the increased volatility of quarterly data versus the its'

counterpart.

136

CHAPTER V

CONCLUSIONS:
IMPLICATIONS FOR POLICY AND FUTURE RESEARC H

To the policy maker and the researcher, microenterprise present a
challenge. Both in the developing and in the developed world, the
significance of their existence and role in the broader economy has been
recognized. However, by definition of who they are and the complex interplay
between all the elements, interests, and needs encapsulated in their
existence, a comprehensive understanding of them has remained elusive. In
recent years, a significant amount of research effort has shed light on some of
the stylized facts of these firms, their prolific presence in most LDC
economies, and their birth and survival. Further, the sector has received
increasing attention from both donor agencies and policy makers. Most
recently, a concerted effort has been made to gain an understanding of the
mechanisms of growth and change for these firms. This dissertation has
sought to contribute to a deeper understanding of these special firm
dynamics.

Jamaica provides an invaluable setting for the study of
microenterprise dynamics. First, the micro sector is an important part of
Jamaica's national economy. The size of the sector has been documented on

two occasions by the Statistical Institute of Jamaica. In 1978, a nation-wide

I37

census estimated a total of 38,000 microenterprises employing roughly 80,000
people. A 1990 census revealed an increase to over 88,000 enterprises
employing over 150,000. This amounts to 15% of the Jamaican labor force.46
Second, the analysis can reveal the dynamics of microenterprise in the
context of macroeconomic crisis. The time period of this analysis
encompassed an economic downturn, including a currency devaluation,
volatility in interest rates, and soaring inflation. Finally, a significant
amount of research has been done in Jamaica to understand the role and size
of the micro sector. These include the two census surveys mentioned above,
several smaller surveys conducted in the late 1970's (Davies, Fisseha, and
Kirton, 1979; Fisseha and Davies, 1981; Fisseha, 1982), a small
establishment survey in 1983 by STATIN (Small Establishment Survey,
1983), and a national survey of microenterprises in 1992 (Anderson, 1994).
The static characteristics of the microenterprise sector in Jamaica
have been defined from this series of studies, and the findings reveal a sector
similar to that of other countries. First, they are dominated by own account
firms, with roughly seventy three percent falling into this category.47 Twenty
two percent of the firms employed between 1-4 employees, while only five
percent employed between 5-9. The average microenterprise employed 1.7

workers. Consistent with the Jamaican phrase "female is to small as male is

 

45 STATIN reports that the labor force in 1990 was 1.06 million.
47 Microenterprises are defined in this work as firms employing less than 10
persons.

I38

to large,”8 female headed firms were more likely to be own account than
their male headed counterparts, although female headed firms represent
almost equal numbers in the microenterprise sector. Jamaican firms are
distributed with a slightly higher urban bias than most countries, with
roughly fifty-five percent of the Jamaican firms located in urban settings.
Based upon the 1992 National Survey, twenty four percent of the firms were
less than four years old, another 41 percent were between eleven and four
years of age, and the remaining thirty-four percent were older than 11 years.
Reflecting the large role of tourism in the local economy, sixty six percent of
the firms were engaged in trade and commerce. Although manufacturing is
important in Jamaica, the preponderance of trade activity diverges from the
experience of most other developing countries, where manufacturing is the
dominant industry.

This dissertation builds on a unique set of information collected fi‘om a
panel of microenterprises in Jamaica.49 This represents the first national
effort to capture detailed information on microenterprises over an extended
time period. The enumeration included annual data collection over five years
of the same firms, with an additional intensive quarterly collection over a two
year period. Even with this extensive data collection, these data are not
without limitations. First, firm birth and death data were not collected.

Second, the survey instrument proved to be somewhat complex in the method

 

48 Quote found in Anderson, 1994.

139

used to collect sales and output data. Finally, the attempt to capture annual
information on capital accumulation was unsuccessful.

Beyond these shortcomings, these data provide a unique opportunity to
shed light on questions surrounding firm dynamics. As the data were non-
retrospective, they did not suffer from bias due to loss of memory.50 Further,
tracking the same set of firms over time (panel data set) allowed a careful
documentation of firm level changes over the life cycle. As such, this analysis
builds on several important research questions relating to firm growth, as
well as provides valuable insight to the Jamaican government and donor
agencies seeking to better channel assistance to the sector.

Two unique approaches to this data were taken in this dissertation.
First, a detailed year-to-year and quarter-to-quarter analysis of secular
trends for key measures of growth performance outlined the dynamic ebb and
flow of this sector. This depiction of changes in the firms over time revealed
important stylized facts of the volatile environment characterizing these
firms over the long and short run. Second, a model of firm growth was
estimated using a lagged dependent variable and IV (instrumental variable)
techniques to control for measurement error. This model extends the
traditional cross sectional approach to growth models (presented in Chapter

III) by incorporating a true dynamic effect into the model specification.

 

49 Specifically, the data incorporates firms identified in the 1990 census, the
1992 National Survey, and the Quarterly Panel Survey of 1993-1994.

50 All of the employment information was non-retrospective, with the
exception of 1991 employment collected with the 1992 National Survey.

I40

 

Importantly, this model better characterizes the dynamic process driving the
growth of surviving firms. A summary of the major findings follows.

In Chapter II, a descriptive analysis profiled firm performance over a
quarterly two-year and annual five-year time frame along several key firm
dimensions: employment, sales or output, and wages. This technique
generated a unique dynamic picture of how these firms changed, ebbed and
flowed over time, and how the patterns of change varied by sector, gender, E
location and firm size. For the quarterly analysis, firm performance was

summarized by examining the quarter to quarter change in firm employment

 

IE...»- w . ..
-.

and wages for a set of firms "alive" in two consecutive time periods. The
measures were broken down by sector, gender, firm location and size. A
similar approach was applied in the five-year analysis, but only for firm
employment. Several key insights came out of this analysis.

First, this micro sector is very tenacious. The period between 1990 and
1994 was characterized by several negative macroeconomic shocks,
particularly between 1993 and 1994 during the quarterly panel survey. The
effects of the shocks manifested themselves in quarter to quarter and year to
year fluctuations in firm performance with periods of double digit loss. For
example, between quarter I of 1994 and quarter II of 1994, employment fell

by almost 10 percent! Although some of this fluctuation is attributable to

I41

seasonal fluctuations, the extreme variation points to the influence of macro
effects.51

Second, the overall level of employment of the existing microenterprise
in Jamaica declined over the 1993-1994 period. Total employment in the
panel of firms at the fourth quarter of 1994 was 18.6 percent below the level
in the same enterprise in 1993.52 The five year analysis reflected a similar
trend, with the greatest decline in 1993 and 1994.

The overall downward trend in output and employment in the existing
microenterprises in the sample is consistent with the picture of an economy
in stress. This was indeed a difficult period for the Jamaican economy.
Overall, real GDP per capita increased only 0.5 in 1993, and inched up in
1994. One of the main contributions of the panel survey is an indication of
the effect of this stress on an important component of the Jamaican economy.

At the sector level, one of the significant findings has been the
desultory performance of microenterprise in the trade and commerce sector of
Jamaica. By all measures - employment, sales, and wage bill -, it performed
the worst over the survey period. Moreover, this sector was subject to more
quarter to quarter variation than the others. Finally, microenterprises in
the manufacturing sector experienced the smallest quarter to quarter

variation in activity.

 

51 Note as well that these fluctuations were even more exacerbated when
examining real sales or output.

52 Real sales fell by 35.7% for this same panel of firms (Gustafson and
Liedholm, 1995).

142

 

With respect to gender, the most striking finding was the relatively
strong performance of female-owned microenterprises in Jamaica. Although
employment and the real wage bill declined for female headed firms, the
decline in the real wage bill was significantly smaller than that experienced
by their male-owned counterparts. It should be noted, however, there was
greater quarter to quarter variation in employment of the female-owned
microenterprises.

By firm size, it is the largest microenterprises (5-9 employees) that
performed the worst during the period. Not only did their employment and
real wage bill decline more than the other firms sizes, but by the end of 1994,
their level of employment was almost 50 percent lower than at the beginning
of 1993. Those firms also experienced the most volatile swings from quarter
to quarter. By contrast, the own account firms fluctuated the least, and were
the only size category where employment and real wages were higher at the
end than at the beginning of the 1993-94 period.53

Finally, a characteristic of the data were wide quarter to quarter
swings in the level of microenterprise activity in Jamaica. From quarter II to
quarter III, 1993, for example, there was a downturn that was picked up by
all the indicators. These swings were less evident in the five year long run

time frame, and the different sectors moved more in lockstep fashion.

 

53 This result may be partially influenced by the fact that one person firms
cannot decline without closing. Mansfield (1962) suggested this as the
primary driver for the apparent negative relationship between firm size and
growth.

I43

 

Chapter III extended the above analysis to examine firm growth by
means of an OLS cross sectional model. This approach parallels the majority
of research on firm growth, permitting a direct comparison to those results.
Several of the key issues addressed were the validity of Jovanovic's 1982
theoretical model, Gibrat's law, and the relevance of several dimensions of
human and firm capital on firm growth. Different from most other studies,
however, the dependent variable was defined alternately in log levels and as
firm growth to assess the impact of the different variable specifications on the
parameter estimates. Four key findings emerged from this analysis.

