An Investigation of Policy Implications of a Polak Model on Botswana's monetary and Balance of payment system Novice Happy Fidzani Introduction When Botswana first got its independence in 1966, it found itself entangled in a number of complications. Not only was it a member of the South African dominated Southern African Customs Union Agreement (SACUA) but it was also a member of the Rand Monetary Area (RMA). Membership into these two organisations had great implications for the extent to which the new administrators, from whom so much was expected, could use economic tools to direct and control the development of the economy. Membership into the RMA denied Botswana the use of monetary tools as it meant that Botswana could not have a completely independent monetary authority of its own. For almost ten years after independence, Botswana had to depend on monetary policies that were independently taken by South African authorities. In 1976 Botswana broke away from the RMA and formed its own monetary authority which came to be known as the Bank of Botswana (BOB). The extent to which Botswana suffered from its membership in the RMA is clearly reflected in the following reasons that were given for breaking away from the RMA: (]) The ability to choose our own domestic interest rate pattern instead of having rates largely determined by the Johannesburg Money Market. (2) The ability to manage Botswana's foreign exchange reserves instead of having South African Reserve Bank to do it for us. (3) The ability to retain Botswana's savings for investment in Botswana, instead of routinely lending them to South Africa as a contribution to financing South Africa's development. (4) The ability to set our own exchange control regulations, instead of having the South African Reserve Bank to do it for us. (5) The ability of the Government of Botswana to place its funds with its own bank, instead of continuing to bank at one of the two Commercial Banks.! The establishment of the BOB in 1976 was therefore viewed as a means by which monetary tools could be used to steer the economy into the direction which favourable BOP position and economic growth could be achieved. Membership into SACUA meant free movements of goods from the relatively much more industrialised South Africa into Botswana. That, coupled with the neglect of the Agricultural Sector resulted in an economy which was not only to remain with a weak industrial base but ~n economy so highly dependent on imported food. It is clear from Table 1 that the ratios of imports to GDP are very high and they are in fact among the highest in Africa. This Table further indicates that a significant proportion of the imports are food stuffs. Althou~h the Botswana economy has mana/led to maintain a healthy BOP position since 1976, the trade account reveals that imports have always been more than exports. This implies that the capital account has played a major role in maintaining this healthy BOP position. If imports are ~oing to continue exceeding exports, a downturn in the capital movements mi/lht lead to disastrous consequences for the BOP position. Notwithstandin/l the fact that Botswana is hi~hly dependent on imports, Table 1 also shows thi't the ratio of exports to GOP is very high too. This is a major result of the rapid expansion of the Minin~ Sector. Table 1, therefore unambi/luously shows that Botswana is a highly open economy and it is therefore bound to be vulnerable to international economic instabilities. This therefore, calls for a carefully managed monetary and credit system. TABLE 1 Botswana Trade Flows and Import/GOP and Export GOP Ratios 1973-1981 in Thousand Pulas Year Imports Exports Trade Balance Food % of Import as % Export as % Imports of GDP of GOP 1973 114963 59,200 -55,](,3 15.ll 65.9 41.32 1974 125,418 81,990 -43,42R 16.9 70.79 44.99 1975 159,288 105,040 -54,248 18.2 69.50 50.11 1976 181,385 153,172 -28,213 19.6 67.25 50.00 1977 239,605 156,653 -82,952 18.2 73.20 45.48 1978 307,090 192,676 -114,414 18.3 70.59 54.69 !979 438,289 1980 537,592 367,253 391,235 -71,036 -146,357 16.9 !5.7 64.72 71.92 51.90 51.01 I 1981 690,308 356,302 -334,006 14.0 I N/A N/A I ------------- ---------------------------------~-- _____---I I Source: Derived from Republic of Botswana. Statistical Bulletin: March 1982. Central Statistics Office. Aims and Objectives In the light of the bold step that was taken bv the Government of Botswana by pullin~ out o~ the RM~ in orde~ to enable itself to ~cquire the power to use monetary tools :~i~~sldteh'rfesltlnc~o~ caJOle the economy, this paper seeks to investigate amon~ other eo owm~. 2 a. whether since the establishment of BOB (J 976) monetary related variables such as exports, capital movements and changes in net domestic credit have had any impact on BOP positions and economic growth. If they have, what are the policy implications of these effects, and for those which seem not to have had any impact, what explanation can be given for this? The impact of these variables on economic growth wiJl be traced through the GDP function and the import function will be used to trace these variables' effect on BOP positions. b. Parameters estimated in the above functions wiJl be used to determine the levels of strategic policy variables such as changes in domestic credit and money supply that would have been compatible with BOP equilibrium wiJl be determined and policy implications of these results will be examined and analysed. AnalyticaJ F,.amewo,.k and MethodoJogy This paper draws greatly from a model that was developed by J.J. Polak. This model was further developed and improved by him and L. Boissonneault