CHAPTERS SOURCES OF GROWTH IN LOW INCOME ECONOMIES: A THEORETICAL AND EMPIRICAL ANALYSIS
CHAPTER EIGHT SOURCES OF GROWTH IN LOW INCOME ECONOMIES : A THEORETICAL AND EMPIRICAL ANALYSIS In chapter five, we had analysed some of the implication of our theoretical model, which explains the positive relationship between average annual per capita growth rates and initial levels of per capita income. One of the implications of that model was that it is possible for countries to escape from vicious cycles to virtuous cycles and catch up or overtake relatively richer countries, by achieving higher levels of (what we have termed) "exogenous" sources of growth. The world "exogenous" needs explanation. In chapter three we mentioned that a large number of crosscountry studies in growth economics have tried to identify the sources of economic growth and one of the strongest contenders is the investment ratio. In our empirical analysis however, we find that though investment ratios do lead to growth, they are themselves explained by initial levels of per capita income. Thus, investment ratios become part of the per capita income - investment ratio - per capita growth cycle, which describes the process of cumulative causation in our model; it is endogenous to this cycle. On the other hand, "exogenous" sources of growth would be those sources which influence growth independently of the per capita income levels; higher values of these sources lead to higher values of per capita growth, for given levels of per capita income. It should be clarified here that "exogenous" in this sense is not to be confused with the exogeniety (in an econometric sense) of an explanatory variable in a I
ChapterS 183 regression exercise, where the explanatory. variable is assumed to be exogenous to the dependent vari~ble. Since we shall be concerned with both forms of exogeniety in our. study, we shall use the word (i.e., exogenous) within inverted commas in the context of per capita income, and without inverted commas when used in the econometric sense. Clearly, the identification of these "exogenous" sources is crucial to the understanding of the process of growth in these economies and for designing appropriate policies that may give a boost to future growth performances. This chapter has two objectives. The first objective is to empirically test for some economic variables as possible "exogenous" sources of growth in developing economies. We shall deal with this in the first section. The second objective is to theoretically expl~;tin these empirical results, and this is done in the next section. Some Empirical Results As we have mentioned in chapter three, recent cross section studies have identified more than sixty variables as significant sources of per capita growth. These include social, political, economic and even religious and geographical variables. Many of these sources of growth are not backed up by rigorous theories. As we have explained in the last chapter, we have chosen two sources based on the fact that they are (a) economic variables (b) their role in the process of growth is explained by existing theories and (c) they have evoked considerable interest in the recent literature. These sources are Human Capital and International Trade and in this chapter we shall start by testing them as sources of growth in developing economies.
Chapter 8 184 For each of these potential sources of growth, i.e., Human Capital and International Trade, we have used more than one measure. Thus each of these measures has been tested as a source of growth. In fact, corresponding to each measure, we have carried out two regres~io~s. In both regressions the dependent variable is the same, i.e., the rate of per capita growth. In the first regression however, one particular measure of Human Capital or International Trade is the only explanatory variable, while in the second regression, that particular measure as well as per capita income are the explanatory variables.. The first regression tests whether the particular measure is a source of growth while the second tests whether it is an "exogenous" source of growth, i.e. whether it leads to growth independently of the per capita income levels. We shall now present our empirical analysis. In the next section we deal with Human Capital while International Trade is dealt with in the section after that. Human Capital In our empirical analysis, Human Capital is the only variable the data on which have not been taken from the World Banks World Tables. These data have been taken from Barro and Sala-i-Martin (1995). There are two alternative measures of Human Capital that we have used, (a) average years of schooling attainment for males (SM) and (b) average years of schooling attainment for females (SF). The data are for the year 1985 and are assumed to hold approximately for the beginning of the period, i.e. 1982.,