First, an examination of the coefficients across the five years reveal
subtle changes in firm dynamics. The result for parish, for example,
fluctuated across the years, with the signs on the coefficients changing fiom
year to year. One explanation for this result posits that microenterprise are
sensitive to macroeconomic and other shocks, and these shocks vary by
region. For future research, careful consideration should be given to the
timing of efforts to examine microenterprises. Findings grounded in a
particular policy or macro environment may not be relevant when change
takes place. For policy makers, factors such as interest rates, inflation and
exchange rate fluctuations do make a difference in the health of the small
firm.

Second, the definition of the dependent variable does make a difference.

With the dependent variable defined as the natural log of firm growth, a

144

 

negative bias is introduced into the model results. This bias is readily
observable for the coefficient on firm size, with an expected slight shift in the
magnitude of the coefficient between the models. The result for firm age was
more dramatic, however, with a change in sign and a general shift in
coefficient significance between the models. Significantly, in the model with
the dependent variable defined as the natural log of firm growth, the firm age
variable is significant and negative. For Jamaica at least, the negative
relationship between firm age and growth is called into question, and more
careful scrutiny of this relationship should be exercised in future studies.
Shifting the focus to the model in log levels, a third result confirms the
negative relationship between firm size and growth, although the
relationship posited for firm age is not supported, either in sign or
significance level. This latter result conflicts with a significant amount of
other empirical work in several countries. In contrast, the age of proprietor is
significant and negative or trending downward. This supports Cressy (1995),
who suggests that age of the proprietor is a good indicator for human capital.
If this measure can be thought of as a proxy for firm learning, then
Jovanovic’s model finds support here as well. This result will be reviewed in
more detail below.
Beyond this, a fourth key finding relates to the difficulties facing female
entrepreneurs in Jamaica. The gender of the proprietor stands out as an

area requiring special policy attention. Unequivocally across model

145

_-_. ..f
7.‘

 

IF

specifications, female headed firms are apparently at a disadvantage to male
headed firms as measured by firm growth in employment. This is consistent
with the analysis of Chapter II which detailed the poor employment
performance of female headed firms. This stylized fact may be further
compounded, for example, by firm location, as female proprietors are more
likely to work out of their homes rather than a commercial center (Anderson,
1994). Consistently in these data, home based businesses were also at a
disadvantage to businesses in commercial centers. The evidence in these
data underscores the difficult challenges facing the female proprietor in
Jamaica.

Chapter IV introduces an important extension to the analysis of
Chapter III. Specifically, a panel data model is developed introducing a true
dynamic effect to the specification (a lagged dependent variable) and IV
(instrumental variable) estimation is utilized to diminish the bias in the
estimated parameters. This model is evaluated for both the five year and two
year quarterly employment data and reveals results broadly consistent with
the cross sectional analysis.

Three central findings stand out. The first relates to the classic
hypotheses between firm size, age and growth. Firm size is negatively
related to firm growth, even when firm size is defined as size in the prior year
or quarter. This is a strong refutation of Gibrat's law. Firm age, by contrast,

is not related to firm growth, although age of the proprietor is. These results

146

 

basically confirm the results from the previous sections, and solidify these
findings for Jamaica.

Second, access to credit is positively related to firm growth. Although
this result did not obtain in the more robust models (the IV specification for
example), the specter of its significance has important and exciting
implications for donor agencies and public policies focused on credit and
lending. The result here is not definitive, but through this dynamic model at
least a "weakly significant" effect can be confirmed.

Third and finally, short and long run growth dynamics appear to be
driven by very similar processes. Evaluating the signs and magnitudes of the
coefficients across the models, the majority obtain very similar results. One
striking difference between the long and short run model relates to age of the
proprietor. Specifically, age of the proprietor switches from a negative to a
positive relationship with firm growth between the long and short run models
respectively. This may be because the variable proxies different effects in the
long and short run. Overall these results imply that firm dynamics reflect
roughly the same process in the long and short run. An interesting and
important area for additional research might be how decisions and risks are
evaluated by the microentrepreneur in both of these times frames as well.

Before discussing policy implications, some observations on the

benefits and pitfalls of panel data are in order.54 Is panel data the panacea

 

54 See Hsiao (1986) or Raj and Baltagi (1993) for a thorough discourse on the
benefits of panel data.

147

for the researcher in lesser developed countries? There are several very
specific benefits and insights that the panel data delivered. First, several
elements of firm dynamics were fleshed out, such as the volatility of the
sector and the differing patterns of various measures of firm performance
over time. These could not have been addressed through a cross sectional
dataset. Second, an assessment of short run performance dynamics was
possible, and the comparison between the short and long run revealed only
subtle differences between the two. Third, the panel data made possible the
introduction of a lagged variable into the specification, improving the fit of
the model and better representing the dynamic process at work. Finally, the
IV (instrumental variable) estimate was introduced to adjust for
measurement error, which turned out to be more substantial in the short run
model.

Several shortcomings of the data, however, should be noted as well.
Panel data take a long time to collect, and in localities with wildly swinging
political or economic conditions, such time is not available. In Jamaica,
difficult weather and travel conditions, migrant proprietors, and poor data
collection infrastructure all complicated data collection efforts. Moreover,
several very useful pieces of flow data are difficult to collect, especially from
enumerators of governmental agencies. This is a common problem in all data
collection exercises, but data with continual random error exacerbate the

error in the estimates (Hsiao, 1986). Finally, the difficulty of panel wear out

148

and attrition is a difficult hurdle, and for Jamaica this was no different. This
effect is difficult to quantify. Panel data bring a different perspective to
microenterprise growth, however, as it extends beyond static considerations
to address dynamic issues. The need on the part of policy makers and donors
to understand dynamic processes better may provide the justification for
more panel work in the future.

Given the above findings, what are the implications for policy makers
and donors? As the broad range of comments below will highlight, there is no
single silver bullet that can bolster all the different dimensions of
microenterprise growth. The policy mix, in contrast, needs to encompass
multiple dimensions and reflect the social and economic tides of the day.
Further, Jamaica provides a particularly challenging microenterprise
environment, as such a large proportion of activity is focused on trade and
tourism. The comments below will first highlight three overarching stylized
facts, which provide a lens through which to view several specific
recommendations.

First, a very encouraging characteristic of these Jamaican firms
revealed in the intertemporal dynamics of the panel is the tenacity of the
microenterprises. Amidst economic decline and marcoeconomic shocks these
firms forged ahead, reflecting the Jamaican mindset, "You can't get me outta
the race."55 Secondly, the environment facing these firms contained large

swings in market demand, reflected in broad swings in employment in

149

 

Chapter 11.56 These swings potentially affected the proprietors willingness to
take business risks to expand or diversify their offerings. Finally, the
fundamental elements governing long and short run dynamics are mostly
congruent. Based upon the set of firm dimensions examined in this work,
policy directives do not need to discriminate between the two.

Within this context, several key policy recommendations can be
suggested. The first relates to firm size. In general, smaller firms grow
faster, confirmed by evidence from both the trend analysis of Chapter II and
the modeled relationships of Chapters III and IV. Two seeds of caution are
necessary regarding this result, however. One, own account firms are
incapable of shrinking their employment other than closure. In this case,
they would drop out of this sample and not be included in the analysis. This
introduces an upward bias into this result. Perhaps reflecting this
phenomena in the quarterly analysis of Chapter II, own account firms
remained robust to changes in employment while larger firms declined;
however, own account firms experienced a more significant percentage
decrease in sales than their larger counterparts. In fact, the firm with 2-4
employees achieved the best performance. This suggests that small, but
maybe not the smallest, firms should be the target for programs.

Second, younger proprietors can be a useful target of policy or aid

programs, given the strongest relationship between growth and age was in

 

55 Comment quoted from Anderson, 1994.

150

the youngest cohort. The insignificant result for firm age adds an additional
dimension to this recommendation as well. It appears from these data and
other research (see Fisseha, 1979) that there is little generational continuity
in Jamaican microenterprises. Further, few Jamaican firms use family labor,
and consequently microenterprises have a high potential for closure with the
death or retirement of the proprietor. In other words, firm learning in the
Jovanovic paradigm dies with the proprietor. Policy should be formed that
recognizes this life cycle aspect of the firm, encouraging for example more
continuity between successive generations of firm management.

Third, enterprises located in commercial districts achieve faster
growth, although it appears to make no difference whether a firm is located
in an urban or rural setting (see Fisseha, 1993 as well). As seen in many
other studies, home base enterprises are often slow growers or stagnant.
This fact is intuitive due to lower traffic levels at home enterprise and
competing interests for the proprietors time. Although home base enterprises
reflect lower levels of financial risk (lower overhead), policies could be
designed to minimize financial exposure in moving from a home based to a
commercial building.

The fourth implication regards the gender of the proprietor. Handa
(1996) points out that there is a very high incidence of female headed

households in Jamaica, as there are throughout the Caribbean. In Jamaica,

 

56 See Gustafson and Liedholm (1995) for details on how sales and output
fluctuated.

151

this percentage is roughly 42%. This implies two things. One, a large
percentage of women in the work force are not only working or running a
business, but also raising children and running a household. As Downing
(1990) suggests, this takes time and energy away from their business.
Secondly, however, Handa points out that Jamaican women are more likely
to choose female headship in their own household than women in other LDC’s
in order to improve their status and influence, for mainly economic reasons.
The high incidence of female headed households coupled with the high
percentage of microenterprise firms headed by females points to the
significance of the policy directives that should focus on female headed
microenterprise firms and female headed households. The importance of this
area of focus is highlighted in the negative relationship found between firm
growth and female proprietors.