and some of their IMF colleagues. I That model uses changes in domestic credit and elements of BOP and monetary variables to analyse changes in GDP and Imports. The model is based on the foJlowing assumptions which this Paper presumes are likely to hold in the Botswana situation. It assumes a constant demand for money function which, in the classical Quantity Theory of money, would imply a constant velocity of circulation of money. In a developing country like Botswana, where the market for financial assets is highly underdeveloped, the demand for money is mainly for use as a store of value and transactionary purposes. It is therefore likely to remain constant as the interest rate does not play any significant role in influencing the amount of money people would like to hold. The model further assumes that capital movements, exports and domestic credit are exogenous to the economy. Capital movements are assumed exogenous mainly because of the weak role played by the interest rate in the financial market. Domestic credit is assumed to be a policy variable which can be used by that Government through its control over Commercial Banks. An Outline of the Model The model is based on the foJloWIng seven economIc vanables: Gross Domestic Product (Y), Imports (M), the Stock of Money (MO), Change in Net Foreign Assets (CNFA), Chance in NET Domestic Credit (CNDC), Exports (X) and Capital Movements (CM). The first four variables (Y, M, MO, CNFA) are assumed endogenous to the model. The last three (X, CM, CNDC) are exogenous. The relation between seven variables can be summarised by the following four equations: mY 0< m < I . .(1) t I .(2) MOt' 0 < k < I k MOt = MO _ + CNFA + CNDC .0) t 1 • (4) CNFA = Xt + CMt - Mt •.• t 3 Equations 0) and (2) show a casual relationship between Imports and GDP, and GDP and Money Supply respectively. The last two are definitional or identities •. It is clear from the above that there is some simultaneity involved in the two causal equitations as both use endogenous variables as explanatory vari~bles. In Equation 0), while imports depend on GDP, GDP also depends on Imports. It IS but a fact that the rapid economic growth that took place in Botswana during the period under study, w':ls to a great extent dependent on imported machinery which played a major role In mining and infrastructural development. Another, but unusual way, GDP in Botswana depends on imports is through the SACUA revenues. Botswana's SACUA revenues come from a common pool of the Customs Union and depend on the ratio of her imports to those of the Common Customs areas as a whole. Botswana receives a first instalment of the revenue which has been accrued in a p,iven year, in the same year. The second instalment comes two years later. But since the second instalment includes adjustment for under-forecasting of the first instalment, marginal increases in imports are only reflected in marginal increases in revenue two years later. In the second equation there exists some simultaneity between GDP and Money Supply. Money Supply influences GDP through the role it plays in a country's Production Function3 and its influence on Aggregate Demand. On the other hand, the rate at which the GDP is growing dictates some changes in money supply. Equations (3) and (4) mainly show that money creation is equal to changes in net foreil'\n assets plus changes in net domestic credit (A MO = CNF A + CNDC ••.••• (5) ). Substituting Equation (4) into (3) shows that MO = MOt_1 + Xt + CMt - Mt + CNDCt •••••• (6). It follows from this that b MOt = Xt + CMt - Mt + CNDCt •••• • • (7). This equation clearly shows that imports are a drain to money creation through the term - Mt.4 This further implies that imports are a drain on GDP via their impact on MOt. This can be demonstrated by successive substitution of Equation (7) into (3) and then into (2) which yields Yt = ~ (MOt_ I + Xt + CMt - Mt + C:NDCt );( •.•..• (8). Since Mt is exogenous to the model and it appears on the nght hand Side as an explanatory variable, a reduced form of Equation (8) yields the following resutls,. Yt = 1r1 At + ,. 2 MOt_1 + Ut ••.••• (9) where At is the sum of all e~ogen~us vanables (Xt, CMt and CNDCt). This reduced form helps avoid unbiased and inconsistent estimates. The impact of all exogenous variables on Y t can therefore be traced through Equation (9). Succ«:ssive substitution of Equation (4) into (3), then into (2) and finally into Eguation (I) YIelds the follOWing reduced form ~or imports; Mt = 113 At + 1£ 4 Mt_l + 0t ••• • • • (10) where 1£3 and 11"4are Import Impact multipliers. This equation can therefore be used to trace the impact of the exogenous variables on imports. !his Paper seeks ~o us~ Equation (9) and (0) to investigate the magnitudes of the Impacts and ~he dIrectIOn ~f the ca~sality. of the exogenous variables on GDP and Im~r~s by usmg. the regressIon analysIs. ThIS approach is a major departure from the t~a~I~lonal practices of either first obtaining the annual values of m and k by direct dIVISion o~ Imports values by corresponding values of GDP and money supply by correspondmg GDP on ~O and Mt on Yt to get m and k respectively. These values of m and k ajre then fed mto a standard formula to get the estimated values of Imports and GDP. 4 Taking a regression of the reduced form has the following advantages over the traditional method: a. It avoids getting biased and inconsistent resutls which may result from regressinj1; endogenous variables on other endogenous variables. b. It gives leverage to trace in a much detailed manner the causality that exists between the three exogenous variables (Xt, CMt CNDCt) and GDP and Imports. This can be done by breaking down Equation (9) into: Y t ~ BO ~ BI Xt ~ BZ CMt ~ B3 CNDCt ~ Bit Mat_I ~ Ut .