α = 1 gives the poverty gap ratio, which is a linear measure of the extent to which household incomes fall below the poverty line.

Transcription

1 We used some special measures of poverty under the broad class of measures called the Foster-Greer- Thorbecke metric[chapter2, globalisation and the poor in asia]. Under this scheme, we use an indicator variable (I i) to denote the deprivation suffered by the i th household. Let x i = income of i th household. For the i th household, I i = 1 if x i < Z where Z is the household poverty line. and I i = 0 if x i Z Let the i th household represent the fraction w i of the population. The Foster-Greer-Thorbecke measures of poverty are defined as Pα = Σ( )α (Iiwi) α = 0 corresponds to the case where only the total fraction of households below the poverty line are counted, without considering how much deprivation an individual household suffers. It is an extremely crude measure of poverty, called the headcount ratio. α = 1 gives the poverty gap ratio, which is a linear measure of the extent to which household incomes fall below the poverty line. α = 2 gives the poverty severity index, which measures absolute deprivation suffered by the BPL households, giving a higher weightage to those households which are further below the poverty line. In our analysis, we have mostly used this index, to have a truly non-optimistic and unbiased estimate of India s poverty situation. Using data on income-distribution of Indian households across different years [McKinsey], we computed the headcount ratio, poverty gap ratio and poverty severity index of India from 1985 to The results were as follows. Year Headcount Ratio Poverty Gap Ratio Poverty Severity (%) (%) Index (%)

2 Next, we tried to assess how much the fruits of India s growth reach the poor [globalisation and poor in asia] Let η be the growth elasticity of poverty, i.e.. Clearly, η will have two components, namely, the one due to the effect of growth alone, and and the other due to the effect of changing inequality. These last-mentioned quantities are measured by two indices denoted by δ and ε respectively. δ= We used the index defined by Kakwani and Pernia, 2000, to measure the degree of benefits reaching the poor, namely, φ = We note that for positive overall growth rate g, δ is negative, since if all household incomes increased at g%, poverty would obviously reduce. Hence if φ > 1, we can say that η <0 and δ < η i.e. actual rate of decrease of poverty is greater than what it would be if the benefits of growth were equally distributed. Thus, the growth is strictly pro-poor. On the other hand, if 0<φ<1, η is still less than 0 but this time, η < δ i.e. actual rate of decrease of poverty is less than what it would be if the benefits of growth were equally distributed. Thus, the growth is not pro-poor; the poor don t receive as much benefits from growth as they should. This is the most common situation in any country. This kind of growth is called trickle-down growth. In some extreme cases, φ may actually become negative, which indicates that δ is positive, i.e. poverty increases In spite of positive growth. This is an example of anti-poor growth. All these interpretations are reversed if the economy is in recession. Then, a negative φ is actually good news for the poor, since this means poverty has reduced despite the recession. This time, the higher the positive value of φ is, the worse the poor have been hit. These considerations lead us to conclude that in the general case, if g and φ are of the same sign, the poor benefit more from a positive growth and are less badly hit by a recession. This conclusion is nicely captured by the commonly used index called the poverty-equivalent growth rate (PEGR), defined as g * = gφ In trying to compute India s poverty-equivalent growth rate, we found that in certain cases, it gave inflated figures, sometimes to the tune of 40-50%, leading to an over-optimistic evaluation of India s performance in poverty-alleviation. So we used g * = g φ. We recognise the benefits of this formulation by noting that

3 for φ>1, φ< φ and for φ<1, φ> φ. Thus, taking φ instead of φ reduces the deviations from the actual growth rate, making the PEGR more down-to-earth and reminding us that a lot of work has still to be done in eradicating poverty. Following are the results obtained by us from the study in the changes of poverty measures of India from We have separately determined the δ s and η s for poverty gap ratio and poverty severity index. We have computed both the classical PEGR and the one defined by us, to illustrate how the classical one tends to overestimate both the positive and the negative trends of poverty alleviation. Real growth rate of India is as per data from the IMF ( We have used the poverty line of $1.25 at PPP per day per person, which comes out as INR 44,250 per household per annum. Year g(%) Eta (Poverty Gap Ratio) Delta (Poverty Gap Ratio) phi=eta/delta (Poverty Gap Ratio) PEGR = gφ modified PEGR =g φ