Fifth, attention should be paid to credit programs, both for startup and
operating capital. Access to credit has been a hot topic in recent years, both
in the academic literature as well as for GO or NGO’s projects. In the panel
data model, both access to startup capital and post-startup capital obtained a
weakly significant result with firm growth. A very interesting result
obtained in the short run model was the significant positive relationship
between post-startup credit and firm growth. This result could be
interpreted as a working capital effect. Supporting this find, Rhyne and

Otero (1994) point out that access to small amounts of working capital is

152

 

perhaps the greatest need of microenterprises. These results support the fact
that credit accessed after startup has a positive impact on firm growth in the
short run, and roughly the same percentage impact on growth as access to
startup capital. Credit programs should pay attention, if not equal attention,
to each type of loan scheme.

The result on firm credit is not definitive, however, due both to the
weakly significant model results and the issue of endogeneity. Self selection
implies only the more successful firms are able to obtain credit. These firms,
already growing with a higher probability of growing faster than others, are
the ones to be approved for loans. Hence, credit looks like an extremely
effective policy program. This tangle cannot be unpacked, and only a
tentative case can be made for the positive effect of credit on firm growth.

A sixth recommendation relates to long and short run policy
considerations. Good news for the policy maker, long and short run dynamics
appear to be very similar, meaning little discrimination needs to be made on
policies between the two. One exception discussed directly above would be
considerations for short run operating loans. A case could be made for
increases in working capital loans to micro firms, which is a short run
consideration. These loans may also help smooth out some of the volatility in
the business cycle, which the firms seem particularly sensitive to. A second
consideration relates to age of the proprietor. In the short run, firm age is

positively related to firm growth. This result holds for all age groups,

153

 

although the magnitude of the coefficient decreases for older proprietors. In
the short run, age may simply be a good proxy for the proprietors energy,
willingness to take risks or succeed. Policies designed to motivate could
target any age group and result in positive economic results for the micro
sector in terms of employment growth. This finding could be particularly
noteworthy in an economic downturn, similar to this analysis period, when
the policy goal is to stimulate short run employment growth.

Finally, a confusing result relates to technical assistance or business
training. This dimension of human capital is difficult to measure, and with
all survey data, is always subject to under reporting or other forms of
measurement error. There are several technical assistance programs in place
in Jamaica, and the intuitive result would suggest these programs to be at
least somewhat effective. With one exception, most of the results were not
significant. In the 1992 cross sectional analysis, business training was
positively related to firm growth. This also happens to be the year of greatest
economic distress. One obvious explanation is that proprietors with better
training were better prepared to deal with the difficult demand environment,
hence the positive relationship with firm growth. This result is singular,
however, and does not lead to a strong recommendation for these types of
programs.

Several important research questions stem from this work. First, the

findings relating to credit and education (training) in general are counter

154

Intuit

 

I? ..

intuitive. For Jamaica in particular there are a plethora of microenterprise
training efforts, but it appears from these data that these efforts had little
effect on microenterprise performance. This result may be due to low impact
projects, or it may flag a need for a different design in data collection to
address these issues. Further, a construct to better understand the long run
effects of credit on a microenterprise need to be designed.

Second, the firm dynamics depicted in this dissertation reflect the
activity of surviving firms only. Although there is strong theoretical and
empirical justification for this, panel data incorporating birth and death
information would greatly enhance the validity of the findings. Most of the
research to date on firm survival has been generated in a static environment
or through a two period data collection scheme. To incorporate this type of
information into a rolling panel would enable a comprehensive analysis of the
dynamic process.

Third, firm productivity, efficiency and market demand represent
other areas of research related to firm dynamics and not explicitly covered
here. Evidence of low efficiency or demand in Jamaica were identified in the
rapid adjustment in firm output from period to period without corresponding
increases in employment. To assess these effects, better flow information on
changes in firm capital and some assessment of local market conditions

would be required.

155

- mm-rflnfl
.

 

W ..-

Fourth, this dissertation did not foray into the household side of the
enterprise. How central is the income from these firms to the household?
How are funds handled between the household and micrenterprise? How
does the health of the enterprise affect key measures of human welfare, such
as nutrition and education? Questions such as these relate directly to the
welfare implications of policy directions, and consequently remain an
important venue for future research.

Fifth, this research brought into question the treatment of firm age,
both on a empirical as well as a theoretical basis. Is the experience of a firm
best encapsulated in the age of the firm or that of the proprietor? Further is
this relationship linear or non-linear as Jovanovic's theory suggests? More
cross country work needs to be tackled on this issue, with careful attention
paid to the regression estimate of the firm age effects.

Finally, this dissertation has sought to confirm or negate some key
predictions coming out of the limited theoretical literature on firm growth.
Mark Blaug (1993) suggests that this type of research is the appropriate
manner in which to confirm or deny the reigning theoretical constructs.
Clearly, there are some significant holes in the empirical predictions, both in
the predictive capability and in the number of dimension included in the
paradigm. More work needs to be done to fill these voids and add more depth

to the existing literature.

156

fi’
'5’ ‘

 

‘1

The plight of microenterprises in Jamaica provides a challenge in as
much as the understanding of the dynamics of their growth and change over
time is truly a complex interplay between cultural, human capital,
governmental and market realities. The assessment of their contribution and
the health of the sector hence poses difficult questions indeed. Unequivocal,
however, is the role they play in the local economy and the important source
of income and livelihood they provide. The peril they face is enhanced or
diminished as a direct function of appropriate efforts to fully comprehend and

interact with their reality.

157

 

APPENDICES

158

 

APPENDIX A

Table A.1 Cross Sectional Model: 1990 thru 1994

Firm Age Interactions Included

 

 

 

 

 

 

 

 

 

  
 

 

   

   

   

    

 

 

  

  

 

Variables Ending Year for Growth Model
1990 I 1991 I 1992 I 1993 1994
Dependent In Growth
Variable
Ln Startup -.2960843 -.0422646 -.0465315 -.1177187 -.1201862
Size (-5.329) (-2.232) (-2.698) (4.292) (-4.427)
Ln Firm -.206484 -.0333079 -.026136 .0000492 -.0272202
Age (-2.710) (-1.062) (-.955) (.001) (-.546)
Ln Firm .0274899 .0023201 .0010431 -.OO30934 .0012247
Age (1.943) (.450) (.815) (-.409) (.153)
Squared
Ln Firm .0967071 .01 1851 .0120388 .0315852 .0322358
Age * (4.693) (1.888) (.034) (3.578) (3.782)
Startup Size
Constant .3727841 .0939525 .0859612 .0459734 .086056
(3.804) (2.031) (2.085) (.669) (1.132)
~Squrecf if “—1 _.0657 .___04722.,,,, I I I '.1——_ ,_- 12*.17“
F Value 15.45 6.94 8.09 14.10 13.59
I (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

 

 

 

 

 

Source: STATIN Survey Data
Values in parentheses are t values.

159

5 1.1-0

 

Table A.2 Complete Cross Sectional Model: 1990 thru 1994
Firm Age Interactions Included

 

 

 

 

 

 

 

 

I"

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variables Ending Year for Growth Model
1990 I 1991 I 1992 1993 I 1994
Dependent In Growth from Startup
Variable
Ln Startup -0.86495 -0.20707 -0.20433 -0.403 -0.47617
Size (9057) (-2.789) (-2.771) (3552) (4109)
Lu Firm -0.58248 -0.17175 —0.11359 -0.03871 -0.05535
Age (.4528) (4.387) (0883) (-0.19) (0229)
Lu Firm 0.09007 0.027625 0.018269 0.008122 0.00541
Age (3.372) (1.161) (0.745) (0.224) (0.128)
Squared
Ln Firm 0.240099 0.047164 0.036245 0.07839 0.106055
Age 1. (5.273) (1.557) (1.315) (1.97) (2.527)
Startup Size
Gender -0.10589 -0.06997 -0.07128 -0.06074 -0.08376
(-2.861) (-2.847) (-3.159) (-2.125) (-2.87)
Age of 1.051671 (dropped) -0.03269 0.484092 0.02921 1
proprietor (3.297) (0157) (2.105) (0.137)
1.162984 0.122561 0.049311 0.38564 -0.11871
(3.572) (2.033) (0.253) (1.689) (-0.558)
1 .226472 0.142688 0.099407 0.473204 -0.02935
(3.872) (2.415) (0.511) (2.078) (-0.138)
1.137687 0.129152 0.074253 0.396034 -0.02355
(3505) (2.249) (0.383) (1.752) (-0112)
1.117079 0.064132 -0.00394 0.37329 -0.11061
(3.544) (1.125) (-0.02) (1.552) (0525)
1.10047 0.051567 -0.01 0.33666 -0. 12592
(3.483) (0.852) (-0.052) (1.484) (0595)
1.013268 0.059042 -0.02771 0.324506 -O.13294
(3.187) (0.927) (0142) (1.422) (0525)
1.018408 -0.00744 -0.06417 0.333099 -0.13973
(3.147) (-0.095) (0322) (1.431) (0545)
Education 0.018046 0.003575 0.002862 -0.02447 -0.03218
(0.488) (0.15) (0.128) (-0.86) (-114)
Business 0.00166 0.025605 0.07535 0.048374 0.018114
Training (0.03) (0.732) (2.315) (1.072) (0.403)
Startup 0.040271 -0.01577 -0.01254 0.05511 0.035611
Credit (0.509) (0379) (-0.324) (1.084) (0.705)
New Credit 0.040069 0.051824 0.063721 0.0604 0.010575
(0.707) (1.4) (1.829) (1.445) (0.256)
Parish 0.006991 0.017005 0.061454 0.038192 0.072014
(0.085) (0.32) (1.242) (0.58) (1.087)
0.012895 0.056597 0.05413 -0.0595 -0.07716
(0.102) (0.684 (0.697) (-0.586) (-0.782)
0.035969 0.006109 -0.03631 0.086675 0.405794
(0.233) (0.05) (-0.382) (0.758) (2.597)