Chapter 8 185 We have run OLS regressions to test for each of the two measures of Human Capital, as a source of growth. In each case, the per capita growth rate is the dependent variable and one of the two measures of Human Capital is the independent variable. Since we are particularly interested to know whether Human Capital explains per capita growth, "exogenously" to per capita output levels, we run a second set of regressions which now include per capita income as an additional explanatory variable apart from the measures of Human Capital. The results from all these regressions are presented in table 1 below. TABLE 1 PANEL A ( MALE SCHOOLING ) Regression Equation : G = a + b SM c -0.00762 0.006755-1.128262 0.2669 SM 0.00273 0.001709 1.601128 0.1183 Sample Size : 3 7 Adjusted R-squared 0.041626 White's statistic : 1.43 (0.488), Test rejects heteroscedasticity
Chapter 8 186 PANEL B ( FEMALE SCHOOLING ) Regression Equation : G = a + b SF c -0.00227 0.004772-0.476440 0.6367 SF 0.00181 0.001551 1.170639 0.2497 Sample Size : 3 7 ' Adjusted R-squared 0.010184 White's statistic : 4.09 (0.129), Test rejects heteroscedasticity PANEL C ( MALE SCHOOLING ) Regression Equation.: G = a + b 1 SM + b 2 Y c -0.008005 0.006876-1.164174 0.2525 SM 0.002186 0.002068 1.057169 0.2979 y 3.60 E-06 7.45 E-06 0.483600 0.6318 Sample Size : 37 Adjusted R-squared 0.020178 White's statistic: 7.51 (0.111), Test rejects heteroscedasticity
ChapterS 187 ' PANEL D ( FEMALE SCHOOLING ) Regression Equation : G = a + b 1 SF + b 2 Y c -0.003667 0.005253-0.698001 0.4899 SF 0.000929 0.002062 0.450581 0.6552 y 5.49 E-06 8.32 E-06 0.660433 0.5134 Sample Size : 37 Adjusted R-squared -0.006022 White's statistic : 8.02 (0.090), Test rejects heteroscedasticity Panel A and B gives results of regression with SM and SF as the explanatory variables respectively. Panel C and D gives results of regressions, which include per capita income as well as SM or SF as the explanatory variable. These results clearly indicate that in developing economies statistical evidence does not confirm the hypothesis that Human Capital is a source of growth, "exogenous" or otherwise. International Trade International trade has been considered to be an engine of growth right from the days of Adam Smith. However, till date, issues related to the effect of
Chapter 8 188 international trade on the process of growth continue to be controversial. One of the problems related to any empirical study of the effects of trade on growth, lies in the identification of an appropriate measure of international trade. Most of the theoretical models in this.coj,ltext, as well as some empirical studies, measure international trade in terms of the degree of "openness" of an economy. (a) The Degree of Openness as a Source of Growth The first measure of international trade that we shall test for, as a source of growth, is the degree of openness. Here we shall use the average tariff rate of a country as a measure of its degree of openness. Data on the average tariff rate are taken from the "Handbook of trade control measures of developing countries supplement: A statistical analysis of trade control measures of developing countries" (UNCTAD, 1988). We define a variable AT that measures the trade-weighted mean total charges due to tariffs and para-tariffs for the group of developing countries in our exercise. This data is a rolling cross section, with each country giving the most up-todate information possible. The actual dates for each country range from 1985 to early 1988. Since this period coincides with the period we have used for the data on growth (i.e., 1982 to 1993), this data serves our purpose fairly well. In fact this database has been used in a number of articles (Pritchett (1996), Erzan et. al., (1989)) that deal with tariff rates. Out of the forty seven countries that we have used for our earlier analysis, this data is available for thirty five countries. Figure 1 below gives the scatter diagram between per capita growth and average tariff rates.
ChapterS 189 FIGURE 1 PER CAPITA GROWTH VERSUS AVERAGE TARIFF RATE 0.05 0.04 0.03 0.02 ~ e 0.01 0 E c.., 0 u.,... " "'"' -0.01-0.02-0.03 40 60 80 100 120 140 1 0-0.04 Average Tariff Rate As in the case of Human Capital, we shall now run a set of two regressions where the average tariff rate is the only explanatory variable in the first regression while the average tariff rate and per capita income are explanatory variables in the second regression. The results are presented in table 2 below.
