THE RELATDNfiHiP CF LNTEN‘HONS TO BUY AND :SUSSEQUENT PURCHASES GtF FARM MACHINERY Thesis in»: i’ha Dam-so of M. S. MiOHIGAN STATE UMVERSITY Leiand D. Lamberf €954. _ I" 1:15.515 LIBRARY Michigan State University ABSTRACT THE RELATIONSHIP OF INTENTIONS TO BUY AND SUBSEQUENT PURCHASES OF FARM MACHINERY by Leland D. Lambert This paper investigates the relationship between farmers inten- tions to purchase machinery and their subsequent actual purchases. A panel was questioned at the beginning of 1959 as to (l) the strength of their intentions to purchase, (2) the amount they in- tended to spend, (3) when they intended to buy, (h) whether they intended to buy a new or a used machine. Since the panel was the Michigan Mail-In Farm.Account c00perators, information was avail- able as to the farmers actual purchases from the account records at the end of the year. The survey of intentions was limited to investments estimated to cost more than $500. This limited most of the analyses to seven of the larger machines: balers, bulk milk coolers, choppers, hay conditioners, tractors, combines, and corn pickers. Both tabular and regression analyses were used to determine the correlation of purchases with intentions. The type of farm oper- ation and income variables were considered in the multivariate analyses. A single equation model was tried for the multivariate analyses and found to be inadequate. The "twin-linear" model which was used, estimated the probability of purchase with one equation and the size of purchase with a second equation. The analyses indicated that the probability of a purchase being made and the size of purchase are dependent partially on different variables. The results were found to be significantly different for different machines. THE RELATIONSHIP OF INTENTIONS TO BUY AND SUBSEQUENT PURCHASES OF FARM MACHINERY By Leland D. Lambert A THESIS Submitted to the College of Agriculture of Michigan State University of Agriculture and Applied Science in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Agricultural Economics l96h 'P' . I‘ -_ 4.... r. ACKNOWLEDGEMENTS The author wishes to eXpress his gratitude to all those who contributed to the organization of this project and in the pre- paration of the manuscript. 0f major importance was the guidance and patience of Dr. Warren Vincent throughout the project. Special thanks also are due to those who assisted with spe- cific areas of the project: Dr. Rdbert Gustafson and Dr. Lester Manderscheid for statistical advice, Dr. J. B. Lansing on method- ology,‘William.Rub1e for computing, and the personnel in the.M,S,U} data processing department who prepared the data for computing. The author is also indebted to other faculty members and to fellow graduate students for suggestions and contributions. Finally, thanks are given to the Agricultural Economics de- parhment for the opportunity and financial assistance provided the author as a graduate student. ii TABLE OF CONTENTS Page ACKNOWLEDGEMENTS . . . . . . . . . . . . . . . . . . . . ii LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . v LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . vi Chapter I. INTRODUCTION . . . . . . . . . . . . . . 1 Purpose 1 Need for Study 2 Survey of Literature 3 II. METHODOLOGY . . . . . . . . . . . . . . . . . . 16 Characteristics of Study 16 Source of Data 16 The General Approach 19 III, DATA PROCESSING’ . . . . . . . . . . . . . . . . 21 Questionnaire of Intentions 21 Data Adjustments 21 Machines Used in the Analyses 23 Data Processing Procedure 25 IV. ASSUMPTIONS AND HYPOTHESES . . . . . . . . . . . 27 Assumptions 27 Hypotheses 33 V. TABULAR ANALYSES . . . . . . . . . . . . . . . . 36 VI.MULTIVARIATEANALYSES.............1&6 VII. CONCLUSIONS . . . . . . . . . . . . . . . . . . 56 General Conclusions 70 Suggestions for Further Study 71 iii BIBLIOGRAPHY......................71+ APPENDIX........................77 Key to Symbols and Abbreviations 79 Coefficients for the behavior equations 80-102 Comparison of Machinery Prices 103 Questionnaire of intentions 106 iv IIST OF FIGURES Figure Page 1. Fulfillment rates by intentions categories ’43 2. Subsets of dummy intentions variables 51 3. Intercorrelations in the baler equations 67 LIST OF TABLES Table , Page 1. Percentage of total dollar purchases grouped by strength of intent to purchase and quarter of purchase . . . . . . . . . . . . . . . . . . 37-38 2. A comparison of mean intended and mean actual purchases . . . . . . . . . . . . . . . . . . . A1 3. Range of data . . . . . . . . . . . . . . . . . 1‘7 h. Equation 1: Tractor Expenditures 80 5. Equation 2: Tractor Prdbability 82 6. Equation 3: Tractor Expenditures 83 7. Equation h: Tractor PrObability 8h 8. Equation 5: Tractor Expenditures 85 9. Equation 6: Tractor EXpenditures ' 86 10. Equation 7: Baler Prdbability 87 11. Equation 8: Bulk Milk Cooler Probability 88 12. Equation 9: Chopper Prdbability 89 13. Equation 10: Corn Picker Probability 90 1h. Equation 11: Combine PrObability 91 15. Equation 12: Hay Conditioner Probability 92 16. Equation 13: Baler Expenditures 93 17. Equation 1h: Bulk Milk Cooler Expenditures 9h 18. EQuation 15: ChOpper Expenditures 95 19. Equation 16: Cambine Expenditures 96 20. Equation 17: Corn Picker Expenditures 97 vi 21. 22, 23. 2h. 25. Equation Equation Equation Equation Equation Hay Conditioner Expenditures Tractor Probability Chopper Probability Baler Probability Tractor Expenditures vii 98 99 100 101 102 CHAPTER I INTRODUCTION Since 1928, Michigan State University through the cooperation of the Agricultural Extension Service, the Agricultural Economics De- partment, and the Agricultural Experiment Station has worked with interested farmers of Michigan on their fem accounting and busi- ness management problems. Until 1957 the program followed tradi- tional procedures of gathering farm account books at the end of the year and processing the data obtained from them on hand calculators. Since 1957 two innovations have been employed: (1) information has been received monthly by mail and (2) data are processed currently using punched cards and electronic data processing equipment. The potential value of these records for marketing research was envisioned and a "Plans to buy” project was begun in 1958. This project utilized the cooperation of participants in the extension accounting program and was financed with research funds made avail- able by the U. S. Department of Agriculture. Purpose This study is a probe into the buying intentions of farmers and their subsequent purchases of farm machinery. Hepefully, the re- sults of this study may help to increase the power of mathematical l 2 models for predicting the demand for farm.machinery from.known variables such as: 1) Level of income. 2) Capital availability. 3) Profit expectations. h) Intentions to buy. 5) Machinery prices. Need for Study Economists recognize that there are a host of factors which af- fect a potential buyers decision to purchase or not to purchase. In the aggregate, many of these factors exert such a minor effect that it is generally considered unprofitable to attempt to measure them. There is a need methodologically to detenmine if the effect of some of these variables can be captured indirectly by consider- ing intentions (or the results of an intentions survey) as a proxy variable which will.measure the combined effect of many of these minor variables. At the micro level, there is a need for a better understanding and evaluation of the relative importance of factors that affect farmers decisions in buying. Such information would enable machin- ery manufacturers, machinery dealers, credit agencies and others to make better demand predictions. 3 SURVEYOFLITERATURE Most of the early research utilizing intentions1 in predictive work, did not include a means for evaluating the actual fulfillment rate of individual respondents (i.e., the ratio of intentions to purchases). Starting in 1927, the Regional Shippers' Advisory Boards of the Association of American Railroads surveyed finms in an effort to anticipate boxcar requirements to aid the railroads in planning shipments.2 Projections based on this study have been relatively inaccurate with an error exceeding a naive model,3 In l9h6, Fortune magazine incorporated a survey of anticipaticns into their "Forum.of Executive Opinion".h Forecasts, incorporating these anticipations, had errors about 23 percent smaller than the error of a straight extrapolation. In l9h7, Dunn & Bradstreet started incorporating sales expecta- lMuch of the early research in this area was summarized at the 1951 Conference on Research in Income and Wealth. The tepic for this meeting was "Short-Term.Economic Forecasting." The papers delivered at this meeting were published in Vblume XVII of Studies 1p Income gag Wealth, a report of the National Bureau of Economic Research, published by Princeton University Press, Princeton, N.J. 3Franco Modigliani and Owen H. Sauerlender, "Economic Exr pectations and Plans of Finns in Relation to Short-Term Forecasting," Studies _i_i_i Income and wealth, vol XVII (1955), 26h-267. 