4 Year g(%) Eta (severity) Delta (severity) phi=eta/delta (severity) PEGR = gφ modified PEGR =g φ From these tables, we observe that India as a whole has performed well consistently in reducing poverty, even during the recent slowdown. Plotted below is India s real growth rate and PEGR from 1985 to 2009, based on our index (g φ), with respect to both poverty gap ratio and severity. These graphs show that the moderately poor have benefited the most from India s growth, as is shown by the graph of the PEGR based on poverty gap ratio, whereas the acutely poor have been marginalised in some cases, as is shown by the graph of the PEGR based on severity.

5 Series1 -> Growth Rate Series2 -> Poverty Equivalent Growth Rate based on Poverty Gap Ratio Series3 -> Poverty Equivalent Growth Rate based on Poverty Severity index Series1 Series2 Series However, these figures do not reflect the performance of India s different regions. Data on four different representative states of India (Karnataka, Maharashtra, Punjab and West Bengal) were obtained from Rural Poverty in India in an era of Economic Reforms by Devendra Kumar Pant and Kakali Patra. Two such tables, containing data only for , are reproduced here for brevity.

6 For the different states, we divided the entries in the corresponding columns of the first and second tables to obtain the relative income-levels of different sections of households. We multiplied this by the average income per household for that state to obtain the absolute levels of household income for different sections of society. Based on this data, we computed the severity index for the rural population of these states over the years. From these, we determined the PEGR (both the classical and ours) for the rural belts of these states from 1986 to PEGR based on the classical formula: Year PEGR (Karnataka) PEGR(Maharashtra) PEGR(Punjab) PEGR(West Bengal)

7 PEGR based on our formula: Year PEGR (Karnataka) PEGR(Maharashtra) PEGR(Punjab) PEGR(West Bengal) i Exactly one entry in the top table is negative, indicating anti-poor growth in West Bengal s villages around In general, too, West Bengal s performance, as far as PEGR is concerned, is seen to be considerably poorer than Karnataka and Maharashtra. Punjab s PEGR s are also low, indicating a possible slowdown after the effects of the Green Revolution somewhat wore out. We have plotted PEGR of different states below.

8 Series1 Series2 Series3 Series In the above figure, Series1, Series2, Series3, Series4 represent Karnataka, Maharashtra, Punjab and West Bengal respectively. We identified three factors on which the severity of poverty of a region can depend[7 th chapter globalisation and the poor in asia]. These are income disparity between urban and rural areas (measured by the ratio of per capita household incomes of urban and rural areas), per capita expenditure by the government on poverty alleviation, and globalisation, which in turn has a nonlinear effect on severity. We have modelled this effect as a cubic. Thus we can write the severity of poverty of a region as y= α 1(income disparity) + α 2(poverty alleviation expenditure) + α 3(globalisation) + α 4(globalisation 2 ) + α 5(globalisation 3 ) +α 6 We have measured globalisation by glob= Data and Analysis for Karnataka: 10 Year Income Disparity Per Capita Poverty Alleviation Expenditure (INR) Glob Glob squared Glob cubed severity

9 Following results are obtained from a linear regression analysis of this model. From this table, we observe that all the factors are significant, since t values are greater than 2. However, somewhat surprisingly, income disparity has a negative effect on severity, implying that severity decreases as income disparity increases. This may be because of multi-collinearity. To determine multi-collinearity, we follow the steps: We perform a VIF test on the regression. If the VIF values are all less than 5, there is no multicollinearity. Otherwise, we remove the independent variable which has the highest VIF value and regress again. Then go to the first step. Finally, the result of the VIF test was as follows:

10 Thus, globcubed and income disparity emerge as the only two independent variables. However, income disparity is still showing a negative effect on severity. We suspected that this might be due to the variables showing opposite trends with time and thus ran a VAR and unit root test to investigate time effects. Following are the results of VAR and unit root test. VAR:

11 Dickey-Fuller unit root test: The Z values of the VAR results show that direct dependence of severity on income disparity is not significant and the apparent dependence between the two is mainly due to both of them depending on time. Thus, globalisation is by far the most important factor in determining severity of poverty in Karnataka. Further, Dickey-Fuller test reveals that globcubed has a unit root. As per regression, the dependence of severity on globalisation alone for Karnataka can be plotted as follows.