Chapter 8 190 TABLE 2 PANEL A Regression Equation : G=a+bAT c -0.002610 0.004790-0.544801 0.5896 AT 0.000227 0.000108 2.105407 0.0429 Sample Size 35 Adjusted R-squared 0.091704 White's statistic: 0.96 (0.615), Test rejects heteroscedasticity PANEL B Regression Equation : G = a + b 1 AT + b 2 Y c -0.012847 0.006456-1.989997 0.0552 AT 0.000273 0.000104 2.625498 0.0132 y 1.30E-05 5.84E-06 2.224299 0.0333 Sample Size : 35 Adjusted R-squared 0.188747 White's statistic : 3.49 (0.478), Test rejects heteroscedasticity
ChapterS 191 The results of the regressions presented above indicate that the average tariff rate is a positive and statistically significant source of growth, "exogenous" and otherwise. This result however, goes against all the theories of trade and growth, which maintain that there is a positive relationship between the degree of openness of an economy and its growth rate. Since average tariff rates are inversely related to an economy's degree of openness, these theories would predict an inverse relationship between the per capita growth and the average tariff rates; This is totally inconsistent with the empirical evidence presented above that indicates a positive relationship between the per capita growth and the average tariff rates. However, another look at the scatter diagram in figure 1 shows that there is an outlier with a very high average tariff rate that is giving rise to the positive result. We now drop this outlier and carry out the. ' regression exercises once more. The results are presented in table 2A below. TABLE 2A PANEL A Regression Equation : G=a+bAT c -0.001836 0.006129-0.299581 0.7664 AT 0.000203 0.000161 1.261146 0.2164 ' Sample Size : 34 Adjusted R-squared 0.017579 White's statistic : L44 (0.486), Test rejects heteroscedasticity
ChapterS 192 PANELB Regression Equation : G=a+b 1 AT+b 2 Y. '. c -0.011954 0.007402-1.615109 0.1164 AT 0.000244 0.000153 1.596074 0.1206 y 1.30E-05 5.93E-06 2.196003 0.0357 Sample Size : 34 Adjusted R-squared 0.122408 White's statistic : 3.68 (0.450), Test rejects heteroscedasticity The results indicate that once the outlier is dropped from the regression exercises, the average tariff rate is not a statistically significant source of growth, "exogenous" or otherwise. The lack of any significant result from the degree of openness as a source of growth is still inconsistent with the theories that maintain that there is a positive relationship between the degree of openness of an economy and its growth rate. Of course, it must be understood that there are various problems associated with the construction of a measure of "openness" of an economy. Firstly, there are so many ways in which the domestic industry.can be provided protection (tariffs, quotas, subsidies, non-tariff barriers etc.} that it becomes difficult to detem1ine the most
Chapter8 193 appropriate measure of openness. Secondly, there are some measures of protection (for example non-tariff barriers on the basis of environmental issues etc.) that are not easy to quantify and hence cannot be measured. As a result, even though the theoretical models described in the last chapter analyse the effect of openness on growth rates, it becomes extremely difficult to test them empirically. From a theoretical point of view, the effect of the degree of openness on growth would depend (at least partly) on the impact that the degree of openness will have on exports and the impact that exports have on growth. In the next stage of our analysis, we have alternatively considered export ratios as measures of international trade. We use average annual exports ratios, i.e. the average rat~o (for the twelve years, 1982-1993) of exports (U.S.$, current prices) to the gross domestic product (U.S.$, current prices) of an economy, as measures of international trade. (b) The Volume ofexports as a Source of Growth The next measure of international trade that we shall test for as a source of growth is the total exports ratio. This is the ratio between the volume of total exports of goods and services and the gross domestic product of a country. Again as in the case of Human Capital, we shall run a set of two regressions where the export ratio is the only explanatory variable in the first regression while the export ratio and per capita income are explanatory variables in the second regression. The results are presented in table 3 below.