3For this paper a naive model is defined as one that pre- dicts that whatever happened last year will happen again this year, e.g., if tractor sales increased 10% in the first quarter of 1958, a naive model would predict a 10% increase in the first quarter of 1959. hModigliani and Sauerlender, op. cit., 267-27h. h tions into their routine questionnaires of the financial status of firms.5 The average error in forecasting from.their studies has also exceeded the naive model. The studies mentioned above do not give a direct correlation between anticipations and subsequent sales. The Shippers' report only gives information on anticipated physical shipments and subse- quent actual shipments. There was no information gathered on actual sales in the other two studies. One of the earliest studies correlating intentions and subse- quent purchases was conducted jointly by the Office of Business Economics of the Department of Commerce, and the Securities and Exchange Commission.6 This survey was begun in l9h8 and dealt with capital equipment only. Friend and Bronfenbrenner7 analyzed this study and concluded: There is a wide disparity in the accuracy with which individual businessmen anticipate their capital outlays, though in the ag- gregate the positive and negative discrepancies tend to cancel out. The degree of accuracy is related to many different fac- tors, including size of firm, amount of investment, and age of existing assets. The largest firms are much more accurate in their anticipations than the smallest firms. Similarly, firms planning large-scale investment (relative to existing assets) perform.better than those planning minor expenditures. It is also interesting to note that where existing plant and equipment is relatively old, firms are less likely substantially to cur- tail their planned expenditures. The predictive accuracy of this study was about the same as the 5Ib1d., 27h-277. 6Ibid., 3oh-307. 7Irwin Friend and Jean Bronfenbrenner, "Plant and Equipment Programs and.Their Realization", ibid., 55. Fortune study. A similiar study was carried out by the Canadian Government at about the same time.8 One of the earliest studies using intentions to purchase con- sumer goods was made by Lansing and Withey.9 werking with inten- tions to buy durable goods, they concluded: I) Predicting of aggregate purchases is much easier than pre- dicting the probability of an individuals actions. 2) Financial ability and change in financial status affects the fulfillment rate. 3) Trends and direction of change between surveys may be more significant than absolute percentage levels. A) The correlation between intentions and purchases was bet- ter for higher priced items than for low priced items. 5) Six months would prObably be a better time interval than one year. In a subsequent study10 Lansing (and Klein) concluded: we are convinced of the superiority of general forecasts for the economy. Ultimately we foresee a combination of survey data about the consumer sector with data from other sources in a model of the entire economy built for forecasting purposes... All three of the broad types of variables which we considered- financial, demographic, and attitudinal--proved to be important... In working with the attitudinal variables, we were particu- larly impressed with the importance of buying plans. The coef- ficient for this term.in the equation was highly reliable, amount- ing to almost A 1/2 times its own standard error. 80. J} Firestone, "Investment Forecasting in Canada", ibid., 113-259. 9John B. Lansing and Stephen B. withey, "Consumer Antici- pations: Their'Use in Forecasting Conswmer Behavior), Studies ip Income gag wealth, 0p cit, 381-hh0. 10L. R. Klein and J. B. Lansing, "Decisions to Purchase Consumer Durable Goods", Journal 9£_Marketi , vol xx (October, 1955), 109-132. 6 Irving Schweigerll made a general evaluation of intentions for use in forecasting and concluded: ...estimates (based on expectations) have generally been correct as to direction, (but) the indications of amount of change in demand have been very rough...Consumers' ability and inclination to plan purchases can vary over time as greater or lesser cer- tainty exists regarding availability of goods and credit, pro- spective incomes, etc. These factors widen the margin of error in interpreting intentions data. The experienced user can make a110wance for such factors. This characteristic indicates, how- ever, that intentions data cannot be handled in a mechanical fashion and that judement is necessary to interpret them. Robert Ferber12 made a general evaluation of some of the methods that might be employed to forecast sales of consumer durable goods by means of sample surveys. His most significant findings were: 1) The pepulation groups doing the most purchasing (on a per family basis) were also the ones doing the most planning. 2) Large items were more likely to be planned than small ones. 3) The planning horizon increased with the amount of contem- plated expenditures. h) The planning horizon varied by type of good. 5) Purchase plans were much more likely to be fulfilled if: a) the approximate time of purchase was known, b) they were accompanied by a high degree of certainty. 6) The majority of fulfilled plans were fulfilled not longer than one month beyond their scheduled date, where a date was given. 7) Fulfilled plans whose approximate timing was not known tended to be fulfilled even sooner than those for which approximate timing was given. llIrving Schweiger, "The Contribution of Consumer Anticipa- tions in Forecasting Consumer Demand", Studies i9 Income gag Wealth, vol xVII (1955), h55-h72. 12Robert Ferber, "Sales Forecasting by Sample Surveys", Journal 9: Marketin , vol xx (July, 1955), 1-13. 7 8) Degree of fulfillment of plans varied by type of good. 9) Degree of fulfillment varied with the respondents present and expected future financial position. Cromartyl3 made an extensive study of the factors affecting the demand for farm machinery. His study was based on census data for the years 1923-5h. Using multiple regression equations he con- cluded that: A 10 percent change in net farm.income has on the average resulted in a 5 percent change in the same direction of machin- ery purchases. A 10 percent change in the January 1 asset po- sition has resulted in a 3 to 6 percent change in machinery pur- chases also in the same direction. There is good evidence to show that a 10 percent change in machinery prices will result in a 10 percent change in the opposite direction for'machinery purchases. There is not sufficient evidence to conclude that a large stock of machinery at the beginning of the year will result in smaller quantities being purchased during the year. Nor is it possible to conclude from the results of this analysis that in- creases in farm.wage rates will result in more machinery being purchased, although higher industrial wages were associated with larger’machinery purchases. In the case of fanm.tractors, a 10 percent increase in net cash receipts for the previous year is associated with nearly a 2 to h percent increase in tractor shipments. Jean Namiaslh made a study of intentions and subsequent purchases of household durable goods. This study was based on data collected by the Survey Research Center of the University of Mdchigan in 1952 and l953.(this data were also used.by'lansing in his studies) She concluded: l3William A. Cormarty, T_h_e Demand :2; Farm Machinery and Tractors, Michigan State University Technical Bulletin 275 (Novemr ber 1959 . 1hJean Namias, "Intentions to Purchase Compared with Actual Purchases of HOuaehold Durables", Journal 2: Marketin , V01 at (July 1959) 26-30. 8 1) Consumers who say that they do 223 intend to buy a house- hold product during a given period seem more likely to carry out their negative intentions than peeple who say they'gp intend to buy. Nevertheless, most of the purchases are likely to be made by the group of consumers who do not plan to buy. 2) Fulfillment of intention to buy is prdbably, in large measure, predicated on income. 3) The larger the holding of liquid assets, the greater seems the probability to buy. A) The existence of personal debt does not seem to deter peo- ple from.buying. 5) Consumers who say that they intend to buy seem more likely to buy if they have a favorable attitude about their per- sonal financial situations, and express Optimism.about market conditions. 6) Consumers who live in towns, small cities, or the Open country probably are more likely to carry out their inten- tions to buy durable household goods than are consumers in big cities. 