 

 

I60

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.10479 0.01305 0.01199 0.04888 0.05237
(0.954) (-0.18) (0.175) (0.582) (0.534)
0.04387 0.054459 0.054729 0.02313 0.032897
(0.503) (0.972) (1.23) (0.341) (0.453)
0.37955 0.07543 0.07815 0.28595 0.09554
(2.151) (0.891) (0.973) (-2.211) (0953)
0.051147 0.021592 0.01348 0.039068 0.00034
(0.77) (0.41) (0.275) (0.525) (0.005)
0.0515 0.05242 0.04827 0.07097 0.05329
(0.521) (0.975) (0.803) (0.93) (0.781)
0.02455 0.022813 0.014731 0.08074 0.12159
(0.275) (0.392) (0.257) (-1153) (-1.709)
0.14575 0.04802 0.03833 0.09039 0.11515
(-1.846) (0.944) (0.804) (-1453) (-1841)
0.045059 0.014392 0.034557 0.01071 0.01505
(0.537) (0.252) (0.575) (0.153) (0.231)
0.00943 0.028281 0.017085 0.007503 0.01348
(0.105) (0.485) (0.315) (0.109) (0.192)
0.04553 0.044879 0.038951 0.04572 0.0422
(-0.578) (0.89) (0.826) (0.715) (0.545)
Location 0.055553 0.034955 0.075299 0.104157 0.059135
(1.198) (0.991) (2.3) (2.539) (1.455)
0.229254 0.128599 0.153584 0.250895 0.03225
(2.517) (2.245) (2.911) (3.825) (0.491)
0.027571 0.205986 0.114153 0.153155 0.13919
(0.173) (2.155) (1.278) (1.155) (1.279)
0.015554 0.01539 0.041755 0.125173 0.030043
(0.178) (0.242) (0.599) (1.783) (0.437)
0.091854 0.053874 0.07979 0.114525 0.035735
(2.054) (2.237) (3.004) (3.354) (1.049)
0.015482 0.025801 0.045452 0.097451 0.027595
(0.124) (0.297) (0.558) (1.003) (0.305)
0.355351 0.127105 0.100525 0.070553 0.015832
(1.597) (0.858) (0.731) (0.317) (0.077)
0.158455 0.045477 0.042053 0.109134 0.058933
(1.757) (0.755) (0.755) (1.427) (0.988)
Rural/Urban 0.07751 0.02583 0.03031 0.00904 0.02153
(-1.843) (0.959) (4.205) (0.284) (0.588)
Constant 0.05429 0.259042 0.23387 0.24987 0.353288
(0.154) (1.55) (0.957) (0.755) (0.978)
R Squared 0.4447 0.1928 0.2541 0.3304 0.3514
F Value 5.25 1.95 2.94 3.43 3.29
I (0.0000) (0.001) (0.0000) (0.0000) (0.0000) I

 

 

 

 

 

 

Values in parentheses are t values.

Source: STATIN Survey Data

 

 

 

 

 

 

      

   

   

   

   

   

   

:3. 3 82..-. 89.3 83.3 83..-. 92.3
“398.- 823.- «are: 258.- 839...- ”$23.- 52.. .5 8.550
32.3 .82.-. Ens 9%
- - - $888.- 3325.- :33. _. 5? .EE 5
.93-. .93-. 89.3
- - . 5838.- 2 958.- 888.... N a}. E: 5
$3. a a: . z 88.: 83.8 32.8 .93.
«328. $28. 2533. 59.58. 589.8. 2828. owe. E: 5
83.5 83.3 .233 a: .8 88.3 88.3 :3. 88563:.»
:38. 3: Es. 3832. 834.3. 3583. 3884. 395. 5

   

 

 

 

 

 

 

 

 

E05323; 3083......» Buggies 2.05.3350 3085.350 305.3.an o...a...a>
a. a. a. .5 :4. :. Eugene:
2.:... «mm 3...: an». 2......— Sn 2.:... «mm 2.:: wan 2...... Gm
Sun: awn": ans—u: moan: a3“: nnn—H:
S 3 .3 3 3. 3
v .252 v .352 n .23.). N .252 N .35.). — .23.).
>— mAO mac mac >— mAO SO 30 3:395er

 

 

2.3-33 .mflq—mem a mAO can mAO

3am .ezam 323330832 56.95:.
".352 Benzene—new teamed a. :— 30? 03% ad 039..

m “£339.54

162

 

 

 

 

 

 

 

 

 

.88. .. .88. .. .88. .. .SN. .. .68. _. .88. _. .856
8.3.8... 8.. . 2.... 8888. 8888. 8888. 8888. 85:338..
.3 a. .. .8”. .. .88. .. .8... .RN. .. .888.
$83... 8288. 8888. 888. 888... 888... .820 8288
wage...-
...8...-. 88.... .84.... 88..-. .88.... .82.. 58.95
888...- mem88. 88:... 888...- 5888. 888. 8.82.8 18.9.
.08..-. .82.-. 8%.... .83..-. .888.-. .82.. 89.8%.“.
88. 8.- .88...- 4888. 88. 8.- 888.- 8888 .o .25.
.82.-. .82.-. .8. .~-. .88.-. 88.8-. 88. .-.
888.- 8888.- $88..- 8883- 88:8.- 882 .-
888. .83-. .88..-. 88.8-. .83-. .88..-.
8888.- $88....- n88m. .- 8888.- m. 88...- ..88. ..-
88.8-. .83-. .83-. .82-. 88.8-. .8... . -.
8888.- 8888.- 838: .- 8888.. 8888: 888. .-
888.-. .82..-. .48. .-. 88.8-. 38....-. $88-.
838.- 8888.- 8. 88.- 2888.- 8889- :88.-
.888. .83-. .8. 8-. 38.8-. . .88-. .. m. .8.
4888.- 88.88.- 88..8.- 8.88.- 888? $88...
.83-. .33-. .82.-. .83.-. .83-. .88....
883.“.- 8888: 8888.- 89.3.- 8888.- 83.88...-
.33. .83-. .mom. 7. .23-. .88..-. 88.8-.
: 888.- 8889- 888.- 8:88.- 8888.. 838...-
.83.-. .888. .83... 388-. .888. .58-.
833..- 888? 8.2.8.- 2 88...- .§.3.- 8848.. 58.52.. 8 ow<

 

 

 

 

 

 

163

 

 

 

.8888.
888888.-
.888.
8888888.
.888.
8888888.
.888.-.
.888888;
.888.-.
8888888.-
.8888.
888888.-
.8888.
8888888.
2888-.
88888.;
.. .88-.
88888.-
.888.-.
888888.;
888.
8888888.
.8888.
8888888.-
.888...
888.888.

 

.8888.
888888.-
.8888.
8.88888.
8.88..
88.8888.-
.888. .-.
888888;
..88. 7.
888888.-
..88.8-.
8.8888.-
.. 88.8.
88.888.
.8888.
88. 8 . 88.-
..88.8-.
88888.-
.888.-.
888.88..-
.888.
8888888.
.8888.
88. .888.-
.888...

E88888.

 

.8888.
8888888.
.8888.
88888.8.
.. .88.
8888888.
.8888.
888888.-
.8888.
888888.-
.888.
8888888;
2888-.
888888.-
.. 8.8-.
888888.;
.88...
888888.
.8888.
8888.8.
.8888.
8.88.88.
.8888.
8888.88.
.88...
8.88888.

 

.8888.
8. .8888.-
.888.
888888.
.8888.
888.888.-
.888.-.
8888888.-
.888.-.
8888888.-
.88..-.
8888888.-
.8888.
8888888.
.8888.
888888..-
.8.8.8-.
88888.-
.88..-.
88888..-
2888.
8.88888.
.8888.
.8888.-
.88...
8888888.

 

.888.
8888888;
..88.v.
8888888;
..88.8.
88.8.88;
.888.-.
.888888x
.888.-.
8888888.
.888.-.
888.88:
.8.8.8
8888.88.
.88.-.8-.
8888888;
.888.
8888.88.
..88.-.
8888.8.;
.8888.
888888.
...8.8.
8888888;
.888..
8888888.

 

.8888.
8888888.
.8888.
88.8888.
.888.
8888888.
8.8.8-.
888888.-
8.8.8-.
88.888.-
.8888.
8888888.-
.8888.
8888888.-
.888.-.
.88. .8..-
.88. ..
8888888.
.88..8.
8888888.
.8888.
8888888.
.888.
8888888.
.888...
8888888.

 

88.5.

 

 

I64

 

 

88.8.8

.888..