Chapter8 194 TABLE 3 PANEL A Regression Equation : G=a+bX Variable Coefficient Standard Error t-statistic Probabif~ c 0.001326 0.006025 0.220042 0.8268 X 0.005958 0.020202 0.294913 0.7694 Sample Size : 47 Adjusted R-squared -0.019414 White's statistic : 5.99 (0.049), Test does not reject heteroscedasticity PANELB Regression Equation : G = a + b 1 X + b 2 Y c -0.001933 0.006217-0.310893 0.7574 X -0.017116 0.016583-1.032132 0.3077 y 1.48 E-05 4~72 E-06 3.136463 0.0030 Sample Size : 4 7 Adjusted R-squared : 0.073855 White's statistic: 10.17 (0.037), Test does not reject heteroscedasticity
Chapter 8 195 The results of the two regressions are presented in the two panels, where panel A represents the.fir~t regression while panel B represents the second regression. The results indicate that the volume of trade is not a statistically significant source of growth, "exogenous" or otherwise~ (c) The Structure of Exports as a Source of Growth The lack of statistical evidence in favour of the total export ratio as a souree of growth is somewhat surprising given that in a number of East Asian economies, exports were found to have acted as an engine of growth. It must. be considered however, that the East Asian countries were promoting the export of manufactured goods particularly. In Kaldor's model also (we have described it in chapter five), the growth impulses were coming from the export of manufactures. On the other hand, Non-Neoclassical development economists were also of the opinion that the expmt of non-manufactures generated negative impulses towards growth. These ideas seem to imply that the nature of exports, rather than it's volume, is important in the process of growth. We shall now empirically test for the validity of this hypothesis. In order to test whether the nature of International Trade is important for the process of growth, we have broken up the total volume of exports of goods and services into two parts - the export of manufactures and the export of nonmanufactures. Correspondingly, we have two ratios- the export of manufactures to GDP ratio, XM, and the export of non-manufactures to GDP ratio, XNM. Vve construct a third variable, XR, which is the ratio of export of manufactures to export
Chapter 8 196 of non-manufactures. Clearly XR is a measure of the structure of exports undertaken by a country. We now carry out OLS regressions, exactly on the lines of what v.:e have done for other measures, to test whether XR is a source of growth, and also whether it is "exogenous". The results are presented in Table 4 below. The results of panel A indicate that XR is a statistically significant source of growth. The results of panel B is consistent with the hypothesis that XR is an "exogenous" source of growth. TABLE 4 PANEL A Regression Equation : G =a+ b XR c -0.002884 0.002661-1.084030 0.2841 XR 0.025564 0.006850 3.731860 0.0005 Sample Size : 4 7 Adjusted R-squared 0.219370 White's statistic: 0.50 (0.777), Test rejects heteroscedasticity
ChapterS 197 PANEL B Regression Equation : G = a + b 1 XR + b 2 Y c -0.011852 0.003895-3.042991 0.0039 XR 0.027232 0.006349 4.289185 0.0001 y 1.36 E-05 4.57 E-06 2.966843 0.0048.. Sample Size : 47 Adjusted R-squared 0.334718 White's statistic : 4.88 (0.299), Test rejects heteroscedasticity In the empirical analysis described above, we have defined the structure of exports as a ratio between manufacturing exports and non-manufacturing exports. Non-manufacturing exports would include non-fuel primary exports, exports of fuels and exports of services. The arguments of the Non-Neoclassical Development economists, however, were based particularly on non-fuel primary exports, and not the exports of fuels or services. Taking this into account, we now define another variable XR1, which is the ratio between manufacturing exports and non-fuel primary exports. Clearly, this is an alternative definition (to XR) of the structure of exports which is more relevant for our analysis, since we are interested in the contrasting effects of manufacturing exports and non-fuel primary exports on growth. We shall now test this variable as a possible source of growth. The results are presented in Table 4A below.