7) For peeple under A5, the presence of children in the fam- ily tends to be associated with greater stability of in- tentions to buy than in other families. wright and Vincent15 working with the same data as was used for this study, made a comparison of intentions and subsequent purchases of tractors. They concluded: 1) 0f the 935 farmers replying to the questionnaire in late December 1958, some 265, or 28 percent said there was "some chance" of them.buying a tractor in 1959, while 670, or 72 percent, said there was "no chance". 2) Actual purchases were made by 13h, or 50 percent, of the "some chance" men, and 91, or 1h percent, of the "no chance" men, for 225 tractor purchases. 3) Expenditures for tractors by the 50 percent "some chance" 15K. 'r. wright and warren Vincent, "Intended and Actual Trac- tor Purchases by Farmers in Michigan, 1959". .Michigan State Univer- sity Agricultural Experiment Station Quarterly Bulletin, Vbl hh, (November 1961) 33h-6o. 9 men, amounted to 65 percent of the total intended, as they paid.more than expected. Tractor purchases by the "no chance" men exceeded the deficit of the "some chance" men, so that total expenditures exceeded intentions by 6 percent. A) Of the 265 "some chance" farmers, 20 said they were "very certain" they would purchase a tractor, A3 were "quite certain", 96 said there was a "fair chance", and 106 a "slight chance". The percentage actually purchasing was as follows: 65 percent of the "very certain", 72 percent of the "quite certain", A9 percent of the "fair chance", and ho percent of the "slight chance". 5) When sorted in 1959 net income quartiles, 31 percent of the high income quartile men indicated "some chance” of purchaeing, 32 percent of the second, 23 percent of the third, and 27 percent of the low-group. As to percentage of those actually purchasing, 52 percent of the high-income quartile purchased, 52 percent of the second group, A3 percent of the third, and 53 percent of the low-income quartile. 0f the "no chance" men, 2h percent of the tOp- income quartile purchased tractors, lh percent of the se- cond group, 11 percent of the third, and 7 percent of the lowrincome group. 6) Combined purchases by‘both the "some chance" and the ”no chance" men in the high-income group was 129 percent of that intended, 108 percent in the second group, 95 per- cent in the third, and 81 percent in the low—income group. 7) Twenty-four percent of the 225 tractors purchased were bought in the first quarter, A5 percent in the second, 13 percent in the third, and 18 in the fourth. 8) Total outlay for tractors by all men was 33 Percent less than intended in the first quarter of the year, 15 per- cent above in the second, 36 percent above in the third, and almost four times as much as the small amount intended in the fourth quarter. 9) There was little difference in the percentages of the var- ious "strength of intent" groups actually buying tractors that intended to, whether sorted by 1959 or 1958 income. wrightl6 made a subsequent study of machines other than tractors. 15K. T. Wright, "Purchases of Major Farm Machinery”; Research report no. 3,.Michigan State University Agricultural Experiment Stat tion. 10 He concluded: 1) Farmers' actual expenditures for major’machinery in the year considerably exceed their January intentions. 2) Strength of indicated intent to buy is a strong factor affecting actual expenditure per farmer. 3) About one-half as many "no chance" farmers actually buy as "some chance" men, but spend one-third less per farmer. h) Predicting the time of the purchase, based upon indicated January intentions, cannot be made with.much reliability, especially beyond sixzmonths. 5) Higher net cash income the previous year is associated with stronger intent to buy and higher’machinery purchases per farmer. 6) Income level of farmers the previous year is somewhat more closely related to actual purchases than income level in the current year. 7) Strength of intent to buy, as indicated in January, and income level the previous year, are significant factors affecting actua1.mmohinery purchases, but there are other important factors also having an influence on purchases. Therefore, predictions on future expenditures for major machinery (other than tractors) by a group of farmers, based on knowledge of strength of intent to buy and income level the previous year, appear to have only a.moderate amount of reliability. Fisherl7 made a study of the relationship between consumer durable goods expenditures and the three variables: assets, credit and intentions. This study was also based on data from.the 1957 and 1958 Survey of Consumer Finances dealing with purchases of durable goods. A three stage estimation process was used in which the first stage dichotomized purchasers and non-purchasers, the second stage 4% 17Janet A” Fisher, "Consumer Durable Goods Expenditures, ‘With.Major‘Emphasis on the Role of Assets, Credit and Intentions", Journal 9:.thg American Statistical Associgtion, V01 58 (September, 1963) 6h8-57. 11 dichotomized cash and credit purchasers and the third stage esti- mated the size of the net outlay. Fisher concluded: The results suggest that sensible, but not simple relation- ships do obtain between certain regressors representing assets and liabilities and purchasing behavior, and that past behavior does provide some extremely helpful clues to the future. The results of this analysis for 1957 also support and extend pre- viously found relationships between purchasing intentions and subsequent behavior." Huang18 made a study of the demand for automobiles using a statis- tical approach similar to that of Fisher. Huang termed his method a "twin-linear estimation technique". The first stage estimates the prObability that a purchase will be made and the second stage estimates the size of the purchase. This study was also based on data from.the 1957 and 1958 Survey of Consumer’Finances. The author (Huang) was interested in estimating the "inventory effect", the "taste effect" and the "trade-in effect"19 associated with purchases of new automobiles. Huang concluded: The consumer's net investment may or may not display the traditional stock effect; we must consider the character of his initial stock as well as his Option to purchase new or used durables and to make a trade-in. There also needs to be more rigorous and detailed treatment of the effect of taste than has so far appeared in the literature. It seems that a preper emr pirical approach to the problems in this area requires simul- taneous use of cross-section, panel and aggregative time-series data. 18Dewid s. Huang, "Initial Stock and Consumer Investment in Automobiles",Journa1 of the American Statistical Association, Vbl l9'rhe trade-in effect is the effect of inventories on the ability to purchase, i.e., a person with a late model used car can purchase a new car with less cash outlay than a person with no car inventory. 12 .Tobing0 made an evaluation of intentions and attitudes for pre- dicting expenditures. His study was based on data from the 1953 Survey of Consumer Finances. In addition to intentions and at- titudes, he included as Objective variables: (1) current income,‘ (2) change in liquid asset holdings from the previous year, (3) change in personal non-mortgage debt from.the previous year. He concluded that intentions made a significant contribution to pre- diction but that attitudes were of questionable value. Tobin's article was criticized by Katona21 and by Fisher22- They found evidence that attitudes were more important than inten- tions for prediction. MMeller23 made a comprehensive survey of the record of forecasts utilizing attitudes and intentions. She concluded: In summary, the analysis indicates that discretionary spend- ing by consumers is determined to a large extent by income level and the state of consumer optimism.and confidence...If, as the data suggest, attitudes reflect the impact of more environmental factors than merely income change, and if complex combinations of these factors have a bearing on spending decisions, it fol- lows that consumer spending is not wholly governed, nor well pre- dicted, by the traditional financial variables...When attitudes 20James Tobin, "On the Predictive Value of Consumer Intentions and Attitudes", The Review 9: Economics gpg Statiptics, Vbl XLI, (February, 1959) 1-11. elGeorge Katona, "On the Predictive Value of Consumer In- tentions and Attitudes: A Comment", The Review 9; Economics gpd Statistics, Vol 111 (August, 1959) 317. 22Janet A. Fisher, "Something More 'On the Predictive Value of Consumer Intentions and Attitudes'", The Review.gf Economics gag Statistics, Vbl XLI (August, 1959) 317-319. 23EvaiMueller, "Ten'Years of Consumer Attitude Surveys: Their'Forecasting Record", JCurnal of the American Statistical Association, Val 58 (December, 1963), 899-917. 13 are also taken into account, the predictive performance of buy- ing intentions is not consistent from one test to another. 