   
 

 

 

 

 

 

 

 

 

 

 

 

 

 

AN-...:oE>o_nEm
Oz 02 02 mm; <2 <2 8888-. . 88< E:
ANN-8.208882qu
88> oz oz 88> <2 <2 8888.
8838...; >_ 8888.8m
.888.8. .888.8. .888.8. .888.8. .888.8. .888.8.
8888888. 8888888. 8888. .. 8888. 88. 88.888. .8888.8. .8388
.8.8.8. .888.8. .888.8-. - .88..-. .888.8-.
88.8888. 8888888. 8888888.- - 8888888.- 888888..-
..8. .8-. .8.8.8-. .888. T. .8.8. T. .888.8. .888..-.
888888.- 8888888.- 888888..- .888888.- 8888888. 8.8888..-
.888... .888... .8. ..8-. .888.8. .888.8-. .8.8.8-.
888888.. 88888... 88.88.8.- 8888888. 888888.- 8888888.-
- - .888.8. - - .888.8-.
- - .888888. - - 8888888.- 8.88888
.888.8. .888.8. .888.8-. .888.8. .888.8. .888.8-.
8.88. .8. 8888888. 88.888.. 8888. .8. 888.888. 88. .8.8.- 8885 .8 .888
.888... .888... .888.8. .888... .. 88. .. .888.8.
888888.. 88888. .. 888888. 88.88. .. 88888... 8888888.
.88.... .888... .888.8. .888... .88.... .888.8.
8888888. 8.8888. 888888.. 8888888. 8.88888. 8888.8..
.888... .888... .888... .888... .888... .888...
888888. 8888888. 888888.. 888888. 8.8888. 88888..
.888.8. .888.8. .888.8. .88. .8. .888.8. .888.8.
8888888. 8888888. 88888. .. 8.88888. 8888888. 88888. ..
.888.8. .888.8. .888.8. .88..8. .888.8. .888.8.
888.8.. .88.88.. 888888. .8.888.. 888888.. 88.888.
.888..-. .8.8. T. .888.8. .888..-. .888. 7. ..888.
888888..- 888888..- 888888. 88.888..- 888888..- 88.8.88.
.88..8-. .8.8.8-. .8.8.8. .88..8-. .8.8.8. .8.8.8.
888828.. 8888888.- 8888888. 8. .8.8.- 88888. 88.8.88.
.888.8. . 8. ..8. .888.8. .888... .888.8. .888.8. 888.888.
.888888. 8888888. 8888888. 8.. .888. 8888888. 8888888. .8 88888.

 

 

 

 

 

 

 

 

165

 

.3388, 8 8.8 8882.88.98.89 E 88383 :88.

Sam swam 25.88am ”meadow

 

coco...

good

coco-c

coco-o

coco...

coco...

8A...—

 

$63

 

wwdm.

 

mm...-

 

3.3.

 

Edm—

 

2...:

 

88.; 8

 

166

 

 

 

 

 

 

  

   

  

   

   

  

  

 

 

 

 

 

 

 

 

 

 

 

Amma- _ -. 88.3.”. S. 8.8-. 83. . -. Ammo-N. Am. 8.»...
a _ ~38...- mvcgbcr $388....- 88803... mm _ mob...- m... 88.- go... ..o 83:00
.888.8-. .888.8-. ..88.8-. 88.8
- - - 88.88:...- m~¢3_cr 8.58.- 8 ow< 5...... .5
9.3. S 388.... $88....
- - - ammo—c. wmmawc... 2mm. 5. N om< .EE .5
£8. _ . _. Ammod 9838. S 888..-. 88...... 23......
Eco—N... moan-8.8... 98.888. 8328...... 8....- wontmor 09.. .EE 5
Ammmsm. ASENN. A. .8. . m. 92... .. Ann-8:8. 8:... S C-.. 855.8388
, 8.88888. - -88888M8.- - 8888888. - ‘ 888.8888. , , 88888.8. - ,N 88888.». , .. --mmwmfid
.aoEmEQEo 8:95.82an «55.82.50 8:05.82an Eofimsnam 8:95.82an - «.8.8.-:3
:— :_ :8— :J :8— 5 8:08.539
85.—E can 88.—E 38 85.—E mmm 88.—E can 85.....— 38 83.....— mmm
one": amen: awe—n: amen: 39H: 83":
c. .8. ..8. .8. .8. .8.
8 .252 >_ v .352 m .232 N .252 >_ n .332 _ .252
SD mac 30 36 36 mac munamqucU
2.3-«28H .88—580% a mAO “:3 mAO

"—0852 unmEmoEEm 3.88.830 8.0 .8388

0 538.34..

167

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.888.8. .888... ..88... .8.8.8. .888... .888. .. 8885
8.8888. .888888. 88. .888. 888.8.8. 8.88888. 8888888. 88.88.8888,.
.888.8. ..88.8. .888... .888.8. .888.8. .88...
88888.8. 888888. 8888888. 88888.8. 8888888. 8888888. 8885 8888.8
wEEE-r
.888.8-. .88. .8. .888.8-. .888.8.. .88. .8. .888.8-. 8888.888.
8888888.- 8888888. 88888.8.- #88888; 8888888. 888888.. 8888888.. I .88
.8.8.8. .. .8... .888.8. .888.8. .888.8. .888.8. 8888888.
8888888. 8888888. .888. .8. .888888. 888888. 8888888. .8 .85.
.888. 7. .888. 7. .888. .-. .888.8-. .8. 8..-. .8.8..-.
8.8888,. 88.888. .- 888888. .- 888.888.- 888.8. .- 888888. .-
..88.8. .8. ..8. .888.8.. .888.8. .88. .8. .888.8-.
8888888. 8888888. 88. 8888.- 888888. 88888.8. 8888.88.-
.. 88.8. .8.8.8. .8.8. .-. .888.8. .88. .8. .888.8-.
8888888. .888888. .88888.. 8.88888. 88. 8.8. 888888.-
.888.8. .888.8. .888..-. .888.8. .8.8.8. .8.8.8-.
888.8.8. 88888.8. 8888.88.- 8888888. 8888888. .888888.-
.. . .... .888... .888.8. .888... .888... .888. ..
.8.88. 888888.. . . .8888. 88888. .. 888.88.. .8.88888.
.888.8. .88.... .888.8. .888... .8.8. .. .888.8.
.888888. 8888888. 888888. .888888. 8.88888. 8888888.
.888.8. .8.8.8. .88. .8. .8.8.. .888.8. .888.8.
8888888. .888888. 8.8888. 8888888. 888888. 8888888.
.888... .888. .. .888... ..88. .. .888... .88. . ..
88.8888. 88F8888.. 8888888. 888.8.. 88888.. 8.888L88. 88.8.8888. .8 88.8

I68

 

 

 

 

 

 

 

 

 

.888.8-. .8.8.8-. .888.8. .8.8.8.. .888.8-. ..88.8.
888888.. 88888.8.- 88888.8. 888888.- 888888.. 88.88.8.
.888.8-. .888.8-. .888.8-. .888.8-. .88. .8-. .. . ..8-.
8888888.- 8888888.- 8888888.- 8.888.8.- 8888888.- 8.88888.-
.888.8-. .888.8-. .888.8-. .888..-. .888..-. .888.8-.
388888.- .888888.- 88888.8.- 8888888.- 8888888.- 8888888.-
.88..-. .. .8..-. .888..-. .888..-. .888..-. .88..-.
888.888.- 8888888.- 8888888.- 88.8888.- 88.8888.- 8888888.-
.8.8..-. .888..-. .888..-. .888..-. .888..-. .8.8..-.
.88888. .- 888.88. .- 8888888.- .8888. ..- 888888..- 8888888.-
.888..-. .888..-. .888..-. ..88..-. .888..-. .88..-.
888. 88 . .- 88. .88..- 8888888.- 8888.8..- 8888 . .- 8888888.-
.888..-. .888..-. .888.8-. .88..-. .888..-. ..88.8-.
.8.8.8..- 888888..- 8888888.- 8.888... 888.88..- 8888888.-
..88... .888... .8.8.8. .888.8. ..88... ..88.8.
88888.. 888888.. 888.88. 888888.. 888888.. 8888888.
.888.8-. .888.8-. .888.8-. ..88.8.. .888.8-. .888.8-.
888288.- 8888888.- 8888888.- 888.888.- 8888.88.- 8888888.-
.888.8. ..8.... .8.8.. .888.8. .8. 2. .888.8.
88888. .. 888888.. 8888888. 8888. . .. 8.8888.. 8888.88.
.888.8. .888.8. .888.8-. .88..8. .888.8. ..88.8-.
8888888. .8888. .. 8888888.- 8888888. 8888888. 8888888.-
.888.8. .8.8.8. .88..8. .888.8. .8.8.8. .88..8.
8888888. 8888888. 88.88.8. 888888. 8888888. .8888.8.
..88.8-. .888.8. .888... .888.8-. .888.8. .8.8..

88888.8.- .888888. 8888888. 88.8888.- 8888888. 88.8888. 88.88,.

 

 

I69

 

 

wooed

 

NNHN. .0

mac...

3%...

32....