Chapter 8 198 TABLE 4A PANEL A Regression Equation : G =a+ b XR1 c -0.002540 0.002578-0.985475 0.3297 XR1 0.008797 0.002300 3.824623 0.0004 Sample Size : 4 7 Adjusted R-squared : 0.228547 White's statistic: 0.65 (0.721), Test rejects heteroscedasticity PANEL B Regression Equation : G = a + b 1 XR1 + b 2 Y c -0.010005 0.003809-2.626909 0.0118 XR1 0.008787 0.002171 4.046818 0.0002 y 1.18E-'05 4.63E-06 2.547908 0.0144 Sample Size : 47 Adjusted R-squared 0.312455 White's statistic : 2.39 (0.663), Test rejects heteroscedasticity
ChapterS 199 The results of panel A indicate that XR1 is a statistically significant source of growth. The results ofpm~el B is consistent with the hypothesis that XR1 is an "exogenous" source of growth. (d) The Econometric Exogeniety of XRl The regression results from panel (b) of table 4A above indicate that XR1 is an "exogenous" source of growth in the sense that XR1 is significantly related to growth independently of per capita income. However, as we have discussed in chapter three in the context of investment ratios, a significant coefficient in a regression does not indicate the exogeniety of an explanatory variable in terms of the dependent variable, as there may be reverse causality. Thus, apart from being exogenous to per capita income, XR1 has to be exogenous to per capita growth in the regression above. To ensure this,.we use Hausman's test ofthe exogeniety ofxr1, the procedure for which is exactly the same as in chapter three, for the case of investment ratios. We shall use. lagged values ofxr1 as instrumental variables. In order to obtainthese lagged values, we construct a panel on the lines of chapter three. The average annual growth. rates and the XR1 values are calculated for four-year periods (i.e., for 1982 to 1985, 1986 to 1989 and 1990 to 1993). The initial per capita income values corresponding to the three periods are those for the years 1980, 1984 and 1988 respectively. Let us now define, the one period and two period lagged values ofxr1 as XR1 (-1) and XRl (-2) and use them as instrumental variables for XR1. The result of the Hausman test for the exogeniety ofxr1 is presented in table 5 below.
ChapterS 200 TABLE 5 Regression (R1) : G = a + b 1 XR1 + b 2 Y Variable Coefficient Standard Error t-statistic Probability c -0.010471 0.004357-2.403107 0.0176 XR1 0.004341 0.001807 2.402873 0.0176 y 1.27E-05 5.58E-06 2.276961 0.0243 Sample Size 141 Suspected Endogenous Variable : XR1 Instruments : XR1 (-1), XR1 (-2) White's statistic : 2.21 (0.696)Test rejects Heteroscedasticity Regression (R2) : XR1 = a + b 1 XR1 (-1) + b 2 XR1 (-2) + b 3 Y c 0.365894 0.206546 1.771491 0.0836 XR1 (-1) 4.782541 2.110589 2.265975 0.0285 - XRI (-2) -4.144729 2.560333-1.618824 0.1128 y -0.000534 0.000332-1.609442 0.1148 Sample Size 47 White's statistic : 37.99 (0.000), Test does not reject heteroscedasticity Residual from R~gression (R2) : ZXRI
Chapter 8 201 Regression (R3) : G = a + b 1 XR1 + b 2 Y + b 3 ZXR1 c -0.016074 0.008689-1.849849 0.0712 XR1 0.002131 0.002717 0.784381 0.4371 y 2.49E-05 1.08E-05 2304018 0.0261 ZXR1 7.90E-05 0.005662 0.013958 0.9889 - Sample Size 47 White's statistic : 3.27 (0.773), Test rejects heteroscedasticity There are three regressions as in chapter four. Regression (R1) is the same as the one represented in panel (b) of table 4A, except that the former is based on panel data (with three successive time periods for each country pooled together, where each period is the average of four years), while the latter is. based on cross section data (with twelve year averages for each country). Here, we shall be testing for the exogeniety of XR1 in regression (R1). In regression (R2) we regress XR1 on its instruments XR1(-1), XR1(-2) and per capita income Y and name the residual, ZXRJ.. Finally in regression (R3) we run the auxiliary regression, which is similar to regression (R1) except that it includes ZXR1 as an added explanatory variable. The insignificant t statistics for ZXR1 implies that XR1 is exogenous in regression (Rl).
Chapter 8 202 (e) Manufactured and Non-fuel Primary Exports as Exogenous Sources of Growth Our empirical analysis has shown that the structure of exports, XRI is a statistically significant source of growth, independent of the values of per capita income. Now, XRI is actually the ratio between the manufactured exports to GDP ratio and the non-fuel primary exports to GDP ratio, each of which has a separate influence on the per capita growth rate.... In fact, the conventional wisdom in development economics is that while manufactured exports have a positive effect on growth, non-fuel primary exports have a negative effect, i.e. higher non-fuel primary exports may lead to lower rates of per capita growth. Are these ideas borne out by the statistical evidence? Are these variable significant sources of growth (or retrogression as in the case of non-fuel primary exports) independently of the per capita income level? This is what we shall now try to ascertain using OLS regression analysis. Let us first consider manufacturing exports. For this we shall run a regression where the per capita growth rate is the dependent variable while the per capita income Y and manufactured exports ratio XM, are the explanatory variables. Next, we consider nonfuel primary exports. Here we shall run a regression where the per capita growth rate is the dependent variable while the per capita income Y and non-fuel primary exports ratio, XP, are the independent variables. The results are given in table 6 below.