'Wu2h applied a two stage decision model to the theoretical hypothe- sis of stock adjustments. The probability of a purchase being made is estimated by the first stage and the size of purchase, given that a purchase is made, is estimated by the second stage. The general stock adJustment hypothesis is that: qt = C((8§ - Bt-1) + dt, where A qt 3 gross expenditure on durable goods in period t. st - desired stock at the end of period t. st-1 - actual stock at beginning of period t. dt - depreciation in period t. O(.- adjustment coefficient. various proxy variables were used by wu to measure st. These includ- ed marital status, home ownership, number of children, income, change in income and others. ‘Using data from.the 1958 and 1959 Survey of Consumer Finances, the coefficients of multiple determination were .0955 and .1106 for the prObability equations and .1359 and .1166 for_ the expenditure equations for the respective years. wu concluded that: ...the determinants of probability of purchase and of net outlay are not completely the same. Many variables which show significant effects in the prObability function do not appear to be significant in the net outlay function...0ne possibility is that it is the relative gap between desired and actual stocks which is important in determining the prdbability of purchase while the absolute gap is important in determining the net outlay. 2hDe-Min wu, "An Empirical Analysis of Household Durable Goods Expenditure". Unpublished paper presented at the winter Meetings of the Econometric Society in Boston, Mass., December, 1963. 1h Looking at previous research as a whole, there has been a trend toward viewing intentions and attitudes as having increasing impor- tance in demand analysis. There has been a recent controversy as to whether intentions or attitudes have more power for predictive purposes. Except for the wright-Vincent studies, the previous research has been weak in several areas. There is a need for additional research to fill in these gaps which this study endeavours to accomplish. Two of these weaknesses are in the reinterview process. Mbst of the previous studies did not get a quantitative measure of the de- gree of fulfillment, the respondents were only asked: "Did you or did you not make the intended purchase". All of the studies (with the exception mentioned above) which Obtained fulfillment data, got their information concerning the fulfillment of plans by a reinter- view process. This prOCess has the following disadvantages: l) The information obtained may be inaccurate either from erroneous reporting or forgetfullness of the respondent. 2) Measuring fulfillment by a reinterview’may bias subsequent surveys, especially if the reinterview is made a short time after the survey of intentions. This study avoids both of these prOblems by measuring the respon- dent's fulfillment rate from his accounting reports. These reports are mailed in monthly and the respondent has no knowledge that his fulfillment rate is being measured, thus subsequent surveys are not biased by reinterviewing, nor by forgetfullness. The survey of intentions for this study was, to the respondent, 15 merely a part of a larger prOgram.which is a service to him, thus it is to be expected that rapport with the respondent would be su- perior to the conventional panel. The high rate of response (89%) is an indication that this was the case. CHAPTER II METHODOLOGY CHARACTERISTICS OF STUDY In this study more attention was given to the "cutting-point prOblem" than has been the case in previous research. Lansing?h describes the prOblem as follows: (he is discussing a study in which there were only three classifications of strength of intent) This method of measuring expectancies gives rise to the so-called gpttingtpoint problem. ‘Should one assume that only those categorized as "definitely will buy" are actually going to purchase? If not, should one include the entire "probably will buy" group in one's prediction or a fraction of them? If a fraction, then what fraction? The predictor needs to decide on some point on the scale so that persons above such a point are going to be regarded as future "pur- chasers" and those below that point as "non-purchasers". Or he has to devise a formula with fractional predictions from each grouping. The customary solution has been to pre- sent the entire scale and base one's interpretation on trend data using the entire column. For this thesis the strength of intent was broken down into five categories, the respondent indicating his intent in terms of vary- ing probability that he would make a purchase. SOURCE OF DWTA History of Mail-In Accounting Project The data for this study came from the Michigan mail-in farm 2%. B. Lansing and s. B. withey, pp. gig, 1:16. 16 17 account program, The University started a farm.record project in 1928 which has Operated continuously since that time. This program was carried out in various ways until 1957 when it was converted to a mail-in type farm.accounting system. With the increasing importance of accounting records for tax purposes and as an aid to better management, the number of cooper- ators in the program expanded to a maximum.of 1700 in 1957 but has stabilized at about 1150 during the 1960's. HOwever, service to farmers is only one objective of the program. Another Objective of the accounting project is to train extension agents in farm.man- agement and as a vehicle for getting specialists out on farms. The record summaries and the farms are used for case studies, class visits, tours and special research projects. Publishing the sumr mary of each year's records provides a continuous source of input- output data and reasonable standards of performance for'Michigan farm conditions. Participation in the program is voluntary and at the time of this research the cOOperators were charged merely a small fee cover- ing the cost of materials needed in the operation. Mechanics of 0peration25 1) Farmers are enrolled in the project by the county agri- cultural agent. 2) The cOOperators mail in monthly, an itemized statement of financial transactions on uniform ledger sheets. ities of Mail-In Accoupting,.Michigan State University Ag. Econ. Mflmeo‘857 (Sept. 26'27, 1961) The details of this prOgram.have changed somewhat since 1961. 18 3) When the reports are received at the University, each transaction is coded and punched into IBM.cards. h) At the end of the year, an accounting summary is prepared from the IBM cards and a cOpy is mailed to the farmer. 5) Farm management specialists visit all counties and with help of agents "check-in" all cooperators. This includes inventory records, crOp production records and additional information. Questions arising out of monthly reports are clarified. Also at the end of each year, the University Farm.Mnnagement specialists compile a comparative summary for each of 17 areas of the state. These summaries contain information about: 1) The size, organization and Operation of commercial farms in the area. 2) Trends that are taking place on commercial farms. 3) The range in gross farm.income, expenses, net farm.income, labor efficiency, etc. h) Factors associated with profitable fanm management. These summaries provide a basis for making a comparative analysis of an individual farm. A cOpy of this summary is mailed to each cooperator. If the Farm Management specialist believes a special condition exists on a specific farm, he may exclude the data on this farm from his area report. The general criteria for the separation is as follows: 1) If he'believes the accounting report may be inaccurate. l9 2) If peculiar conditions such as fire, disease, sickness, etc., on a farm.in a small subsample caused an extreme change in the average results from.one year to another. 3) If the type of Operation was atypical, such as a muck farm, which would contribute to misleading area averages. For part of this study, the data which was excluded from these area reports was also excluded from the analysis. There was a total of 887 farms included in the 17 area reports. THE GENERAL APPROACH A brief summary of the general approach used in the study follows: The members of the mail-in accounting project were surveyed'by mail as to their intentions to purchase farm.machinery during the following year. The respondents were asked to classify the strength of their intention to purchase into one of the following categories: 1).E2£Z certain - have already started or am.making arrange- ments. 2) ngpg certain - considerably better than 50/50 chance. (of making purchase). 3) Egg; ghgpgg - about 50/50 chance of making purchase. h) glighpighapgg - less than a 50/50 chance of making a purchase. 5) Eglghgggg - of making a purchase. The respondents were also asked to indicate the quarter of the year in which they intended to purchase a new or used machine, and the 20 amount of the estimated expenditure. These intentions were then classified by strength of intent, quarter of intended purchase, and whether the intention was to pur- chase a new or used machine. Tabular analyses were then made of the data to get a preliminary indication as to which variables should be considered for multi- variate analysis. Then each of seven machines was analysed with multiple regression using a two equation model similiar to those used by Fisher26 and Huang27. This model, is elaborated in detail in Chapter VI . 