•..•. (II). Equation (10) can be broken down into Mt = AO ~ AI Xt ~ AZ CMt ~ A3 CNDCt ~ Ait Mt_1 ~ Ut .... (lZ). c. This approach is more flexible in that new variables can always be added to the analysis whenever it is felt that the model omits some other important variables. Statement of Hypothesis Equation (10) implies a linear relationship between Mt and At and Mt_I' Since At contains two major variables (Xt and CMt) which are not only sources of foreign exchange but also Important sources of demand, It IS expected that "It 3 will be positive and significant. The import dependency of the Botswana economy is expected to give a very high coefficient for At. Since Botswana did not have any foreign exchange and import restrictions during the period under study, it is also expected that CNDCt had a positive and a strong impact on imports since it was also a source of demand for imports. That is mainly due to the fact that Botswana has to import both consumer and producers goods. 1\ It in Equation (10) is difficult to preduct because Mt-l has two opposing effects in the Botswana case. As mentioned before, Mt is a drain on money creation, therefore, through its impact on Mat_I' Mt_1will have a negative impact on Mt. But as has already been discussed above, Botswana relies on SACUA revenues which are based on her level of imports. It is probable that past levels of imports, through the first SACUA instalment, will influence their future levels through both Income and Foreign exchange effects. It is expected therefore, that Mt_1 might have a positive impact on Mt. As to which of these two effects is stronger it is difficult to tell a priori. But since the monetary effect is not that automatic, as monetary authorities can always sterilize it, it is expected that SACUA revenues will have a stronger effect and therefore IfIt will be positive. Regarding Equation (9) Yt ~ "111 At ~ "if Z Mat ~ Ut it is expected that the combined effect of the exogenous variables (Xt, CMt CNDCt) will be positive and significant. Capital movements have been in form of real investment in large projects such as the Shashe project, the diamond mines, roads and some other infrastructural related constructions. In this respect capital movements contributed to real output. The three exogenous variables are also a source of demand as pointed out before and they are therefore bound to have a positive impact on GDP. Since Botswana experienced an export boom during the period under study (due to increases in beef and diamond prices), it is expected that they contributed to GDP growth. This export boom was channelled to increases in demand through salary increases and direct earnings to those households that own cattle. s The Polak Model in its reduced form, implies that the exogenolJs variables (CNDCt, Xt and CMt) have identical effects on Mt or Yt. This Paper rejects this and predicts that, since exports operate mainly through monetary and aggregate demand e~fe~ts, and CNDCt operates mainly through aggregate demand effects and to a limited extent through monetary effect but with no foreign exchange effects, th~ ~wo are not likely to have equal or identical effects on either Imports or GDP. Similarly, C~t have their effects through foreign exchange earnings and real output effects ~ut ~Ith limited monetary effects. It is unlikely therefore, that CMt will have Identical effects as those of CNDCt and X. Equations (J J) and (J 2) will be used to test this hypothesis. Sources of Data and Definition of Variables With the exception of GDP data, all data used for estimation purposes in thi:; Paper are from the International Financial Statistics (IFS). The GDP data were obtained from the Botswana Statistical Bulletin in annual form and own method based on some indices were used to break it down into quarterly figures. The method used will be explained in later sections of this Paper. Variables used are defined as follows: Net Foreign Assets (NFA) - These represent foreign assets of the banking system. They are composed of bank deposits due to foreign banks less deposits held by foreign banks plus reserves. Net Domestic Credit (NDC) - These include domestic credit offered by the banking system. Government credit to parastatal organisations is not included. Capital Movements (CM) - This covers all the items of the balance of payments that are not classified as imports, exports or reserves. The following formula was therefore used in this Paper to derive these figures, CMt = Mt - Xt + CNF At Money Supply (MO) - Includes demand deposits plus currency in the hand of the public. Quasi money is only included in M02' . G~oss Domestic Product (GDP - These are National Accounts figures given at market prices. As already indicated above, the GDP figures were not available in quarterly form and S? the fOllowmg method was used to interpolate for the quarterly figures. GDP annual figures by sector were obtamed for each year and sectorial indices based on sectorial ~uarterly .data were derived and used to break these annual figures. These sectorial !nt~rpolatlOns were then added up to give total GDP quarterly interpolations. The mdlces were derived as follows: The Mining Sector Index - Monthly mining output data was used to construct the quarterly weights. The Manufacturing Sector - Since the Botswana Meat Commission contributes more th~n 90% to r:nanufacturing, its monthly output data was used to construct quarterly weights for thiS sector. Construction - Due to the unavaila~i1ity of any data on that sector's quarterly production, monthly figures for equipment and machinery imports were used to construct the index. 