12 Thus we see that at low-to-middle levels, globalisation has a somewhat ambiguous effect on poverty, but as globalisation increases rapidly, poverty declines. Data and Analysis for Maharashtra: Year Income Disparity Per Capita Poverty Alleviation Expenditure (INR) Glob Glob squared Glob cubed severity

13 Following results are obtained from a linear regression analysis of this model. From this table, we observe that glob, globsquared and globcubed are significant, since t values are greater than 2. However, somewhat surprisingly, income disparity has a negative effect on severity, implying that severity decreases as income disparity increases. To determine multi-collinearity, we follow the steps as in the previous case. Finally, the result of the VIF test was as follows: Thus, globcubed and percapita poverty alleviation expenditure emerge as the only two independent variables. Predictably, per capita poverty alleviation expenditure is showing a negative effect on severity. So, the results for Maharashtra show no anomaly. As per regression, the dependence of severity on globalisation alone for Maharashtra can be plotted as follows.

14 Thus we see that at low-to-middle levels, globalisation has a somewhat ambiguous effect on poverty, but as globalisation increases rapidly, poverty declines. Data and Analysis for Punjab: Year Income disparity Per capita poverty alleviation expenditure (INR) Glob Glob squared Glob cubed severity Following results are obtained from a linear regression analysis of this model. From this table, we observe that all the factors are significant, since t values are greater than 2.

15 However, somewhat surprisingly, income disparity has a negative effect on severity, implying that severity decreases as income disparity increases. This may be because of multi-collinearity. To determine multi-collinearity, we follow the same steps as in earlier cases. Finally, the result of the VIF test was as follows: Thus, globcubed and income disparity emerge as the only two independent variables. However, income disparity is still showing a negative effect on severity. We suspected that this might be due to the variables showing opposite trends with time and thus ran a VAR and unit root test to investigate time effects. Following are the results of VAR and unit root test. VAR:

16

17 Dickey-Fuller unit root test: The Z values of the VAR results show that direct dependence of severity on income disparity is not significant and the apparent dependence between the two is mainly due to both of them depending on time. Thus, globalisation is by far the most important factor in determining severity of poverty in Karnataka. Further, Dickey-Fuller test reveals that globcubed has a unit root. As per regression, the dependence of severity on globalisation alone for Punjab can be plotted as follows. Thus we see that at low-to-middle levels, globalisation has a somewhat ambiguous effect on poverty, but as globalisation increases rapidly, poverty declines.

18 Data and Analysis for West Bengal: Year Income disparity Per capita poverty alleviation expenditure (INR) Glob Glob squared Glob cubed severity Following results are obtained from a linear regression analysis of this model. From this table, we observe that only income disparity is significant, since t value is greater than 2. However, somewhat surprisingly, income disparity has a negative effect on severity, implying that severity decreases as income disparity increases. To determine multi-collinearity, we follow the same steps as in earlier cases.

19 Finally, the result of the VIF test was as follows: Thus, globcubed and income disparity emerge as the only two independent variables and even among them, globalisation is not significant. However, income disparity is still showing a negative effect on severity. We suspected that this might be due to the variables showing opposite trends with time and thus ran a VAR and unit root test to investigate time effects. Following are the results of VAR and unit root test. VAR:

20

21 Dickey-Fuller unit root test: The Z values of the VAR results show that direct dependence of severity on income disparity is not significant and the apparent dependence between the two is mainly due to both of them depending on time. Thus, globalisation is by far the most important factor in determining severity of poverty in Karnataka. Further, Dickey-Fuller test reveals that globcubed has a unit root. As per regression, the dependence of severity on globalisation alone for West Bengal can be plotted as follows. Thus we see that for West Bengal, unlike the other states, at low-to-middle levels, globalisation leads to decrease in poverty, but as globalisation increases rapidly, poverty increases again. This may be due to other factors like politics. However, we must not look too deeply into this result, as it is merely predicted and nothing of this sort has been observed yet.