26Janet A. Fisher, 2p. g._i_t_. 27David s. Huang, 92. _<_:_i_t. CHAPTER III DATA PROCESSING Questionnaire of Intentions A questionnaire of intentions to buy was mailed to all of the mail-in account COOperators on December 22, 1958. There were 10h2 or 89% who returned the questionnaire. There were 935 of these who both returned the questionnaire and completed the 1959 accounting year. The questionnaire asked for purchase plans concerning major in- vestments (arbitrarily defined as a purchase in excess of $500). This included buildings and equipment as well as machinery. A D cOpy of the questionnaire and the accompanying letter appear on appendix pages 106-110. The analysis of the data required that an inference be made as to whether the respondent purchased his machine new or used. In 2h cases it was not possible to make a reasonable inference from the questionnaire of intentions and the accounting reports. Ten of these respondents were contacted by phone and the remainder by mail to Obtain this information. The telephone contacts were made in January 1961 and the mail contacts in February 1961, DWTA.ADJUSTMENTS Some of the accounting records were incomplete since the 21 22 coOperator failed to complete the year. The intentions and incom- plete purchase data for these respondents were excluded from.the analysis. In a few cases the respondent indicated an intent to buy a com- bine and/or corn picker and subsequently purchased a uni-harvester. As it did not seem reasonable to consider these as unfulfilled in- tentions, these intentions were reassigned to uni-harvester. On a few of the questionnaires, the respondent indicated an in- tent to purchase but failed to indicate dollar intentions. In these cases the mean intent for that machine was assigned as the best estimate of the individuals intention, In separating the purchases into new and used machines an in- ference was made using the following as clues: 1) In many cases the accounting report read "new'machine purchased". 2) Whether the respondent intended to buy the machine new or used. 3) The size of the purchase in relation to the size of the intention. h) The price paid in relation to the manufacturers list price for the model purchased. As indicated above, in those cases in which a reasonable inference could not be made, the respondents were contacted for the infor- nation. It is possible that some of the farmers may have failed to list, on their*mail-in accounting report, machinery traded in. Most of 23 these errors would have been detected at the end of the year since a check is made with the farmers machinery inventory. Also, if a very large prOportion of the purchase price was represented in the trade-in, the purchase price would have varied from the retail price and this would have been detected in drawing the inference as to whether the machine was new or used. Even if such an error were made it would not effect the aggregate purchase figures and would not effect most of the analyses which follow. There were several farmers who purchased two machines of the same kind. In those cases in which there was no intent to purchase either machine, or an intent to purchase both machines, the purchases were summed and treated as a single purchase. In those cases in which one machine was intended and the other was not, the farmer was "divided" into two Observations, one of which purchased as intended, the other making a purchase without intentions. In these cases, the degrees of freedom was reduced by the amount that g was inflated. There were twenty of these cases, fifteen of them being tractors. MACHINES USED IN THE ANALYSIS The original "Plans to buy" project was designed to study "Major" farm investments. By definition a major farm.investment was consid- ered to be any purchase costing $500. or more. It is probable that many machines, such as grain drills and manure spreaders, could cost more or less than $500., depending on the size and model. It is pos- sible that the respondent intended to spend less than $500., but actually spent more than $500. In such a case he would not have . . . . u , o ' . . .7 n u a ._ > , A "a . . , ,. . . i . , . _ , ‘ a v i A r 7 , r ,. . a . ‘ . . u . . o ' n o o 2h listed an intention to buy but would be tabulated on the records as making a purchase in excess of $500. For part of this analysis, such errors would seriously bias the results, for other parts of the analysis this possibility would not be important as the correlation is between only those respondents who had intentions and their subsequent purchases. The machines which do not normally cost $500. at retail were omitted. The remaining machines were classified into three cate- gories: Group I - There is a high degree of certainty that the machines in this category cost more than $500. at retail. The machines in this category are: baler, combine, corn picker, hay con- ditioner, tractor, bulk milk cooler, chOpper, uni-harvester, and picker sheller. These machines were included in all of the analyses. Group II - This category includes gutter cleaners and silo un- loaders. It was felt advisable to exclude these machines from part of the analysis for the following reasons: Gutter cleaner: There was no way of determining whether the inten- tion and/or purchase was for a complete unit or for only part of a unit. this was also a prOblem.for milking equipment and wagons). Silo unloader: .Many of the respondents indicated an intent to buy a silo for I dollars and subsequently bought a silo unloader for that figure. It seems reasonable to as- sume that some of these respondents actually intended to buy a silo unloader but, on the questionnaire of intentions, 25 indicated an intent to spend X dollars on the silo in the form of a silo unloader. These machines were included in part of the tabular analy- ses but were excluded from.the multivariate analyses. Group III - This category includes those machines which could cost more than $500. for the most expensive type and model, or could cost less than $500. for a less expensive type and model. The machines in this category are: grain drill, manure spreader, pipeline milker, corn planter, manure loader, and wagon. These machines were excluded from most of the tabular analyses and all of the multivariate analyses. DATA PROCESSING PROCEDURE After the survey of intentions was returned by the respondent, the infonmation was punched into IBM cards. At the end of the year, the following data was transferred onto these cards from the mail- in accounting cards: 1) The actual dollar purchases of farm.machinery. 2) The 1958 and 1959 income. 3) The month of purchase. An inference was then made as to whether the machine was pur- chased new or used and with or without trade. This data, and also the purchase of part interest, was added to the cards. Thus the data cards contained information concerning the respondents: l) intended purchases in dollars. 2) actual purchases in dollars. 26 3) 1958 and 1959 income. h) strength of intent to purchase. 5) quarter of intended purchase. 6) month of actual purchase. 7) intent to purchase a new or used machine. 8) actual purchase of new or used machine. 9) type of machine intended to purchase. 10) type of machine purchased. 11) type of farming Operation. The tabular analyses were then made by utilizing IBM card sorting equipment. A preliminary regression analysis was made using a Control Data l60h computer. The remainder of the regression analy- ses were computed with a Control Data 3600 computer. CHAPTER IV'- ASSUMPTIONS ANDDHIPOTEISES Assumptions It is recognized that a large number of factors affect purchases, in addition to the ones measured in this study. For purposes of the statistical analysis, we formally assume that the net combined effect of these unmeasured factors can be treated as a random "disturbance", the distribution of which is the same from year to year. That is, if an unstudied variable biases purchases, then it is assumed that it biases purchases by a similar amount every year. Thus the regression coefficients from one year'may be used to predict purchases for the following year. Some of these factors are listed below and are classified into two categories: A. The first category includes those factors which are believed to be constant, at least in the aggregate, and thus would not cause a significant difference between successive surveys. The factors in this category are as follows: 1) It is possible that the process of the respondent recording his intentions may have some effect on his fulfillment rate. If this effect does bias the results, it is assumed to be a constant. 