8.8.8208 M

 

 

 

 

 

 

 

 

 

 

 

 

 

R... .5582qu
oz oz oz 88> <2 <2 .8888. - 888 E...
AN-...:oEzo_nEm
88> oz oz 88> <2 <2 .8888.
838...; Z 88.3—om
.888.8-. .888.8-. .888.8. .888.8. .888.8-. .888.8.
88. .888.- 8.8888,. 8888888. 8888888. .888888.- 8888888. 8.8.888
.888... .888... .888... .888... .888... 888...
88.8888. 888. 888. 8888888. 8888888. .88. 888. 8888888.
.8. ..8-. .888.8. ..88.8. .88..8-. .888.8-. .8.8.8.
8888888.- 888.888. 8.888.8. 888.888.- .888888.- 8888.8.
.888... ..88... .888... .888... .888... .8.8. ..
8888888. 8888888. 88.8888. 88.8888. 8888888. .888888.
.888.8. .888.8. .888... ..88.8. .888.8. .888...
8888888. 888.88. .888888. 8888.88. 8.88888. 8888888.
- - .8.8.. - - .888...
- - 8888888. - - 8.88.88. 8.88888
.888.8-. .8.8.8-. .88. .8. .888.8-. .8.8.8-. .888.8.
888885.. 888.88.. 8888888. 8888888.- 88888.8.- 8888888. 8885 ..8 .83.
.888... .888... .888... .888... .8.8. .. .888...
8888888. 8888888. 8888888. 8888888. 8888.88. 8888888.
..88.8. .88..8. .888. 3 .888.8. .888.8. 8888-.
8. .88.8. 8.88888. 8888888.- 8888888. .8888.8. 8888888.-
.8.8.. .888... .888... .888... .888... 888...
888888.. 888888.. 888888.. 888888.. 88888.. 888888..
.888... .. .8.8. .888... .888... .888... .888...
8888.88. 8888888. 88.88. 888. .88. 8.8.888. 88.8888.
.888.8-. .888.8. .88. . .. .888.8-. .888.8. ..88. ..
88888.8.- 888.888. 8888888. .8888.8.- 8888.8. .888888.
.8.8.8-. .888.8-. .88. .8. ..88.8-. .8. 8.8-. .8.8.8.
8888888.- 8888888.- 8888888. 8888888.- 8888888.- 8888. .8.
.888.8. .888..-. .888.8-. .888..-. .888..-. .888.8-.
8.8888..- 88888. ..- .888888.- 888888..- 8888. ..- 8888888.-
.888.8. .888... .88..8. .888.8. .888. .. .88..8. 8888.888.
8888.88. 8888.88. 888.88. 888888. . .88.88. 88.8888. .8 888888.

 

 

170

.8839, 8 8.8 888.858.8588. a. 8838.8 .2

8.88. 88:88 2.8.888 H88.888

 

_

coca...

coco...

coco...

coco...

coco...

good

“.8.:

 

mmdm.

 

ma.mb

 

98.08.

mm.oma

 

 

8.2.

 

3.5

 

828> n:

 

l7l

APPENDDi D

THE SURVEY INSTRUMENT

These comments relate to the survey instrument used for the
Quarterly Panel Survey (QPS) only. This instrument was specifically
designed for this project with the intent of capturing data for the STATIN
National Accounts publications. In 1993, STATIN only reported national
account data for firms over 10 persons. The design of the instrument, which
will not be detailed here, covered employment, sales, output, and fixed
investment for microenterprises. The following comments relate how various
aspects of the instrument delivered their intended results.

Overall, the instrument was effective in capturing the intended
information. Information on employment was the most complete and
accurate. The data on wages, sales, output, and fixed assets were not as
robust, although for all but fixed assets the information was mostly credible.
The most serious difficulty encountered was logistical. The sampling was
random and nationwide in scope. Consequently, the enumerators had a very
difficult time visiting all of the firms within the specified data collection
period each quarter. As most of the enumerators relied on public
transportation, visits to remote rural areas was time consuming and often
would not be repeated if the proprietor was not available. The success rate of
visits could have been improved with a different sampling scheme (cluster

sampling, for example).

172

 

 

Most micro-entrepreneurs do not keep records, so the instrument itself
was detailed and designed to operate as a worksheet. This had benefits and
pitfalls. On the pro side, some very good basic data on employment, wages,
sales, and output was obtained. Beyond non-response, the information
obtained particularly on sales and output appeared credible and matched
that of other enterprises of the same size.

Three things in particular did not work, however. First, the more F-
complex portions of the worksheet were typically not filled out. When asking

for "fractions of output" related to the reference week, for example, little

 

reasonable information if any was conveyed. Secondly, the reference week i. .
was the only period of time that respondents could reliably provide
information on. Data for the month or quarter was either not reported or was
simply a multiple of the weekly data. Finally, very little information on firm
assets or raw materials was collected. Proprietors could not recall purchase
prices or dates. I

For a context similar to Jamaica (national survey, census bureau data
collection), an improved instrument would incorporate the following. One, a
worksheet design, which worked fairly well in this project, but simplified to
peel away complex and confusing thinking (don't ask for fractions of weeks,
for example). Two, collect information only on the reasonable basics, such as
employment and sales. Information on fixed assets or raw material use are

hard to come by. Third, sample in a way to minimize logistic bottlenecks to

173

information flow. Finally, if the intent is to conduct a national survey,
careful thought should be put against which sectors should contain bolstered
sample for drill down analysis. Panel data is expensive to collect and
challenging to manage. Even a small sample of 350 results in a massive

amount of data to manage, but little drill down ability within sub-sectors like

manufacturing due to small sample size.

174

 

 

 

gods“.

 

scu>ucucn no open

 

uoau>uon=m no «In:

 

 

 

upro8>uoucn uo «In:

 

 

.88.: .83.... 8. .

 

8888.8...
8.88835

 

0.05.0.8: .8 88.102
n.uouoaunoum

 

amaze

 

>oco=u8uncou

 

saw-nah

 

88.5..
8:828:38...

 

 

 

.8.8.8.... 88.. 2.8.8
858.3338

 

 

.suuguuuucooH N . H

.ll88. non-808 £213....» .258. £93: :25 8.382 2.35 83.825 05 88 338.835 .A

88883858838... 38 a? .8 888.-.3

 

.gfiluuouooagu

.SEQODMgu—Ukh.

ouuunu0uclouuut uo >o>unu «calm unuOHuoao sauna-an

non HH 5

son .855.
A<_hsz—8—ZOU

5.53: .3 858.8882. ..8.-88:58

175

«come uni-:25 “50> o!— uouuuso 0.80:) 0:» “cu a >c5= lo: .22

 

 

 

   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.v.‘uu
«coma >323: c.0525 u=o> as: >38 Mom 9:5: >:a§ 30.. .0525... 0920 an: 0:» 9355 6.1:
8:090 >325: c.0525 56> In: 8.00: non Inl- v >c2. 30: ..8.-8:9: 0053 an: on» 2.8-=5 .uiaa
£25000» van—:93 on .8.—=00 uni...— uo v02!— .8:- .Movaou ..8.—Elm no can»: .Om>u Oll- uo anon—.803 ”Ha I ca .800
3.83.) any no 0088.09 o...» 00 .8080» ac: mean. can vauooaaoo .8 nu... noun: anon: 86.8-«on mgu I8 85:»; as: 83.6 .800
:hoguumoun 0:» >19 nut—30 :0: Coca yo: noon 8:. c.3525 .40 >3 v.8!— na maneuunoun 9.3 no Nov—no: ozu to: 2.051 3 0.1:. 8Q 130
21:93 a u 3. no. 3:533ch nova-.2— In uni .8.-87:!“ no 0.5-.733. and-8&0»... 23 Id .8.-SI on: >von>h05fl "a 1.00 £90:
.8 A- 8.2. a... .8 8: 8.5.8.. .8 A- 8:33.: ..8 A- 88...... .8 A: >33 Mug-mug
.v AI 3..- Ooounxdianlo 3 AI uoflao on... “N A: gun-nan 8n A: In: Ion—u «..8.-81:3... 80 AI 64‘ vol mafia—Mg
.m AI ans-lo 8' AI 0.5:“... 8" AI hoxuoz nun-50: 8N AI hOAIOI haul-h 8” AI Magi-noun 8%
8:88:88
\8\\\\\\\ m .. < .8. o .8
88. 1.80... 88. .30... 8.3: .8 882. .8... 88. .30... on: .8... «>88 .8. 88. .30... 88.-583. 882888: 8...: n... 838. Z...
N macho 1:
. .8296 .o.
.m
.0
.8.
.o
.m
.v
.n.
.N
hound-uncut .u
.88. .83 .3. .3. .8. .8. .8. .8. .8. .8. .8. .8. .8. .3
0.81.— nsconéuo uOu 0.5 3...: v0.10: .3. no.“ «1... 9:5: no \ .->Ia 0.6 can.— «uvuouou anus->98 .1: 30.: our... hex-no: uo
8.88.-«I3. 3... 88 8o 8 1.3 «o 8 88. .38.. 8 .38. veto: 88. .8.: 83.86: :38 8:58! 5.3.. 23.. 3:8
1 . 41 .« novuo
Ea Cosh—Inca Orin—5 En! .uoz wean—.5 In”! coco-nuns. 9.355 of: in IQ! hog—no: «Gav-“>89: no cauunquOIg
«0 I

 

 

 

8 on:
.

 

u.‘HH

ill. 53.62:. no use .g

176

 

mathob.

 

Ilsa-.1...

 

8.1.... .2

 

3:8 i5... .2

 

8......3 ...
its...

 

82::- 888 .2

 

812—.— .2

 

5‘ -=

 

 

838: .2
3:5-

 

... E
E 8. .8...

 

s.
.8. $.32

.88 333.331.. I. ..

 

 

.8.. E

 

 

E E I...

25......8. ..
in... 11.1..

 

"..8.. :55. I...

 

:8... .. :55...

 

.8. .15-3.8
u! i.—

 

..5=....=§ 3. I: .1 .8

 

.n

 

.~

 

 

..
n..8......8....8.8.5......

 

.8 # .8
5 an :0.—

s.
.5. 88:82:.

.8.
an .5.—

E
«a 3:

.8.
2. .3:

 

z.- E

 

 

5.3. 88:...- E

 

5...- 85...: .8.

 

.8... 88:...- 5

 

.8. . ..
.238— !i 8888:.-
l.. I...-
.... is}... .. .3.

 

 

 

mg HH Eon

aging n23

177

 

~ Sauna-6 can. . 2.20: can. u 3.00: 05.