Chapter8 203 TABLE 6 Regression Equation : G = a + b 1 XM + b 2 Y c -0.011354 0.004008-2.833068 0.0072. XM 0.154704 0.071211 2.172469 0.0358 y 1.19E-05 5.33E-06 2.233034 0.0312 Sample Size : 43 Adjusted R-squared : 0.234573 White's statistic,: 5.12 (0.274), Test rejects heteroscedasticity Regression Equation : G = a + b 1 XP + b 2 Y. c 0.000569 0.004382 0.129809 0.8974 XP -0.041425 0.019392-2.136199 0.0387 y 1.04E-05 5.08E-06 2.058178 0.0460 Sample Size ; 44 '. Adjusted R-squared : 0.125671 White's statistic : 7.41 (0.115), Test rejects heteroscedasticity
Chapter 8 204 The results indicate that (a) XM is a significant and positive source of growth, independent of per capita income levels and (b) XP is a significant and negative source of growth, i.e. higher the ratio of non manufacturing exports to GDP, the lower is the per capita growth rate. It is also independent of per capita income levels. (f) The Econometric Exogeniety of XM and XP The regression results in table 6 above m(!,y again be criticized on the ground that XM and XP are not exogenous to per capita growth rates. To test for their exogeneity, we have carried out the Hausman test. We use the two lagged values of XM and XP i.e., XM(-1), XM(-2) and XP(-1), XP(-2), as the respective instruments. As in the case ofxr1, we construct a panel where the values of per capita growth G, average manufactured exports ratio XM and average non manufactured exports rate XNM are calculated for three four-year periods (i.e., for 1982 to 1985, 1986 and 1989 to 1990 to 1993) while the initial per capita income values are for the year 1980, 1984 and 1988. The results of the Hausman Test for the exogeneity of XM and XP respectively are presented in table 7 and 8 below. Table 7 gives the results of the tests for the exogeneity of XM while table 8 gives those for XP. Consequently, there are three regressions each in table 7 and table 8. In both tables, regression (R1) is the same as the ones represented in table 6 except that it is based on panel data, while the ones in table 6 were based on cross section data.
Chapter 8 205 TABLE 7 Regression (R1) : G = a + b 1 XM + b 2 Y c -0.015821 0.004600-3.438981 0.0008 XM 0.190492 0.086959 2.190597 0.0305 y 1.35E-05 6.61E-06 2.042919 0.0433 Sample Size 120 Suspected Endogenous Variable :XM Instruments : XM (-1), XM (-2) White's statistic : 1.32 (0.856), Test rejects heteroscedasticity Regression (R2) : XM = a + b 1 XM (-1) + b 2 XM (-2) + b 3 Y c 0.003248 0.003748 0.866615 0.3919 XM (-1) 1.455793 0.234023 6.220723 0.0000 XM (-2) -0.439260 0.326611-1.344903 0.1871 y 1.71E-06 8.79E-06 0.194440 0.8469 Sample Size 40 White's statistic: 26.71 (0.000), Test does not reject heteroscedasticity Residual from Regression (R2) : ZXM
Chapter 8 206 Regression (R3) : G = a + b 1 XM + b 2 Y + b 3 ZXM c -0.019314 0.009143-2.112382 0.0417 XM -0.070196 0.163407-0.429578 0.6701 y 3.42E-05 1.39E-05 2.453026 0.0191 ZXM 0.772196. 0.426056 1.812426 0.0783 Sample Size 40 White's statistic : 6.58 (0.361), Test rejects heteroscedasticity TABLE 8 Regression (R1) : G = a +b 1 XP+b 2 Y c -0.000912 0.004457-0.204614 0.8382 XP -0.067360 0.022799-2.954524 0.0037 y 1.17E-05 5.19E-06 2.262453 0.0253 Sample Size 138 Suspected Endogenous Variable :XP Instruments : XP (-1), XP (-2) White's statistic : 3.41 (0.491), Test rejects heteroscedasticity
ChapterS 207 Regression (R2) : XP = a + b 1 XP (-1) + b 2 XP (-2) + b 3 Y c -0.001166 0.006890-0.169273 0.8664 XP (-1) 1.047688 0.102619 10.20953 0.0000 XP (-2) -0.125758 0.103160-1.219056 0.2298 y -1.72E-07 7.51E-06-0.022836 0.9819 Sample Size 45 White's statistic: 7.74 (0.257), Test rejects heteroscedasticity Residual from Regression (R2) : ZXP Regression (R3) : G = a + b, XP + b 2 Y + b 3 ZXP Variable Coefficient Standard Error t-statistic Probability c -0.010249 0.007934-1.291840 0.2036 XP -0.076724 0.051202-1.498451 0.1417 y 2.52E-05 8.96E-06 2.813033 0.0075 ZXP 0.251076 0.193170 1.299764 0.2009 Sample Size 45 White's statistic: 6.38 (0.381), Test rejects heteroscedasticity Here, we shall be testing for the exogeneity of XM and XP in regression (Rl ). In table 7 regression (R2), we regress XM on its instruments XM(-1), XM(-2) and per capita income Y and name the residual, ZXM. Similarly, in table 8 regression (R'2),
