Growth Acceleration in India

T

Economic and Political WeeklyApril 14, 20071382various indicators of economic growthshould therefore be analysed from 1992-93 onwards11instead of from 1990-91.The pre-1991 reform period of equallength12is considered (beginning from1980-81 up to 1991-92) in order to com-pare the average trend growth rate in thetwo periods. To put it in another way, theunderlying exercise compares whethertrend growth rates in the two periods aresignificantly different.To carry out this analysis, the trendgrowth rates for the two periods using asemi-logarithmic linear13trend equationare estimated separately.ln (GDP)t = α + β1t,for the pre-1991reform period...(3)andln (GDP)t = α' + β'1t,for the post-1991-reform period ...(3)'The estimated regression results (witht-ratios in parentheses) obtained are:Pre-1991 reform period: 1980-81 to1991-92ln (GDP)t = 12.84 + .053t, t = 1 ......12.(1087.95)* (32.96)*...(4)DW =1.5Post-1991 reform period: 1992-93 to2003-04ln (GDP)t = 12.758 + .059t,t = 13 ......24.(540.64)*(47.11)*...(5)DW = .84Comparing equations (4) and (5), we findthat the trend growth rate in the post-1991reform period is 5.9 per cent and is higherthan that of the pre-1991 reform period.Now, in order to investigate the issue ofa structural difference in the trend equa-tions for the two periods, we tested the nullhypothesis that the set of coefficients inthe pre-1991 reform period [equation (4)]is equal to the set of coefficients in the post-1991 reform period [equation (5)]. Assum-ing the two equations to be structurally thesame we test the null hypothesis (jointly):α = α'β1 = β'1To find the difference in the coefficientsof intercept as well as slope in the pre- andpost-1991 reform periods using a “dummyvariable approach”,14we introduced bothadditive15(D) and multiplicative16(Dmultiplied by t) dummies to equation (3)and estimated the following expandedregression equation by pooling all 24observations.ln(GDP)t=α + α1Di+β1t + β2(Dit) +ui,t = 1 ......24...(6)where Di = 0 for all the observations forthe pre-reform period (i e,1980-81 to1991-92) and 1 for the observations for thepost-reform period, i e, 1992-93 to 2003-04.From this it would follow:E (GDPt/Di = 0) = α + β1tfor the pre-1991-reform period...(7)andE (GDPt/Di =1) = (α + α1) + (β1+β2)t=γ1 +γ2tfor the post-1991 reform period ...(8)It is important to note here that equations(3) and (3)' are same as equations (7) and(8) respectively, with α' = γ1= (α + α1)andβ'1=γ2= (β1+β2). The coefficienta1 is the differential intercept (it tells byhow much the value of the intercept inpost-1991 reform period differs from thatof the pre-1991 reform period) and β2isthe differential slope coefficient (indicat-ing the difference in the slopes of regres-sion lines) for the period for which Di=1,i e, post-1991 reform period. The signs ofthese coefficients indicate the direction ofchange in the intercept and slope duringthe post-1991 reform period. The resultsobtained by estimating equation (6) usingGDP data are given asln (GDP)t= 12.84 – .084Di+ .0528t= (1212.1)*(-2.89)* (36.72)*+ .0062Ditt = 1 ......24. ...(9) (3.05)*DW = 1.24Since in the above regression the t-teston both the differential intercept anddifferentialslope coefficients are statisti-cally significant at the 1 per cent level, weinfer that the regressions for the two periodsare different, i e, the trend growth equationfor the post-1991 reform period is struc-turally different from that of the pre-1991reformperiod. The estimated annual trendgrowth rate of 5.9 per cent in equation (5)during the post-1991 reform period is signi-ficantly different from the annual growthof 5.3 per cent estimated for the pre-1991reform period in (4).The test of the stability of the entireregression, using a F-statistic,17in equa-tion (9) was also performed by testing thenull hypothesis that coefficients of boththe additive and multiplicative dummiesare simultaneously equal to zero, i e,α1 = β2= 0As the value of the F-statistic obtainedwas 8.55, which is statistically significantat the 1 per cent level, we rejected the nullhypothesis and conclude that the trendgrowth equation for the post-1991 reformperiod is statistically different from that ofthe pre-1991 reform period.18The estimatedregression equation for the post-1991reform period has shifted upward with ahigher slope. The exercise performed aboverefutes the assertion that there is no sig-nificant pick up in the average annualgrowth rate in the post-1991 reform period.Turning our attention to thetiming of theturnaround in the growth rate, both addi-tive and multiplicative dummies beginningfrom the mid-1980s up to the mid-1990sare considered. This is to test whether theturnaround in growth occurred in responseto the mild reforms of the mid-1980s orFigure 1: Trend in Gross Domestic Product(measured at 1993-94 prices)Source: Based on Economic Survey, GoI, 2004-05.GDP (Rs 1,000 crore)16001400120010008006004002000