6 Electricity and Water - Monthly figures used. Wholesale and RetajJ - Equipment and machinery fjgures were deducted from the total quarterly import figures and the results were used to construct the Index. Given the fact that almost all goods sold by wholesale are Imported, this method was not a bad way for estimating quarterly contribution of this sector to GOP. For the Agricultural Sector, Government Sector and Financial Sector, no quarterly data were available. It was therefore assumed that these sectors' contributIOn to GOP was in a linear trend and this trend was used to Interpolate for these sectors. Econometric Estimations and Interpretation of Results Times series data, such as those used in this Paper, are normally subject to. serial correlation. Serial correlatIOn understates the standard errors of the regressIOn coefficients resulting In the inflation of the t-statistics. This might result In the over optimistic conclusion that the explanatory variables have a significant effect on the dependent variable while In actual fact they do not. InvestigatIOns using the Ourban H test revealed that all regressIOn results In this Paper suffered from serial correlatIOn and so the Cochrane-Orcutt Iterative Technique was used to correct for this. a. The Import FunctIOn The first step In the estimatIOn of the Import functIOn was to estimate the Import aggregated functIOns and the results were as follows: Mt 0.10 0.304At 0.703Mt_1 t Ct Std. Errors (2.41) (0.0709) (0.083) t-Stat. (0.046) (4.292) (8.4':17) R.2 = 0.978 DW Stat. = 2.58 Of 15 II. The Aggregated ImEort Function in Real Terms In order to take into account inflationary effects on the model, real values for Mt and At and Mt_1 were used and the results were as follows: 0.006 0.455RAt + 0.526RMt_1 Std. Error (4.281) (0.078) (0.106) t-Stat (0.002) (5.682) (4.962) R2 = 0.918 OW Stat. 2.5 Of = 16 The estimation of Equation (19) both In nominal and real terms reveals a strong and positive relation between imDorts ami aggregated term At. This supports the hypothesis that an increase (decrease) In any of the exogenous variables will lead 7 to an increase (decrease) in imports. Since Equation (l0) is in reduced form, the At reltression coefficient of 0.304 implies that an increase in At by P 1.00 will lead to an increase in imports by PO.304: it is an import impact multiplier. The results show that the impact multiplier in the real terms case is larger than in the nominal terms case •. . The Mt_1 regression coefficient is also positive and statistically significant. at the 5% level. As explained above, this coefficient is expected to be negative due to the impact of imports on money supply. It has also been explained, however, that the accrual of SACUA revenues in a lagged form is likely to make this coefficient positive. The second step in the estimation of the import function was to disaggregate the above function and regress Xt, CMt, CNDCt mdividually in order to capture the individual impacts of these variables on imports. As in the above case, the regressions were run both in nominal terms and real terms and the results were as follows: a. In Nomin~erms Mt c 0.176 + 0.330Xt + 0.274CMt + 0.270CNDCt + •636Mt_1 + et Std. Error (2.661) (0.09) (0.0 I) (0.143) (0.09) t-Stat. (0.066) (3.669) (2.766) (1.882) (7.701) R2 , 0.970 DW Stat. = 2.500 Of = 14 b. ~~~~~~s RMt 0.246 0.507RXt 0.329RCMt+ 0.40 I RCNDCt+ O.522RMt_1 + et Std. Error 0.907) (0.095) (0.123) (0.167) (0.105) t-Stat. (0.063) (5.325) (2.678) (2.402) 14.9(5) R2 = 0.929 OW Stat. = 1.995 Of = 15 As expected, in both cases, the results show that exports, capital movements and domestic credit are all positively related to imports. They are all statistically slgmflcant m the real terms function. and CNDCt is not Significant in the ~omlnal terms function. The results further show that, of the three exogenous anables, exports have the strongest impact on imports. That is not surpnsmg as the Botswana mmm~ sector, which is exclusively an export sector is a major Contnbutor to GDP. ' Th d ' .. e Ispanty In the Impacts of exports. capital movements and domestic credit ~nMlmports clearly proves that the assertion by Victor Argy and Polak that Xt. eff t and CNDCt have Identical effects on imports is not supported by data. The ect IS Identical only In terms of the direction of the causalitv and not in terms of the magmtude of the causality •. The reason why CNDCt is not statistically significant in the nominal terms regression is not difficult to determine in Botswana's case. Unlike in most countries where both parastatals in Botswana do not get their credit from banks but from the Government through the Public Debt Service Fund. Furthermore, at least during the period under study, instead of borrowing from the Commercial Banks, the Government actually lent to Commercial Banks. It follows from this that the CNDCt data do not represent the total credit taking place in the economy. It is, therefore, bound to reveal a weak imoact of CNDCton imoorts. It is also possible however, that in a multiple regression like this one, the explanatory variables might be highly correlated with each other. It IS possible for example that, through their impact on the money supply, exports may be hil!.hly correlated to domestic credit. This may lead to multi-collinearitv which would inflate the standard errors of the regression coefficients resulting in low t-values. This further leads to the conclusion that some of the explanatory variables have no significant impact on the dependent variable while in actuality they do.6 In view of the likelihood of the existence of multi-collinearity .in the model, some tests usinl!, the correlation coefficient matrix and the Glauber-Farrer method were made and the results are shown in Tables 2 and 3. The results of the tests reveal that CNDCt is not highly correlated with any of the other explanatory variables but instead that Xt and Mt-I are highly correlated. It is clear from these tests that the t-values of Xtand Mt_1 are the ones that have been understated and not that of CNDCt• TABLE 2 Summary of the Correlation Coefficients Matrix Correlation Coefficients between Value Xt and CMt 0.4!