Chapter8 208 we regress XP on its instruments XP( -1 ), XP( -2) and per capita income Y and name the residual, ZXP. Finally in regression (R3) we run the auxiliary regression, which is similar to regression (R1) except that it includes ZXM as an added explanatory variable in table 7 and ZXP in table 8. The insignificant t statistics for ZXM and ZXP in table 7 and 8 respectively implies that XM and XP are both exogenous in regression (R1 ). A Theoretical Analysis In the last section, we have tested the hypotheses that human capital and international trade are sources of economic growth, and presented the results of these tests. We shall now try to give possible theoretical explanations for these results. In this context, our endeavour will be to explain these results in a manner that is consistent with the theoretical perspective on growth presented earlier (in chapter six). Let us start with the results on human capital. Our regression results show that m a group of developing countries, statistical evidence does not confirm the hypothesis that human capital Is a source of growth, "exogenous" or otherwise. However, one of the results indicates that male schooling is statistically significant if the level of significance is taken to be more than 12 % (say 15%). This may be interpreted as a very weak confirmation of the hypothesis that human capital is a source of growth but as we explain shortly, this is an incorrect interpretation. As we see in the next regression (table 1 panel C), on adding per capita income as an extra explanatory variable to this regression, this measure of human capital (i.e., male
ChapterS 209 schooling) becomes insignificant even for this level of significance. This happens because human capital is very closely related to the per capita income levels of these economies. In fact it shows a close statistical relationship with per capita growth rates (as evidenced in the regression with male schooling - although it is significant only when the level of significance is more than 12 %) simply because per capita income levels determine both the variables. What is the theoretical explanation for the proposition that per capita incomes determine the (stocks of) human capital? In order to understand this we look at the process of human capital formation in developing economies. The most common characteristic of human capital formation m developing economies is that it is constrained by the low levels of income of individual households, which forces the children to start earning their livelihood, rather than go in for schooling and education. Thus the per capita income levels (translated into household income levels) determine the levels of human capital forrhation. It should be clear from all these that human capital is endogenous to the per capita income levels and cannot be considered an exogenous source of growth. We now take up the results on international trade. The first result here is that the volume of trade is not a statistically significant source of growth, "exogenous" or otherwise. Even though this result is surprising, given that there is a voluminous literature on the effects of trade on growth, subsequent tests make it clear why we get this result. We find from these tests that it is the nature of the products exported that determine the effect on growth, rather than the total volume of exports. In particular, manufacturing exports have a positive effect on the growth rates whereas non-fuel primary exports have a negative effect. Clearly, when we consider the total volume of
. ChapterS 210 exports, these two opposite effects cancel each other and hence the total volume of exports has no effect on the growth rates. Why does the nature of the commodity exported determine the effect on growth rates? In order to understand this, let us go back to our model of growth in chapter five. There, due to the assumptions that (a) the economy is a closed one and (b) there are no demand constraints, the savings generated in the economy gets transformed into investments in terms of domestic capital goods. Ifwe drop the assumption of a closed economy, however, savings may be transformed into investments both in terms of domestic capital goods as well as imported capital goods; In fact, most developing economies have limited capacities of producing capital goods and hence, only a part of the investments are in terms of domestic capital goods while the rest are foreign capital goods that are imported in exchange of these countries exports. It is easy to understand that in such a situation, the