2) If the survey of intentions had been taken in mid- summer rather than midwinter, the fulfillment rate would likely have'been higher on harvesting machinery 27 3) ll) 5) 6) 28 and lower for tillage implements. In comparing results from different surveys or surveys taken in different years, the questionnaires should all have the same mailing date to eliminate this variation. Some respondents may have forgotten to enter expenditures on their accounting report. It is unlikely they would enter expenditures which they never actually made. Thus, there would be a.bias downward. If it is assumed that such errors occur at random, then the bias would'be relatively'constant. Machinery appearing on the January 1959 accounting report may have actually been purchased in 1958 and the entry delayed. Also, purchases made in December 1959 may not have been entered on the accounting report until January 1960. If it is assumed that such cases occur at random, then the January errors would tend to balance the December errors. It is assumed that many machines were purchased because of an unanticipated failure of the existing machine. If it is assumed that in the aggregate such failures occur at a reasonably constant rate, then there would not be a significant variation between successive surveys. Conversely, it is possible that the respondent anticipated a failure which did not occur. Although this is much less likely, it is also assumed to be relatively constant. It is likely that there was some change in machinery 29 prices between the time of the survey of intentions and the reapondents decision to purchase or not to purchase. It is assumed that if there was a price change it did not affect the decision to buy or not buy. A comparison of factory list prices, as reported by the National Retail Farm Enuipment Association,29 for the most popular makes and models appears on appendix pages 103 to 105. The average change in price for the machines compared follows: Fall 58 to Fall 58 to Spring 59 Fall 59 balers +2.35 +3.2% combines +0.5 +2.9 corn pickers +1.1 +2.1 tractors 0.0 0.0 uni-harvesters -O.3 +3.h choppers +2.1 -0.3 Total for all machines +0.6 «+1.7 This increase is similar to the wholesale price change 29foicia1 Tractor and Farm ui ent Guide (compiled by National Retail Farm.Equipment.Association), Farm Equipment Retailing, Inc., St. Louis, Missouri. Prices quoted are F.O.B. factory suggested retail prices. Machines that had not been in production more than two years were ex- cluded. Some reports indicate that the average prices paid for farm machinery during the period increased, however, these reports did not separate out the increased production cost and utility of the machines arising from changes in technology. 30 reported by "The Farm Cost Situation" publication:30 In recent years, wholesale prices of farm machinery and equipment have increased about 3 percent during September-December. This was not true in 1959, when they rose less than one-half of 1 percent during the com- parable period. 7) The reapondent might not have been aware that an improved machine would be available and, therefore, did not intend to buy. It is assumed that this affect is reasonably constant from.year to year. 8) The respondent may have changed his plans for such reasons as: l) toll road severance. 2) rental changes. 3) change in crOp plans. h) added or reduced acreage. It is assumed that such changes in Operation are reason- ably constant in the aggregate. 9) The respondents plans may have been upset by a disaster such as fire, hail, accident, windstorm, sickness, etc., which resulted in unanticipated expenditures which upset his fulfillment rate. Also, the respondent may have re- ceived unanticipated income such as priZes, inheritance, etc., which could influence his fulfillment rate upward. If it is assumed that, in the aggregate, these influences occur at random.then there would not be a significant 30The Farm Cost Situation, Agricultural Research Service, United StEtEs‘D‘eEar-t‘Tent—f—o Ag'fioulture publication No. ABS 143-125 (PCS-28) May 1960. B. 10) 11) 12) 13) 31 variation in the aggregate fulfillment rate. The respondent may have changed his plans because the hired.man quit or his son was drafted into the service or returned from the service, etc. It is assumed that these developments occur at random. If some reapondents purchased machinery to emulate their neighbors, (which they had not otherwise intended to buy) this effect would increase the aggregate fulfillment rate. However, if such an effect is present, it is assumed to be reasonably constant. It is assumed that there was no change in Government programs, or anticipation of future changes, that had a significant effect on the fulfillment rate. In a few cases, the respondent purchased less than a full interest in the machine. In some of these cases the respondent may have intended to buy a full interest but actually purchased a part interest. Such an error would tend to bias the dollar fulfillment rate downward since the purchase would likely be less than the intended expenditure, however, these cases are likely to occur at random, and the aggregate bias in one survey would tend to equal the bias of the previous survey. The second class of assumptions includes those factors which are believed to change from year to year but which change at a reasonably constant rate. Many of these changes would be 32 discounted by the respondents when they indicated their in- tentions to purchase. The assumptions in this category follow: 1) The reapondent may have decided to adopt new technology that was not contemplated at the time of the intention survey, such as a green chopping program.or minimum tillage. If it is assumed that this occurs at a reason- ably constant rate, then there would not be a significant difference between successive surveys. 2) It is possible that a purchase may have been made because a custom.machine was not available as the respondent had anticipated. 0r conversely, the purchase may not have been made as planned because a custom.machine became available which the respondent had not anticipated. It is assumed if there is an aggregate secular change in this factor that it changes at a reasonably constant rate. 3) It is assumed that the equity requirements and other policies of the lending agencies were, in the aggregate, relatively constant throughout the year (1959). There was a slight increase in interest rates as reported by the publication "The Farm Cost Situation".31 A survey made by the American Bankers Association in September 1959, indicated 31U'.S. Department of Agriculture The Fanm Cost Situation, Agriculture Research Service, ABS 1:341:11 '(F'CS-‘2"7")',"N"'o“v."l959, ""27. 33 that bank rates to farmers have increased since last fall from 6.55 to 6.76 percent on non-real estate loans. If it is assumed that this slight change in interest rates did not effect purchase plans or credit availability, then there would be no effect on the fulfillment rate. Conditions Peculiar to the Year Since about 195M, the dairy plants have exerted pressure in the form of premium payments or loss of market in an effort to convert to bulk handling systems. This has, no doubt, had an effect on the fulfillment rate of this particular equipment. The "whole farm" soil bank farm.program.was in effect during 1959. This resulted in many marginal farmers "selling out"; which, no doubt, increased the supply of used machinery. During the Spring of 1959, the University Extension Service carried out an extensive educational program to acquaint farmers with the benefits of hay conditioning equipment. It is likely that this program.affected sales of this type of equipment especially among the farmers in the.Mail-In Accounting Project. HYPOTHESES It is recognized that a farmer's purchases of machinery are depen- dent on a large number of causes which vary in intensity and interact in a complex:manner. It is also likely that many of these factors are dependent on the personality of the fanmer. It is Obviously unprofitable to attempt to measure all of these variables. The 3h forecaster must equate the marginal cost of measuring an additional variable with the estimated marginal value of prediction gained from that variable . It is hypothesized that a significant part of the low level causes associated with purchases can be captured by measuring intentions to purchase. Stated in equation form we hypothesize that: Y " f(X1:Xa:-~:Xdlxd+1: -- uxk ka+1w--:Xn) where'Y . an individual's expenditure on a given item. Xl,XQ,...,Xd - those variables whose marginal value in prediction exceeds their cost of measurement. Xd+1,...,Xh - those variables which affect purchases but whose cost of measurement exceeds their value in prediction. It is hypothesized that some of the variables in the second cate- gory (Xd+1,...,Xk) can be shifted to the first category by using the proxy variable, intentions, to measure them. This paper concentrates on the variables Xd+1:---:xk: although some of the variables Xl,...,Xd are included in order to remove their affect. In more specific terms, it is hypothesized that: I. There is a positive relation between strength of intent and subsequent purchases. That is, as the strength of intent increases, the probability of purchase increases. II. There is a positive relation between the size of the in- tended purchase and the size of the actual purchase. As the size of the intention increases, we can expect the size of the actual purchase to increase. 