£3: .2 no .30.» 3.. J30: 3. an 00:09 Quechua.» non.
non—anon... :0» 0.2.0) .230 coouzuua uo 0330» on» 93.33230 .aHHHIAIIIIII

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

< < < < < a 4 a o a
> > > > >
< < < < < .8.—:96 It».
> > > > > 2605.283: .3
.o
.o
...
.o
.n
.v
.n
.u
.u
32 3. 3. 2.. .3 3. .3 .2 A". .2
an «38. Jo » .38. £0 03!. an 03.... .3 .38. «.3000: a?» no 3.»... new
«0 loan 00.20»?— 8000a
".933 .ug uOh H.009 £060: .uIfl NO.- Hg 1! OOGOBOHO¢ NON 3.0.»: #1003 van—D oNOuCH In “0 ad‘s-20H!

 

 

 

 

 

 

 

 

II «o I .00..—
23 £3.32... no noon. 32.... non ~03 .8.... 03?... ..o 83.... a. 2...... a! ..o. g on .89.

8.8. 0.3.5 F :03. 032.39.... we 0.32....— 33 a; .8 .8 .g

GIHH

178

 

 

 

 

fi

3.00.. as. I.» 05...... 33...... .3. 02.0 an: 5.... 2.» 5...: ..o as... no 000...... 00.8.... 0..» on... a...

an. minor—0A »« .flm: 000M... 003 00d0fl0¢
KO 300 0:0 .v0n5»00u=:00. 000008.— ».
00090-0 >00 u0 0.1.0.. an .0»0.—. 0..» 0. »0..3

.u»0 .000..?..0m 0:0 .05. Souu 00:000.. 0...: 00..» .Hw: 0000.... an .uka.
.=u>nkb¢ .NON’EOI! In: a; g GE E gngfi AIIIII

. £09.05 2....

2.»..0: 0:... .
33.! .n. no .00.. .3 £00.02 .3 00 00.30.. 00:000.?- 5000 00.300000 50> 01.0.. .00.:- uo 0.5.0.. 0..» 8.000.008

. ..oo: 2....

.N.flnun All

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.Nn 2.00
.E .1808 0...» new ..0»o» 0..» 0. 000.. 2.9.00 00.»!02... 0...... .3 a o. 2.60
.0:0u Nth .ulx 0...» no...) 0.. .001 0..» uo >00 £0.25 0..» A»... >09 .ulx 0..» no. 00400 u0»0» 0..» 000080 0M0: .m 1.00
.600 .2... $0002.00; £00005... >020: 6.0. auuub 00» no >06 0..» »..A .0 2.00
.nu ..- h ins
.3230. 3...... .v ..8. o» .260 u. .o ..8. :60 ..qu no. 30- ul». 0..» no 3...: 45.9... 0..» 9.0.. u... 3 :30
.m ..00 .5 gaging. 0.. .585... 000.... 0'03. 0..» «o 00...... 3:... 0..» no 000 no... 00......— 0..» 0:00! ..10043... 2.. 2.00
:3. £00 0..» new 0.00 00».— u0n0l 0..» no .401... ..0uo» 0..» »..A .0 400
63.0.: 0.00 01.0 0000! 0..» 00.30 nonuloo» 0095.5 0.. 0.000 00.00 005,-. 00..»0 .000... 0a >0... 090000u00 0‘» 9500.. 0.00 0.0». no.0: ..n 2.00 .a
.IIuOu no 0000.. 0.5.00.3 0..» 0.. v00: 0.0 >05» 00 05:00! 0I00 0..» 2.0.. 0. 0:0 .0 .n .n 0.30.00 .a
< < A < a o ..
> >
< /\ 0:00:0ua000..:
> >
. 9
u n
.m
.1
.n
.w
z.
.2. .2. 3.. 8.. .o. 3. 2.. 3. .m. _ .v. .n. .m. 3.
cum»: 2.... i 3. .v... .8 ..on ..oz .0 ».a.> .25.. 0...; m 3... ..8.... do .30: «0.9.3.... 53 ..00 03......
>00 uaoausm .0 >09 .0 Scum 0.0m 000......0m
E OUEOHONOE nOu A000... ..Udi OMEEDU 0....» 90.: kg conflufluom uOu IOSAI> diva vac... .hOuCn IH U... Glacilnnvhs.
"000..
0...... 03:00 .5. a 3 u so 30.... 09...... so... was»... 4430].. a .8... as»: .33

 

 

..000 0.00.0 0 «0 00.09 0:» »00 000. 00:» no «.0 no and 0:»
0d 000». 00nnouo 0008» 00» 00.0» ab 03h .aaa000«>a0ca u0»c0 0» ..000 00» no >000 00» 000 »0£» 000». no ocunaouu any 0. :0: 0.83 .«dldfldfl

..00dann00 000 ..o.0. vablon no .00q000nuu00 000 00000 000:»00uac00 no» ..o.0. 0u00 00 0000 0a» 000 000 »000 00> .0000000 no v.00 0a»

0000 00 .00000 00 an .0»0 .0udauu no 0»0800n n ..0000020 «0 0x000 v .0059 .0000000 no 0»«00 00 000: 02 0.000 .o»0 .08000 .uduauu

.Iulfluqa 00 0000 000.0» 00:» 0000 000040 ..008 0:» >0 00.00 nan .0uusuu 000 000un 00 .00: :0 >0 00000000 004000 ..o.0. »«c: 00» 00000
00»... c...» .80»..qu 3.. 0.. ». .3 »0..». 03008 »...n n. 00». ..- u. .03.. ..o :3. .70.. 0.52. on 3.5» 8:31. .o a...» 2.... .ululnlnna

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.33....- 5 a: 03.. on o» .33....- o...§§2.8 . 34300430403300 .0
no 203002.... goo»- 0..» an... .00.. 03.500330... 0 up. .n .3352... ..8.... 0..» u... .00.. 033.00.09.00 0 no. .0
.8oou?».&»ao .0 gaging .3032... .33.... 33.6 ..o 380 .0
0.0.00»... 30.. 0..» :2 .00.. 00.00.60.500. 0 no. .0 .qua $03.32.... zoo»- o..» 3.. .00.. 05.500.22.00 0 .3... .0 .30”.
tit... tit: .. 0 ... o a xiii: 2:2: .. 0 a o ...

083.... 000». 00.0.4000 I 08.6.... 000». 2.0.5.0.. v.

n. n.

a. a.

a. 3

o. o.

a o

o o

.. ..

w o

m n

v v

n n

u a

. .

.2. .o. .o. 2.. .o. .m. .0. .n. .2 .3

on all... 3.... 8.03003... .30 00.5.. a. an 008...... 3.5 09...... 0... 00305800..
00000000 0usncu 00¢ .02 n00 .00
030.. 3.3.3.... .2... no. .300»... 30.. 0......» 030: 5.3.3.6 2.02.9.5 so. .5950 .300
.III «0 III .0000

..§\a£§03u§.08.008..§02§00§.08§ .0033.§0000000.§

N I H

180

REFERENCES

18]

 

REFERENCES

Acs, Zoltan, and Audretsch, David. "New-Firm Startups, Technology, and
Macroeconomic Fluctuations," Small Business Economics, vol. 6, 1994.

Aislabie, Colin. "Sudden Change in a Model of Small Firm Growth," Small
Business Economics, vol. 4, pp. 307-314, 1992.

Anderson, Patricia. The 1992 Jamaican Microenterprise Survev. Mimeo,
June 1994.

Arrighetti, Alesandro. "Entry, Growth and Survival of Manufacturing
Firms," Small Business Economics, vol. 6, pp. 127-137. 1994.

Ashenfelter, Orley; Deaton, Angus; and Solon, Gary. Collectingianel Data in
DevelopinLCountries: Does it Make Sense. Living Standards
Measurement Study, Working Paper No. 23. The World Bank.

Audretsch, David and Mahmood, Talat. "The Hazard Rate of New
Establishments," Economics Letters, vol. 36, pp. 409-412, 1991.

Behrman, Jere R., and Deolalikar, Anil B. "The Poor and Social Sectors
during a Period of Macroeconomic Adjustment: Empirical Evidence for
J amaica." The World Bank Economic Review, Vol 5, No. 2. 1991.

Behrman, Jere and Deolalikar, Anil. "...of the Fittest? Duration of Survival
of Manufacturing Establishments in a Developing Country," he
Journal of Industrial Economics, vol. XXXVIII, no. 2, pp. 215 — 227,
December 1989.

Birch, D.L. Usifl Dun and Bradstreet Data for Micro Analvsis of Regional
and Local Economics. Cambridge, Mass. MIT Program on
Neighboorhood and Regional Change. 1979.

Blaug, Mark. The Methodology of Economics, or How Economists Explain.
The University of Cambridge, Cambridge, England. 1993.

Brock, William and Evans, David. "Small Business Economics," Small
Business Economics, vol. 1, pp. 7-20, 1989.

Cabal, Miguel F. "Entry, Exit and Growth of Micro and Small Enterprises in

the Dominican Republic 1992-1993." Michigan State University Ph.D.
Dissertation, East Lansing. 1994.

I82

Cabral, Luis. "Sunk Costs, Firm Size and Firm Growth." The Journal of
Industrial Economics, vol. XLIII, no. 2, pp. 161 — 172, June 1995.

Chappell, William F., Mayer, Walter J ., and Shughart III, William F. "Entry,
Exit and Industry Performance," The Journal of Economics, XIX, no 2,
Fall, 1993.

Chesher, Andrew. "Testing the Law of Proportionate Effect," The Journal of
Industrial Economics, vol. XXVII, no. 4, pp. 403 - 411. June, 1979.