price that these exports fetch in the world market would determine the volume of imported capital goods, and hence total investments in these countries. Let us now consider the exports of non-fuel primary products. It is usually the casein developing economies that primary exports are used in order to import foreign capital goods and the economies with higher primary exports are usually the ones that have a lower capacity to produce capital domestically. (Notice that this is impossible in a one-: commodity model where the part of the output that is saved, i.e., not conswned, can be automatically used for capital formation.) Let us assume that the higher the primary exports ratio, the lower is the ratio between investments in domestic capital goods and the total savings in the economy. For simplicity, we also assume that all primary exports are exchanged for foreign capital goods; Thirdly, we assume that the price of primary commodities is falling vis-a-vis the manufactured capital goods.
Chapter 8 211 This last assumption is in line with the Prebisch-Singer hypothesis. Under these three assumptions, for the same savings ratio, an economy with higher primary exports may. end up transforming its savings into a smaller volume of investments (domestic plus foreign), in which case its growth rates would be lower. In order to explain the above proposition clearly, let us take a simple case of two countries (A and B) over two periods (T 1 and T 2 ). Let the terms of trade between primary and manufactured (including imported capital goods) products be 1 in period T 1 and this falls to 0.5 in period T 2 Let the output and savings for both the countries be 100 and 20 units respectively. Country A has a lower primary exports ratio compared to B and hence a higher ratio of investments in domestic capital goods to. '. savings. Let the investment in domestic capital goods in Country A be 15 and that in B be 5. Clearly the average (over the two periods T 1 and T 2 ) investment for Country A, IA is given by, (15 + 5 X 1) + (15 + 5 X 0.5) 2 = 18.75 The corresponding value for average investmentfor Country B, I 8 is given by, (5 + 15 X 1) + (5 + 15 X 0.5) 2 = 16.25 Thus, with a fall in the terms of trade against primary goods, the country (Country B in our example) with a higher primary exports ratio (and a lower ratio between
Chapter 8 212 investment in domestic capital goods and savings) has a lower investment ratio. Since growth is capital constrained in these economies, this implies that a country with a higher primary exports ratio also has a lower growth rate. One aspect of the above example is that it assumes that in the first period (i.e. T 1 ), the investments are equal in both the countries and this is a strong assumption. However, even if Country B had a higher investment in period T 1, a sufficient fall in the terms of trade would again give us the same results as above. Thus, a sufficient condition for our result is that the investment ratio of the countries with higher primary exports ratio should not be so high as to completely counter the fall in investments brought about by the fall in terms of trade. Let us now consider the case of manufactured exports. Since manufactured exports do not face any terms of trade loss, a higher manufactured exports ratio can lead to higher levels of total investments - for the same investment ratio in terms of domestic prices - and hence growth. Of course, manufactured exports can lead to growth due to other reasons as well. The production and export of manufactured goods exhibit externalities of various types (for example, it led to inflow of foreign capital in South Korea) that lead to higher rates of growth. A brief summary of the chapter Let us now summarise very briefly the contents of this chapter. The first part of this chapter carries out an empirical exercise that tries to identify some of the
Chapter 8 213 sources of economic growth. In this context our empirical findings are (a) human capital is not a statistically significant source of growth (b) the degree of openness of an economy is not a statistically significant source of growth (c) the volume of exports is not a statistically significant source of growth (d) exports of manufactures are a source of growth (e) exports of primary commodities are a source of retrogression. In the second part of this chapter we put forward a theoretical analysis of these results...