35 III. There is a negative relation between: a) the length of time between the date of the survey and b) the fulfillment rate In other words, as the length of the planning span increases, the power of intentions data for prediction decreases. CHAPTER V TABULAR ANALYSES The data were aggregated to get some idea of the gross relation- ships involved. The following tabulations were made: A. Total dollar purchases made by each chance group as a percent of total purchases. This analysis includes data on both new and used machines purchased either with or without trade. The "no chance" group was broken down into two subcategories: a) those who indicated no chance of purchasing any machine, b) those who indicated some chance of purchasing same machine other than the one they actually purchased. The breakdown by quarters is on the basis of the quarter of purchase, 1.6., if the respondent had intentions to purchase anytime during the year and made a purchase in the first quar- ter, then that purchase was tabulated into the first quarter. It was not possible to classify by quarter of intent becuase the no chance group did not indicate an intent. The results are tabulated in Table 1. Observations: l) The proportion of total dollar purchases, of specific machines, made by those respondents who indicated some intention of purchasing, declined steadily from 70% in 36 37 Table 1. Percentage of total dollar purchases grouped by strength Some Chance of Purchasing Machine Intended 1 2 3 1+ 5 ' 6 W _" (2+3+1*+5) of very quite fair slight total purchase certain certain chance chance some chance lst quarter 16.56 30.57 1h.05 8.60 69.78 2nd quarter 5.08 9.75 19.19 12.50 h6.52 3rd quarter h.h0 5.h5 15.78 7.77 33.h0 hth quarter h.68 6.9h 6.66 13.39 I 31.67 38 of intent to purchase and quarter of purchase. 7 8 9 10 11 no chance some chance (7495 (3:95 this of purchasing no chance total total machine)? some machine anything no chance all 16.63 86.h1 13.59 30.22 100 30.35 76.87 23.13 53.h8 100 35.10 68.50 31.50 66.60 100 38.72 70.39 . 29.61 68.3h 100 *The respondents in this category had intentions to purchase one or more of the eleven machines included in the tabulation, however, the machine which was purchased was not one of those intended. 39 the first quarter to 31% in the fourth quarter. In other words, in the first quarter, 70% of the total dollar pur- chases were made by those reapondents who had intentions to purchase that specific machine. In the fourth quarter only 31% of the total dollar purchases were made by those who had intended to purchase that machine. This trend tends to support hypothesis III, 1.6., the shorter the planning span, the higher the fulfillment rate. 2) The proportion of total dollar purchases made by those respondents who indicated no chance of purchasing anything increased from lh% in the first quarter to 32% in the third quarter. (The fourth quarter percentage was only 30%). The trend for this group was not nearly as strong as for the some chance group. 3) The prOportion of total dollar purchases made by those respondents indicating some chance of purchasing some machine (not necessarily the machine intended) varies from 86% in the first quarter to 69% in the third quarter.32 Crhe prOportion in the fourth quarter was 70%). This is an indication that the respondents have a machinery "budget", 1.6., if they have intentions to purchase machinery, they usually purchase machinery, even though it may not be the machine which they indicated on the questionnaire of in- tentions. 32This proportion would likely have been higher if the study had not been limited to 11 machines. ho B. A comparison of mean intended and mean actual purchases. It was deemed desirable to know whether the actual purchases differed from intended purchases because the number of intentions differed from the number of purchases or because the size of the purchase differed from the size of the intention or whether both effects were operating. This tabulation measures the ratio of mean actual purchase to mean intended purchase. The following data was excluded from this analysis: 8) Those who purchased with a trade-in, since a purchase made with a trade would mask the relationship.33 b) Those respondents who intended to purchase with a trade-in. c) Purchases of used machines as the value varies with the age and condition. d) Those respondents who did not intend to purchase. The results are tabulated in Table 2. Observations: The average expenditure was about equal to the average intention for nine of the fourteen machines. Of the five 'deviants', the average purchase exceeded the average intention in four of the five 08368 . 33‘There was no way of knowing whether the respondent traded in the item intended. 41 Table 2. A comparison of Mean Intended and Mean Actual Purchases. Ave. purchase number of number of Ave. intention intentions purchases* baler 104.2% 9 10.5 combine 139.2% 3 3 corn picker --- 1 O hay conditioner 108.3% 32 22 tractor 113.5% 17 24 bulk tank 97.2% 10 11 chopper 95.4% 24 14 gutter cleaner 103.2% 17 5 silo unloader 103.0% 16 10 picker sheller) uni-harvester ) 96.9% 1 2.5 corn planter 80.4% 7 7 grain drill 118.4% 3 5 manure Spreader) manure loader ) 103.7% 11 13 pipeline milker 102.7% 7 9 wagon 104.9% 26 16 *The number of purchases exceeds the number of intentions in some cases as those reSpondents indicating intent to buy with trade but actually bought without trade were included in the calculation of mean dollar purchase. #2 C. Fulfillment rates by categories. This tabulation was designed to measure the effect on the fulfillment rate of the three variables: a) strength of intent, b) quarter of intended purchase, c) whether the intended purchase was for a new or used machine. The data was classified by these categories. As there were some respondents who failed to indicate the quarter of intend- ed purchase an additional column was added for the no response category. This made a total of forty categories. The following data was excluded from.this analysis: a) those respondents who did not report intentions to buy. b) the group II and III machines. (These machines are described on page 24). This tabulation was designed to measure the aggregate fulfillment rate for each of the forty separate categories. There was consider- ably more data in some categories than in others. For better visual interpretation of the results, a three-dimensional diagram.is shown in Figure 1. This figure should be interpreted as follows: 1) the fulfillment ratio figures were made prOportional to the lesser of the two quantities: a) aggregate dollar intentions b) aggregate dollar purchases 2) The figures in black indicate the purchase of new machinery “3 quarter let 2nd 3rd hth not quarter quarter quarter quarter indicated . I I I I l I [.17 I .er I I I I 917‘- I .85 I .70 I I L'I-‘l- /I_ _ 7u—” 7:- .” 7I_ 7—77— —" '— I 1.2+ I .36 I I I .w I}: _I;§3_ __I :‘_ _ _u___ __ “I _ __ __ /I /l /l /I /l I -22 I .72 I .n I .7, I .72 I'77 [57 I26 I .94 I /l— — 74fi—7a“ "7I“—7I__ - I .. IL I .62) I S? I .55 I .37 I .75 l.5l I .50 I .69 I .57 /__ ——7__7—_—7—-— Figure l .flguantity of data" maven 310,000 10,000 To 20,000 20,000 70 30,000 30 70 40.000 40 T0 50,000 OVER 50,000 very certain quite certain fair chance slight chance 3) 1I) hh as a prOportion of intentions, 1.6., the ratio: aggregate dollar purchases aggregate dollar intentions The figures in red indicate a similar ratio for used machinery. The figures for new machinery are shown on the front plane and used machinery on the back plane. These proportion figures are a measure of the degree to which a respondent in a given category can be expected to fulfill his intentions. Observations: 1) 2) 3) There is a weak positive (direct) relationship between the fulfillment rate and the strength of intent to purchase, 1.6., as the strength of intent increases, the fulfillment rate increases. However, the difference appears to break down into only two categories with the very certain and quite certain groups in one category and the fair chance and slight chance in the second category. There is a strong negative (inverse) relationship between the fulfillment rate and the length of time between the date of the survey and the date of the intended purchase, 1.6., the longer the planning span, the lower the fulfill- ment rate. This relationship tends to support hypothesis III. The fulfillment rate of used machinery is slightly more sensitive to the other two variables than is