~2 Xt and CNDCt - 0.0099 Xt and Mt_1 0.883 CMt and Mt_1 - 0.528 CMt and CNDCt - 0.351 CNDCt and Mt_1 0.158 9 TABLE 3 The Glauber-Farrer Test F-Value Conclusion Xt on CMt. CNDCt• Mt_1 22.33 0.78 Reject Ho. eM, 0n Xt. CNDCt, Mt_1 5.275 0.416 Reject Ho. CNDCt on Xt. CMt. Mt_1 3.44 0.28 Accept Ho. Mt_1 on Xt. CMt, CNDCt 31.336 0.834 Reject Ho. Ho: There is no multicollinearity The high correlation between Xt and Mt_1 (0.88) can be explained by a close examination of Equ ation (6): ~~Ot = MOt_I + Xt + CMt - Mt + CNDCt. This equation clearly shows that autonomous changes in imports will lead to changes in MOt. From this we can speculate that a fall in MOt resulting from an increase in Mt may lead to a decline in aggregate demand. The declme In aggregate demand may lead to a decrease in the domestic consumptIOn of traded goods which may finally lead to future increases in exports. [n this way, past chanp,es in imports will affect future levels of exports and therefore result In a high correlation coefficient between Xt and Mt_I' The only commodity that qualifies for this explanation in Botswana is beef as it IS the only maior export which is also consumed domestically. The third step in the estimation of the import function was to determine elasticities of imports with respect to exports, capital movements and changes in domestic credit. This was done by taking natural logs of all these variables and regressing the results accordingly. This approach has a slight advantage over the previous approach in that instead of giving import impact multipliers, it gives import elasticities which may be more useful m policy analysis. The results of the regression were as follows: 1. LnMt = 0.238 InXt + 0.094 InCMt + 0.016 InCNDCt + 0.717 InMt_1 Std. Error (0.079) (0.029) (0.019) (0.079) T -Stat. (3.02) 0.208) (0.87) (9.12) R2 = 0.972 OW Stat. = 1.75 Of = 16 As in the former case, the signs of the regression coefficients are as expected and all vanables except CNDCt are significant. [t is even clearer in this case that imports respond more to chanl1.es in exports than to changes in any of the other explanatory variables. The results indicate that while a I % change in exports leads to a 0.24% change in imports a I % change in capital movements leads only t<.>a 0.09% change in imports. Implied in these results is that economic policies directed towards changing exports will be more useful for stabilisation purposes than. those directed towards attracting foreign capital and changing domestic credit. b. The GOP Function Similar steps to those taken in the estimation of the import function were taken in estimating the GOP function. However, in order to take into account the proportionality that exists between GOP and MO as portraved in Equation (2) (Y = k MOt), all the regressions were done through the origin. The results are as follows: I. aa. The Aggregated Income Function in Nominal Terms Y= a.35At + 1.503MOt_1 + et Std. Error (0.154) (0.2244) (2.30) «(;.164) R2 0.931\ DW Stat. 2.132 Of 1(; bb. The Aggregated Income Function in Real Terms RY = 0.417RAt + 1.59RMOt_1 + et Std. Error (0.140) (0.257) t-Stat. (2.97) (4.51 Jl R2 = =.549 OW Stat. 2.26 Of 1(; In both the nominal and real terms estimations, the siJl,ns are as expected and the coefficients are statistically siJl,nificant. This confirms the hypothesis that exports, capital movements and changes in net domestic credit have a positive impact on GOP. The positive MOt_I coefficient signifies the importance of money in the production function and how monetary policies at a given period of time can affect future changes in GOP. The Real GDP Function shows a relativelv low R2. Normallv when R2 is low and t-values are high it indicates that some important variable has been omitted. An attempt to use the two stage least squares method to Include imports in the function vlelds some unsatisfactory results. A comparison between the At coefficient in the GOP function and that of the Import function reveals that chanJl,es in At have a stronger impact on GOP than on imports. While a PI.OO increase in At leadsto a PO-35 increase In GOP it will lead to onlv a PO.30 in imports. The closeness of these two coefficients illustrates the ~penness and import depenclencv of the Botswana economy as portrayed in the introduction. ii. aa. The Oisaggregated Income Function in Nominal Terms 0.557Xt 0.254CMt + 0.556CNOCt + 1.345MOt_1 (0.243) (0.183) (0.297) (0.2lB) Std. Error (2.29) (J .393) (J .874) (4.755) t-Stat. OW Stat. = 2.067 Df = 14 R2 = 0.946 11 bh. _ The Disaggregated Income Function in Real Terms RY= 0.8R4RXt + 0.392RCMt - 0.0&6RCNDCt + 1.0&RMOt_1 + et Std. Error (0.205) (0.21 J) (0.169) (0.259) t-Stat. (4.315) (0.1&6) (-0 •.506) (4.178) R2 = 0.067 [)W Stat. = 2.012 Df = [If Th" regression results for th" nominal income function shows the exp"cted signs except In the real Income function In which the CNDCt coefficient is unexpectedly negative. However, since this coefficient is not statistICally