Cressy, Robert. "Pre-entrepreneurial Income, Cash-flow, Growth and
Survival of Startup Businesses: Model and Tests on UK. Data " Small
Business Economics, vol. 8, pp. 49-58, 1995.

 

Deaton, Angus. "Panel Data from a Time Series of Cross-Sections " Journal
of Econometrics. 1985.

 

Deaton, Angus. "Estimations of Own- and Cross-Price Elasticities from
Household Survey Data," Journal of Econometrics, 1987.

Dobson, Stephen and Gerrard, Bill. "Growth and Profitability in the Leeds
Engineering Sector," Scottish Journal of Political Economy, vol. 36, no.
4, November 1989.

Downing, Jeane. "Gender and the Growth of Microenterprises." GEMINI
Working Paper No. 5, October, 1990.

Dunne, Paul and Hughes, Alan. "Age, Size, Growth and Survival: UK
Companies in the 1980’s," The Journal of Industrial Economics, vol.
XLII, no. 2, pp. 115 — 140, June 1994.

Dunne, Timothy, Mark J. Roberts and Larry Samuelson. "The Growth and
Failure of US. Manufacturing Plants," Quarterly Journal of
Economics, vol. 105, no. 4, November 1989.

Economic and Social Survey Jamaica. 1992. The Planning Institute of
Jamaica, 1993.

Ettlinger, Nancy and Tufford, Michelle. "Evaluating Small Firm Peformance
in Local Context: A Case Study of Manufacturers in Columbus, Ohio,"
Small Business Economics, 8. 1996.

I83

Evans, David. "Test of Alternative Theories of Firm Growth." Journal of
Political Economv. vol. 95, no. 4, pp. 657 — 674. 1987.

Evans, David. "The Relationship Between Firm Growth, Size, and Age:
Estimates for 100 Manufacturing Industries," The Journal of
Industrial Economics, vol. XXXV, no. 4, pp. 567 — 581. June 1987.

Davies, Stephen P., Mead, Donald C., and Seale, James L. Jr. "Small
Manufacturing Enterprises in Egypt," Economic Development and
Cultural Change, The University of Chicago Press. 1991.

Fisseha, Yacob. "Dynamic Study of Jamaican Micro- and Small Scale
Enterprises", Draft paper, June, 1993.

Fisseha, Yacob,. “Management Characteristics, Practices, and Performance
in the Small Scale Manufacturing Enterprises Jamaica Milieu,”
Unpublished Doctoral Disseratation Thesis. Michigan State
University, Department of Agricultural Economics, 1982.

Fisseha, Yacob and Davies, Omar. "Small-Scale, Non-Farm Enterprises in
Jamaica: Initial Survey Results," Working Paper No. 8, MSU Rural
Development Series. 1979.

Fisseha, Yacob and Davies, Omar. "The Small Scale Manufacturing
Enterprises in Jamaica: Socioeconomic Characteristics and
Constraints," Working Paper no. 16, Department of Agricultural
Economics, Michigan State University. 1981.

Grootaert, Christiaan and Kanbur, Ravi. "The Luck Few Amidst Economic
Decline: Distributional Change in Cote d'Ivoire as Seen Through
Panel Data Sets, 1985-1988," The Journal of Development Studies, Vol
31, No. 4, April, 1995.

Grosse, Robert. "J amaica's Foreign Exchange Black Market." The Journal of
Development Studies. Vol. 31, No. 1. October, 1994.

Gustafson, Todd and Liedholm, Carl. "Patterns of Change Among Jamaican
Microenterprises: Results from A Quarterly Panel Survey, 1993-1994,"
GEMINI, 1995.

Hall, Bronwyn. H., "The Relationship Between Firm Size and Firm Growth in
the US. Manufacturing Sector," The Journal of Industrial Economics,
vol. XXXV, no. 4, pp. 583 — 606, June 1987.

184

 

Handa, Sudhanshu. "The Determinants of Female Headship in Jamaica:
Results from a Structural Model." Economic Development and Cultural
Change, 1996.

Hart, RE. and Prais, S.J. "The Analysis of Business Concentration " Journal
of the Royal Statistical Society, A. 119, 1956.

 

Hay, Donald and Louri, Helen. "Investment in Inventories: An Empirical
Microeconomic Model of Firm Behaviour," Qtford Economic Papers.
1994.

Hausman, Jerry and Taylor, William. "Panel Data and Unobservable
Individual Effects," Econometrica, vol. 49, no. 6, pp. 1377 — 1398.
November, 1981.

Hotchkiss, Julie L. and Moore, Robert E. "Gender Compensation
Differentials in Jamaica." Economic Development and Cultural
Change. 1996.

Hsiao, Cheng. "Benefits and Limitations of Panel Data." Econometric
Reviews. 1985.

Hsiao, Cheng. Analysis of Panel Data, Cambridge University Press,
Cambridge, 1986.

J ovanovic, Boyan. "Selection and the Evolution of Industry," Econometrica,
Vol. 50, No. 3, 1982.

Lever, Marcel. "Firm Size and Employment Determination in Dutch
Manufacturing Industries," Small Business Economics, vol. 8, pp. 389
— 396, 1996.

Liedholm, Carl. "The Dynamics of Small-Scale Industry in Africa and the
Role of Policy," Gemini Working Paper No. 2. 1990.

Liedholm, Carl and Mead, Donald. "Small Scale Industries in Developing
Countries: Empirical Evidence and Policy Implications," Michigan
State University International Development Paper no. 9, Michigan
State University. 1987.

Liedholm, Carl and Mead, Donald. "Dynamics of Microenterprises: Research
Issues and Approaches," Michigan State University, 1990.

185

 

Lillard, Lee and Smith, James P. "What do we Really Know about Wages?
The Importance of Nonreporting and Census Imputation." Journal of

Political Economy. 1986.

Lucas, Robert. "On the Size Distribution of Business Firms." Bell Journal of
Economics, vol. 9, No. 2. 1978.

Maddala, G.S. "The Use of Variance Components Models in Pooling Cross
Section and Time Series Data," Econometrica, Vol. 39, No. 2. 1971.

Mansfield, Edwin. "Entry, Gibrat’s Law, Innovation, and the Growth of
Firms," The American Economic Review, vol. 52, pp. 1023 - 1051,
December 1962.

Mata, Jose. "Firm Growth During Infancy," Small Business Economics, vol.
6, pp. 26-39, 1994.

Mata, Jose and Portugal, Pedro. "Life Duration of New Firms." The Journal
of Industrial Economics, vol. XLII, no. 3, pp. 227 — 245, September
1994.

McPherson, Michael A. "Growth and Survival of Small Southern African
Firms," Unpublished Doctoral Thesis, Michigan State University.
1992.

Parker, Joan C. "Defining and Measuring Business Growth," Draft Paper.
1991.

Pischke, Jorn-Steffen. "Individual Income, Incomplete Information, and
Aggregate Consumption," Econometrica, Vol. 63, No. 4. 1995.

Polachek, Solomon W. and Kim, Moon-Kak. "Panel Estimates of the Gender
Earnings Gap," Journal of Econometrics. 1994

Prais, S. J. The Evolution of Giant Firms in Britian. Cambridge University
Press, Cambridge. 1976.

Preliminary Report on the Sample Census of Small Businesses. 1990. The
Statistical Institute of Jamaica.

Reid, Gavin. "Early Life-Cycle Behaviour of Micro-Firms in Scotland," Small
Business Economics, vol. 7, pp. 89-95, 1995.

I86

 

Reid, Gavin. "Mature Micro-firms and Their Experience of Funding
Shortages," Small Business Economics, 8, 1996.

Robin, Jean-Marc. "Econometric Analysis of the Short-run Fluctuations of
Households' Purchases," Review of Economic Studies. 1993.

Schmalensee, Richard. "Inter-industry Studies of Structure and
Performance," Handbook of Industrial Organization, Volume II, eds.
Schmalensee and Willig, Elsevier Science Publishers, 1989.

Simon, Herbert A. and Bonini, Charles P. "The Size Distribution of Business
Firms," American Economic Review, Vol. 48, No. 4. 1958. E...

Small Establishment Survey. 1983. The Statistical Institute of Jamaica.
Stanley, Michael H.R., Buldyrev, Sergey V., Havlin, Sholmo, Mantegna,

Rosairo N., Salinger, Michael A., and Stanley, H. Eugene. "Zipf Plots
and the Size Distribution of Firms," Economics Letters, 49, 1995.

 

Statistical Review. The Statistical Institute of Jamaica. September, 1992.

Strassman, Paul W. "Home-base Enterprises in Cities of Developing
Countries," Economic Development and Cultural Change, Vol. 36, No.
1. 1987.

Storey, David J. "Firm Performance and Size: Explanations from the Small
Firm Sectors," Small Business Economics. 1. 1989.

Wagner, Joachim. "Firm Size, Firm Growth, and Persistence of Chance:
Testing Gibrat’s Law with Establishment Data from Lower Saxony,
1978-1989," Small Business Economics, vol. 4, pp. 125-131, 1992.

Wagner, Joachim. "Exports, Firm Size, and Firm Dynamics," Small Business
Economics, vol. 7, pp. 29-39, 1994.

Wagner, Joachim. "The Post-Entry Performance of New Small Firms in
German Manufacturing Industries," The Journal of Industrial
Economics, vol. XLII, no. 2, pp. 141 — 154. June, 1994.

Wint, Eleanor. "Microenterprise Support and the Double Bind of Gender,"
Unpublished paper. 1994.

 

187