the fulfill- ment rate for new machinery. That is, as the degree of 1&5 certainty increases and the planning span decreases, the fulfillment rate for used machines increases faster and reaches a higher level than new machines. Summary: Although the data is sparse in some of the subcategories, the tabular analyses generally support the hypotheses. For most mach- ines, farmers do an accurate Job of projecting the amount of money they will spend on a machine providing they do make a purchase. The indicated strength of intent to purchase and the length of planning span are both indicators of the prdbability of a purchase being made. The tabular analyses indicated that all of the intentions variables should be included in the multivariate analyses. There was, also, an indication as to which machines should be considered for further analyses. CHAPTER VI MUETIVARINTE ANALYSIS The tabular analyses indicated that there were a limited number of Observations in some of the subcategories when the data were cross- classified. For this reason a decision was made to limit the multi- variate analyses to the seven classes of machines with the larger number of purchases. These machines were: (1) baler, (2) bulk milk cooler, (3) field chOpper, (h) hay conditioner, (5) tractor, (6) come bine or uni-harvester, (7) corn picker or picker sheller. The range of the data and the means for intentions and purchases for these ma- chines is tabulated in table 3. A single equation model was tried on tractors and found to be in- adequate due to the problem of indivisibilities in purchases, 1.0;, the equation predicted many purchases in the range 0-$800., yet very few purchases were made in this range since a tractor is not divis- able into increments this small. The model used is essentially a two stage process in which the first stage estimates the prObability of purchase and the second stage estimates the size of purchase given that a purchase was made. Both equations were assumed to be linear functions. The indepen- dent variables used were: A. Income variables. 1) Current disposable income defined as total cash receipts 1:6 h? NM.0HOH 00mm NOH 00.05HN 00mm 00H mm.m¢0H 00mm mm ~0.m00 00NH NOH N0.m0~H mmmm 00m 00.0mmm exam 000a 00.m¢0 00H~ mug some swam 3oH momeaousm «n.00NH 000m 00m mm.m¢NH 0005 00m H0.000H 000m 00m 0H.Hmm 00HH 00H mm.nomH 0mm~ 00H 00.000N 00mm 000a 00.000H 000m 00¢ some 5mg: 30H mcoauamusa poaamsm uoxowm no uoxoqm auoo umum0>uezuac= no meanaoo uouoeua nocoquapoou mam Homeoso uoHooo mafia masm uoamm megsome .muea mo omcmm .m canes 178 minus total case expenses plus purchases of machinery and vimprovements. Purchases of machinery and improvements were thus omitted from.oxpenses in order to arrive at the amount of money which was available prior to any purchases of machinery and improvements being made. 2) Disposable income lagged one year, 1.6., the 1958 diaposable income. 3) Change in disposable income from the previous year. Type of operation. It was deemed desirable to take into account the type of farm operation. This was difficult since most of the farms in the sample were dairy farms. However, an arbitrary separation was made with the dairy farms being subdivided. These types of Operation were then entered into the equations as a dummy variable system. The criterion for the separation was based on the arbitrary definitions used by the Extension Farm Management staff. The farms were grouped as follows: CrOp Farms: Farms with crops as a primary enterprise were placed in this classification. There were 90 farms in this category. AtypiCal Farms: As indicated earlier in the paper, this classi- fication includes those farms which had situations suffici- ently peculiar to be excluded from.the area summaries used for comparative purposes by the Extension Farm.Management Specialists. There were 196 farms in this category. 1+9 Grade A Dairy Farms: These farms had grade A milk production as a primary source of income. As a secondary enterprize these farms had either crOps or hogs or else no other secondary source of income. There were #59 farms in this category. Grade B Dairy or Other Livestock: As a primary source-of in- come these farms had one of the following enterprizes: (1) Manufacturing milk, (2) Retail milk, (3) Calf pro- duction, (h) Beef production, (5) HOg production, (6) Sheep production. There were 86 farms in this category. Miscellaneous: Nest of the farms in this classification had Grade A dairy as a primary enterprize but had something other than hogs or crops as a secondary enterprize and were thus excluded from the Grade A Dairy Classification above. Also, included in this category were those farms with one of the following as a primary source of income: (1) Poultry, (2) Horses, (3) Fur animals, (h) Labor off farm, (5) Timber production. There were 118 farms in this category. Intention variables. The intention variables (for the machine intended) were broken down into three sets of dummy variables. One set is concerned with whether the intent was to purchase a new or a used machine, another set deals with the strength of intent and the third set considers the length of planning span. These sets overlap with the dollar intentions set for the sub- set of respondents who did not have purchase plans. This con- 50 tributed to intercorrelations, eSpecially in the baler equations, since there were no intentions to purchase balers in the third or fourth quarters. The relationships between these sets can be illustrated as shown in Figure 2. For the three sets of dummy variables, one subset was dropped in each equation to avoid singularity. D. Intentions to purchase other machines. 1) For the probability equations this variable was measured as the number of other machines which farmer 3 had inten- tions to purchase. This number included both new and used machinery intentions but was limited to the seven machines used in the multivariate analysis. 2) For the equations estimating the size of purchase (hence- forth called the expenditure equations) this variable was measured as the number of dollars which farmer t_intended to spend on other machines. The preliminary regression analyses were made on tractors using the single equation model: Yt‘q'tzhxit‘mt where‘Yt - actual expenditure on tractors by farmer t, at :- constant term. Xi - independent variables used as regressors. ut a pOpulation residuals. These arise from the effects 51 newsman“ seasons pom: popsoucw seasons 30: cocoons“ ocwzoma pom: nos 30: Homage: QMHm£ a zsawpaoo mnm>= E] [J D .96 has Hafiz so» pmgp one so» Camppmo 3o£ opwoaoca .nowo nob A.Eopa some now ngsaoo oco ombv whoa 0p swam :oh on pnoEQsto pen: : sopH m smpH m sopH H smpH "szwbm mm eczema szom mm mmmme A.mH .mmsv Op owv .oocwflo 0% flu A.® .mosd on owv .ooswno 050m _HH anode come swap egos pmoo Han: sous: loses manesaocflv ezmzmeam gees mm.eo has has Has: 56% message on» one pass .oeoa ma page .Asom uses one 6H .HspOp 6H.||||||ua paces has op sesame H use .62 mu «.4 .cfinoommp onp o>onw $ psonw had on pooexo H one .mmw flu asauoompp w o>m£ on mean 50h on comb. D 302 D mooms so 36G pa hug on Edam :0h 09 hepEoomQ .nonsoboz .meOpoo mongopdom .umsmai :33. D QQSb .hwz .Hamae. flu 109 % % _U_U an. .2 D _U 62 8% DD 3- I “U oz mew mew oz mow Am mow. D U D D D D U D comb zoz womb 362 cows 362 mm: 302 D D D D D D. D U D D D D D D D D U D D U D D D D d 23H m sop... m as: H :53 .cpuoowap a poo |£pp3 pmoo Op pp poomxo 50% £058 303 poops opmopocp .copooaoo mp apnocanp on 09633 Eopp mono mom .366 .23 fl manage» obopw pasoEw opdoaqu .oopooa 1N0 mp owwhp muons Eopa Some mob menoowpp a o>w£ op swam soh on wooms no so: pp hep op swam ooh om mongoose .nopEo>oz .popopoo mongopaom .pm5m5¢ .hHSh scam .mmz .prd<. nous: .hpmsppmm .hnwsnew .pp hsn pnwpfi :oh xupnp soh con: psopw ophoaccp .Eopp some pom .mocmno om\om swap mmoH hapmpocpmcoo I zoocwno pnmpam: .OOGGQO Om\Om 0.5096 l =®OGd£0 .HHdrYH: ..JH 0min“ .NH .HH .OH 110 moomp mop oesowpo mpomfiopo mpcmEm>omoEp mcpwppon com hmospnooE pmppo .mLOpompp op pmm>cp 0p msmpo poopm oXmE op ommo ooh opooz Ahco ppv mpsoEEoo pm£3 DUDE] DUDE] DUDE] HHH Boflhorm HH Bthomm H EDMhomm I A.mp .mmoo op omv Ammoon opp omflp pmnpov .oosono oz "mqumoozmm mo Zeppoompmzoo mo aozamo maom mp mmmme mm D .mp “mzohmmpm hm 2H mmpppa am_op ooh pmoo op pp poooxo ooh SooE 3o: poopm opmopocp .poomopo zoom pom .mwcmpo om\om m smsp mmop I :mosmpo pswppm: .om\om pooom I :monmSo ppmh: .ooomno om\om cmsp pop Ipmp hppmpmopmsoo I zapmphoo oppodz .mpnmEomsmppm MCpme So no ooppmpm hnmmppm m>m£ I chmppmo hush: .pp op pppz oOh pmpp ohm ooh opmp I960 30: opmoposp .poonopo Some pom moo 0p capo ooh op pmnz bwpommopo Seemipop sEopoo mflmlombo . p coppmmod 0p ow .oocmpo 680m A 6 v accepp swap egos pmoo pap: nepe: go home mmopeppsn asap pooh op mpomhomm mszmooamm go zeppopmpmzoo MOh¢z ham op pppz ooh moozmno opp ohm pop: .ooop Sp mp pmgp .nooh pxos opp 2H .wH .hp .op .mH $1283? USE CI'ELY MICHIGAN STATE UNIVERSITY LIBRARIES O 3015 5716 3 1293