significant, we cannot conclude that the results imply that CNDCt is negatively related to GDP. In both functions capital movements and changes in domestic credit are not statistically significant at the 5% level. As explained in the Import functIOn, domestic credit data do not represent the total credit takln~ place In the economy. The insignificance of CMt on GDP can be explained by examining the way In which the CM data were derived. These data were derived by uSing the formula CM " CNFAt + Mt - Xt. Since CNFAt is likely to consist of Import credits to the Botswana economy, it means that CMt is also likely to consist of these. Since import credits increase imports Without necessarily increasing G[)P, It IS therefore possible that the impact of eMt on GDP is likely to be weak even though it is strong on imports. An investiRation for multicollinearity reveals a negative correlatIOn between Xt and CNDCt of -0.622 and a positive correlatIOn between CMt and MOt-l of 0.558. To the extent that parastatals do not compete for bank creeflt With the private sector, the main borrowers are likely to be big farmers and private companies. It is possible therefore that dunng periods when beef export earnings increase, biR farmers' demand for credit declines and during slack periods (due to foot and mouth disease) their demand for credit increases. ThiS might result in a negative correlation between exports and CN[)Ct yielding the large negative correlation coefficient of -0.622. iii. aa. The Nominal GDP Elasticity Function lnY = 0.270lnX + O.052lnCMt+ 0.03lnCNDCt + 0.852MOt_1 + ~ Std. Error (0.101) (0.036) (0.242) (0.1005) t-5tat. (2.67 J) 0.450) 0.242) (8.4S2) R2 " 0.950 DW Stat. " 1.886 Df 14 bb. The Real GDP Elasticity Function LnRY" 0.5171.nRXt +0,013LnRCMt+ O.OlOlnCNDC + t Std. Error (0.104) (0.048) (0.027) (0.118) t-5tat. (4.973) (0.274) (0.365) (5.526) R2 = 0.70S DW Stat. = 2.004 Df" 14 The results of the COP elasticity function reveal that in both cases COP is not elastic to changes in domestic credit and capital movements. Explanations for this which were given in earlier sections still hold in these two functions. A comparison between the COP and Mt function reveals that COP is more elastic with regard to export changes than imports. Since the reduced form of the Polak model is over-identified, m and k values which are needed for projection purposes could not be recovered. Therefore, the two-stal?;e-Ieast-squares method was used to estimate m and k and the results were: Mt -23.44 + 0.81 Yt t-Stat. (-1.552) (8.01) R2 0.727 OW = 0.764 Of = 18 Yt -9.01 + 2.133MOt+ e t t-Stat. (0.44) (7.71) R2 = 0.778 OW Stat. = 1.10 Both results show the expected signs and a strong relationship. As expected, the propensity to import is less than I. The propensity to import of 0.81 clearly shows how import dependent the Botswana economy is. The COP results show that k = 2.133 which means that as expected .~ is less than I (0.468). k c. Test for the Stability of the Two Functions The period after 1978 was characterised by major changes which are likely to have affected aggregate demand in the Botswana economy. OurinFl the period 1978/79 there was an export boom which was both a consequence of the increase in diamond prices and the increase in production. OurinFl the same period SACUA revenues increased by 50%1 (Budget Speech: 1979). Furthermore, as a consequence of the export boom and the political elections which were due to take place, there was a major upward revision of the salary structure in the economy. All of these factors are likely to have changed the propensity to import (m) and the velocity of circulation of money (k). Since the proportion of an increase in credit, exports and capital movements that leaks to imports and COP is directly related to the prooensity to import and the velOCity of circulation of money (Polak and Argy),8it means that if the two functions were affected by these changes, the leakages of credit, exports and capital will be higher than those portrayed by the estimated functions. It is therefore important that the stability of the functions should be tested before their results are used for projections and policy analysis. In this regard the dummy variable test for the period 1976-1978 and 1979-1981 was made and the results were as follows: 13 I. Import Functions: Mt = -6.546 + 1.99D +O.379At + O.073DAt + 0.737Mt_1 Std. Error (6.430) (10.047) (0.098) (0.105) (0.088) t-Stat. (-1.01) (0.198) 0.3884) (-0.686) (8.4) R2 = 0.977 DW Stat. - 2.83 Df = 13 RMt = 8.068 + 9.2620 + .462RAt -0.127DRAt+ 0.33Mt_l + et t-stat. (7.83) (9•.588) (0.100) (0.12.5) (0.039) R2 = 0.93.5 DW Stat. 2.137 Of = 14 ii. GOP Function: y= -42.189 + 77.42D + 0.728At - 0.729DAt + 1.699MOt_l Std. Error (16 •.554) (20.935) (0.195) (0.223) (0.284) t-Stat. (-2.549) 0.698) 0.736) (-3.269) (5.9it5) R2 = 0.972 OW Stat. - 1.83 Of = 14 RY = 21.238 + 43.26D + 0.291RAt+0.2870RAt+ 0,663RMOt_l + et Std. Error (16.731) (16.076) (0.158) (0.216) (0.248) t-Stat- (1.269) (2.691) 0.847} (-1.325) (2.679) R2 = 0.873 OW Stat. - 1.96 Of = 16 The insignificance of the 0t and OAt coefficients reveal that neither . the intercept or the scope of the import function was affected and so It remamed stahle. This implies that the propensity to import was not affected by the chanRes that took place during the period after 1978. The GOP function result reveals that it was only affected in nominal terms but not in real terms. This implies that those chanRes after 1979 contributed more to price chanRes than to real production. Policy Implications of the Results Balance of Payment positions of open economies'with undiversified