Economic and Political WeeklyApril 14, 20071384during 1997-98 to 2002-03. However, thedifference in averages fails to be signifi-cant if the growth estimates of 2003-04 andfurther are also included (comparison 6).The deceleration phase of the late 1990sseems to be a transitory phenomenon. Fromthe provisional, quick and advance growthestimates for the economy during the lastthree years, i e, 2003-06, it can be inferredthat the recovery period is already under-way (Figure 2). The GDP is estimated togrow at 8.5 and 7 per cent during 2003-04and 2004-05 respectively. Even in the firsthalf of the current fiscal year (2005-06),despite the low performance of agricul-ture, the economy registered a stronggrowth of 8.1 per cent with robust perfor-mances by industry (9.4 per cent) andservices (12 per cent) and is projected togrow at the same rate during rest of the year(advance estimates of the Central Statistical1980-81 to 1987-88 (instead of 1981-82 to1987-88), the average growth is 5.1 per cent.Since this is higher than the average annualgrowth of 4.9 per cent attained during theFifth Five-Year Plan (using Panagariya’sreasoning), it may be concluded that theaverage growth of the 1980s is differentfrom that attained in the earlier period.Scrutinising the validity of Panagariya’ssecond conclusion, let us consider his ownargument used to support his first conten-tion. If the growth rates in GDP experiencedduring the three years, i e, 1988-91 can makethe average growth of the 1980s higher,then it is important to draw attention tothe fact that the conclusion regardingthefragility of growth during the 1980sis simply based on only one year’sgrowthrate, i e, 1988-89. If average growthof only this period is omitted from theperiod 1980s (defined by Panagariya) thenthe F-statistic fails to be significant, whetherthe growth rate of 1.3 per cent during the“crisis year”, i e, 1991-92 (Table 1) isincluded or not.In Table 4, we recalculated theF-statisticforall the periods considered by Panagariya(after excluding the growth rate for 1988-89) with respect to the growth in the 1990sand found that the F-statistic fails to besignificant even at the 10 per cent level.It is only when the average growth duringthe 1980s is considered from 1977-78 on-wards, the F-statistic is significant. If theperiod of the1980s is not accepted to bea high growth periodbecause of the pres-ence of “crucial growth period” of 1988-91, then it would be justified to questionthe acceptance of the “fragility” of growthduring the 1980s which is simply basedon one year’s growth estimates.These results can also be confirmed fromTable 5 where the comparisons of averagesand variances of growth rates duringselectedperiods are presented.When thepre-1991 reform period, i e, 1980-81 to1991-92 (or 1990-91) is compared to thepost-1991 reform period, i e, 1992-93 to2003-04, we find that the variability of thelatter period is significantly lower andaverage growth is not significantly higher(comparisons 1 and 2). However, when thevariability in growth during 1980-81 to1986-87 is compared to the period 1988-89 to 2003-0420 (comparison 3), we findthe variance in growth is higher for thelatter period, though the F-statistic is notsignificant. Further, even if the “crisis year”,i e, 1991-92 is excluded from the latterperiod (comparison 4), the above resultremains unaltered. Does this mean it is thehigh growth period of the 1980s that isinstilling the variability in growth andmaking it fragile? The validity of both theconclusions remains questionable.In order to assess the argument regardingthe deceleration in growth since the mid-1990s [Rakshit 2003], we used dummiesfor the period beginning 1997-98 up to1999-2000. It is observed (Table 2) thatbeyond 1997-98, the sign of the multipli-cative dummy is not significantly nega-tive. This validates the assertion of a declinein the trend rate of growth. The claim is alsoverified by comparing the average growthratesduring the two sub-periods(i)1992-93to 1996-97, (ii) 1997-98 to 2002-03 (com-parison 5 in Table 5). The average annualgrowth of 6.7 per cent during 1992-93 to1996-97 is higher and statistically signifi-cant at the 5 per cent level than theaverageannual growth of 5.25 per centTable 3: Average Annual Growth during Selected Periods(Per cent)1951-521960-611970-711951-521980-811980-811980-811980-811992-93tototototototototo1959-601969-701979-801979-801989-901987-881990-911991-922002-033.63.952.953.55.85.15.785.45.9Source: Calculations based on growth estimates given in Economic Survey, GOI, 2004-05.Table 4: Testing of Equality of Variances of Average Annual Growth Estimatesduring 1980s and 1990sPeriodVarianceVarianceaF-statisticF-statisticaF-critical Valuea:5 Per Cent1981-82 to 1991-926.13.414.12.14F (9,10) =3.021980-81 to 1990-914.62.393.11.50F (9,10) =3.021981-82 to 1990-914.82.203.31.38F (8,10) =3.071977-78 to 1990-9112.510.968.56.87F (12,10) =2.911992-93 to 2002-031.59bNotes:aCalculations for each period are made after excluding the growth rate of 1988-89.bPanagariya (2004) calculates this value as 1.5 (the difference is on account of provisional andquick growth estimates considered by Panagariya for 2001-02 and 2002-03 respectively).Source:Columns 2 and 4 are taken from Panagariya (2004).Columns 3 and 5 are calculated using growth estimates shown in Table 1.Table 5: Comparisons of Means and Variances of Growth Ratesbetween Selected PeriodsSNoPeriod 1Period 2F-statistict-statistic11980-81 to 1991-921992-93 to 2003-04Average5.46.12-.89Variance5.82.012.89**21980-81 to 1990-911992-93 to 2003-04Average5.86.12-.45Variance4.582.012.27*31980-81 to 1987-881988-89 to 2003-04Average5.16.09-1.17Variance2.74.33.6341980-81 to 1987-881988-89 to 2003-04@Average5.16.4-1.78*Variance2.72.9.9551992-93 to 1996-971997-98 to 2002-03Average6.75.252.23**Variance1.291.021.2661992-93 to 1996-971997-98 to 2003-04Average6.75.71.22Variance1.292.37.54Notes:@ Calculations are made ignoring the growth estimates of ‘crisis year’, i e, 1991-92.* Significant at 10 per cent ** Significant at 5 per centSource:Calculations based on growth estimates given in Table 1.

7.8

7.3 7.3

1.3 5.1 5.9 4.8 6.5 6.1

Average Annual Growth Rates of GDP (per cent)

2

3

4

5

6

78

9

0

EPW