export sectors, like that of Botswana, are usually vulnerable to fluctuations in capital movements and export prices. A sudden decrease in export earnings or foreign capital inflows would lead to BOP deficits which might have further adverse effects on other sectors of the economy. On the other hand, Equation (6) (MOt = MOt_l + Xt + CM _ Mt + CNDCt) t shows that an increase in exports and CMt will not only lead to increases in the BOP surplus but will also lead to increases in the money supplv. Once the money supply increases in excess of its demand, this will lead to price increases or/and increases in the demand for imports. If prices increase this may result in an adverse effect on the export sector and the general production level in the economy. It is therefore important that the monetary and credit policies must be used to guard aRainst adverse effects on both the BOP position and economic growth. The Polak model tries to provide a framework through which disturbances in BOP and money supply due to international fluctuations can be dealt with. According to the model, dOrrlp.stic credit is the only policy variable out of the three exogenous variables. Export and capital movements are assumed to be difficult to affect in the short-run due to the weak role played by the interest rate and structural rigidities. The manner in which domestic credit operates can better be understood by looking at Equation (10) (Mt = 113 At + "4 Mt_1 ) where At = CNDCt + X + CMt •..... (13). Substituting Equation (13) in (10) Yields Mt = "Ii 3 (Xt + CMt + CNDCt) + 11 4 Mt_1 ... . " (14). By Equation (4) we know that CNF At = Xt CMt - Mt and that at BOP equilibrium CNF At = O. Substituting Equation (14) Into Equation (4) and setting the result to zero and finally solvlnp, for CNDCt yields CNDCt = I -"3 (CMt + Xt) -"4 -1fT -'If) Mt_1 ..... (J 5) (IMF:1981). ThiS equatIOn determInes the level of domestic credit which Will be compatible with BOP equilibrium. Using the estimates of 1'3 and "Ii 4 this formula was used to determine the level of changes in domestic credit that would have been compatible with BOP equilibrium in Botswana during the periods under study. The results are as shown by CC in Table 4, Column 5. Substituting CC In EquatIOn (J 3) yields CCMt = '1\3 (Xt + CMt + CC) + "4 Mt-I which gives the level of imports which would have been compatible with BOP equilibrium. These results are shown In Table 4, Column 8. Substituting CC and CCM In Equation (6) yields CCMOt = MOt_1 + Xt + CMt - CCMt + CCt which gives the levels of money supply that would have been compatible with BOP equilibrium at given levels of Xt and CMt. The level of GDP that would have resulted at BOP equilibrIUm is determined by Equation (1): CCY = I~ (CCMO). The coefficient -k was estimated by USIng the two stage least squares prlbcedure and substituted Into EquatIOn (J) YieldIng the results which are shown In Table 4, Column 14. To facJlHate analysIs, Table 4 IS dIvided Into Part A, which shows results for selected periods during which there were BOP deficIts, and Part B which shows results for those periods during which there were high BOP surpluses. Domestic credit results (Column 4-6) show a clear distinction between what would have happened during the deficit periods in order for an equJllbrium to be reached and what would have happened during the surplus period. CC results In Part A indicate that the deficIt would have been removed by a decrease in domestic credit. Part B, on the other hand, indicates that an equilibrium would have been achieved by increasing c~edlt. These results clearly show that an expansion of domestic credit in excess of the demand for money will lead to BOP deficits. Once excess money supply develops the public Will try to get rid of it by Increasing their demand for Imports. By EquatIOn (4) (CNFAt = X + CMt - Mt) we know that when Mt increases CNFA will decrease eventually resulting in a BOP deficit. The economy, therefore responds to excess money supply by exporting its money through BOP deficits (Guitlan In IMF: 1977)51 Also implied in these results is the message that whenever an economy IS faced with a BOP deficit it must cut down on its domestic credit or ration it to the more strategic sectors. This will reduce the money supply and the demand for imports. The results also show that at given levels of exports and capital movements a BOP deficit may rise not because of changes in exports or capital movements, but because of excessIve monetary expansion. Theoretically, a BOP deficit will be self-correcting in that it will lead to a decrease in the money supply which will lead to increases in the interest rate which Will attract foreign capital. As already mentioned in earlier sections of this Paper, In many developing countries interest rates do not playa major role In the attractIOn of foreign capital and so their BOP deficits are not lik~ly to be self-correcting. IS Results in Part B of Table 4 clearly show that movin~ from a BOP surplus to an equilibrium requires an increase in domestic credit since domestic credit which IS compatible with the BOP equilibrium is lower than that for a BOP surplus. This implies that an expansion of credit or the money supply which fails to meet the economy's demand for money might lead to a decrease in the demand for Imports which may finally lead to a BOP surplus. The cost of such action can be seen by comparin~ Columns 14 and 13. These columns show that BOP surpluses are assoCiated with lower GOP than at BOP equilibrium. This has serious Implications for the Botswana Government's intention that, hecause of the high uncertainties it faces, It has to accumulate lar~e reserves which can be used during periods of high need. The results show that this accumulation cannot take place without a cost. Since BOP surpluses are associated with low GOP, one might conclude that surplus may, in the long-run, affect the general production in the economy. When this happens the accumulation of reserves will have killed the goose that laid the golden egg, the export sector. For imports and the money supply, Part A shows that BOP equilibrium is associated with lower levels of imports and money supply than those that exist at defiCit levels. Part B on the other hand, shows that moving from a BOP surplus to eqUilibnum entails an increase in imports and in the money supply. One important point to note from the Table 4 results is that although it is true that moving from a deficit to a surplus requires a decrease in credit and moving from a surplus to equilibrium requires an increase in domestic credit, it is not true that the larger the deficit the more decrease in credit you need. For example In 1978QIdeficits of P9.44m would have required credit to decrease by P40.159m while in 1981Q3 a deficit of P 15.0I required a decrease in domestic credit by only P31.353m. This is because, even though they are held constant, exports and capital movements still play a role in the determination of the actual level of credit that should take place. This implies that if we were to drop the rigid Polak assumption that in the short-run exports cannot be influenced, in those economies whose exports have a hil!:h supply elasticity and a low demand elasticity, devaluation combined with proper monetary and credit policies would give a stronger tool for stabilisation. In real life it is, however, one thing to determine an amount of domestic credit that can be compatible with the BOP equilibrium and it is another thin!?, to be able to actually achieve the set tar!?,et. It is much easier to decrease credit than to increase it. Credit can be easily decreased by using conventional methods such as increasing required cash ratios, fixing prohibitive lending rates and fixing credit ceilings. Increasmg credit on the other hand, involves providing incentives to borrowers who mayor may not respond to the incentives. So as far as credit policy is concerned, we can pull on the string, but we cannot push on it. Credit policy might therefore be useful In decreaSing the BOP deficit but not for reducing a BOP surplus. This point is further remforced by the fact that during the period under study, Botswana Commercial Banks experienced excess liquidity and so it would have been very difficult to decrease the surpluses by increasing credit. Conclusion In concluding, o~e might want to address the question whether, judging from the results of the estimation of the model for Botswana, the pulling out of ~swana from the RMA was of N - - '" - 00 0- 00 N a- N 00 It'\ ~ ..... -~ ..... N ~ 0~ It'\ 00 '" 0\ 000 -'" '" 4- N ..... N 00 - ::E U N -D 00 v:> 00 ..... ..... N '" '" "" ~ v:> 00 N - N -'" '"- '" '" N N 00 00 N 00 N v:> 0 v:> '" ..... '" N N 0 ~ ~ "- - '" ~ v:> N '" .,; - v:> >< -D '" '" '" 0- 0- '" N "- '" N '" - 0 00 -"" N - v:> ..... v:> '" .........: < -. -. < < v:> 0 "- 0 ~ 00 v:> a- ..... a.. < E -L: :J - '" <%>- .....t N '", , v:> Z Z ..... ~ ~ - -'" ....: -D 00 N ..... ~ ~ v:> @ a- - 0 - '" N 0- '" .,; '" v:> 0 "- 0- 0- 0 ("'I - 00 -'" - '" N N -3 <'i "" '"- N 00 ....: N <'i '" 00 >- 0- Cl) c.. -~ U U v:> N N v:> 00 00 0 00 N N - - .::: '" ..... ~ ....: 00 v:> 0 ..... N N 00 ~ ~ 00 v:> 0- ~ < < q - 0 ..... "" 00 - 0- .,; Z ~ ~ N --- ..... 0- ..... v:> N z ~ v:> ~ 0- 0- ~N - 00 N a.. ..... ~ ..... - Cl) 00 0 ~ D'! q 0 :c '" N .,; '" -;: III c.. - N 0 ~ N N .,; - -~ 0 N E "- ..... 0- 0 00 ..... '" '" 0 u c: Cl) - 0 ::E u U 00 "" '" N ~ - "" - '" '" v:> ....: 0 0- N 11'\ '" - '"- ~ '" N N 11'\ .,; -D 00 N '" .D Cl) 11'\ 0 v:> - ~ ......,; '"'".,; a- v:> 11'\ v:> N 11'\ 11'\ ~ 0 11'\ - ~ - Cl) ..; 0\ 0 N "" -"" -- -- -- > 0 0 -D ~ ~ ~ III .r:: ::E - '"~ 0 11'\ 00 00 - N 1:l "3 a- 11'\ 11'\ '" - v:> 11'\ - 0- v:> a- a- 00 "" '" - 0 11'\ N 00 v:> N 00 ~ 0 ~ .,; ~ a- co < ... .r:: ~ a- 11'\ 00 4- 0 N a- ~ .::t 0 ..... N ~ '" - '"- 00 00 v:> N 00 N 0 .,; .,; .,; I- Q) N ....: ....: ..... - - - -'" '" -"" - 00 U > Cl) ...J U 4- "" ~ 0 11'\ N \D 00 -'" N a- a- "- .::t N N c.. III 1:l 11'\ 0 0 00 '" 1:l 0 0 c: ... ~ N N ~ III U Cl :g 0 a- \D a.. '" ..; '" 11'\ N ~ '" .,.:; III '" a.:J 00 00'" '" - q ....: 00 .::t N N - - '"a- N ci ....: ...:J N \D Z u V ~ !9 v:> '" '" N '" '" 0- ~ U Cl ...c: -;' ..... 0 0 I I -.,; "" '" N .,; N 00 \D N I I '" .::t 0- c: a- 00'" .. ." 11'\ 11'\ N 0 a- .::t a- N D'! -..... - 0\ 0\ 0 -D 0 0\ Cl) Cl) ~ U "" 11'\ u E t N I 00 00 E>.. .::t \D Cl U >- "" I I I I N Cl III ~ ~ ..... 00 v:> III 0 ""'" a- 11'\ N N Z ~ c.. v:> 0 0 c.. 00 -; ....0 "" ~ .,; '" Z 00 U ..... 0 U ~ N - 0\ 0 N N "" ....: 00 -;' '0 t -Dt ..; -;' -;' '" N t >- .... «I Z U Cl) u c: cu <'i ~ ..... ..... 00 a.. a- N ------ ----- 0 N "- 00 00 "- 0- "- a- "- 00 00 00 ..... ..... "- ..... a- q. a.. a- 00 Cl) a- 0- a- 0- a- >- 19 APPENDIX I Basic Data Used (in Pmillion) y M X NFA 1976Q3 79.4400 46.0800 42.7320 44.4600 1976Q4 73.7210 47.4400 35.4080 64.7000 1977QI 73.5660 47.0500 36.5870 60.5000 I977Q2 84.0700 52.6500 41.8280 66.8000 1977Q3 90.6750 57.7500 40.1180 67.6300 1977Q4 87:1300 62.9600 33.0670 83.1400 1978Ql 83.5900 53.3100 35.7520 73.7000 1978Q2 92.8000 66.8300 52.1260 93.5100 1978Q3 109.410 75.4200 46.8700 101.340 I978Q4 144.450 83.7600 48.7520 124.630 1979Ql 132.600 83.6900 74.2340 132.000 1979Q2 147.730 105.520 105.668 180.690. 1979Q3 171.660 109.<120 90.338