Document(s) 6 of 16

In this chapter we work on the basis of the 59 NSS regions in the 38th, 50th and 55th rounds rather than the 16-odd major states of India. Our focus is the rural sector and the inter-regional variations in rural poverty. Work using time series from successive NSS surveys has firmly established the connection between poverty reduction and agricultural productivity growth (Ahluwalia 2002). Ravallion and Datt (2002) use state-level cross-section and time-series data pooled together to re-establish the connection. The contrary view of Beasley and Burgess (2004) is probably due to a dubious fixed-effect model which has been rightly criticized by Peter Timmer (2005). Few studies have used the cross section data available for the NSS regions. One exception is Palmer-Jones and Sen (2003). This chapter attempts to push the work based on NSS regions further.

Poverty map for NSS regions

Jain and Tendulkar (1988) had studied the regional variation in poverty incidence based on the unit-level data for 1973–1974 NSS regions available to them. The number of regions used was 56. They had divided the regions into four quartiles using the headcount ratios calculated for each region. This enabled them to draw the "poverty map" for India which is reproduced in Map 6.1 The basic data on the different regions – the upper terminal value of the headcount ratio for each quartile, along with the shares of the population involved – are given in Table 6.1.

The overwhelming impression from the map of 1972–1973 (Map 6.1) is that the regions with varying incidence of poverty form reasonably clear blocks of contiguous areas. The high poverty NSS regions (in the fourth quartile), numbering fourteen, form a continuous East–West belt stretching all the way from West Bengal to Rajasthan in the west. Similarly the regions with the lowest headcount ratio (in the first quartile) are concentrated in the North-West of the country. The other regions, constituting the second and third quartile ranges of the headcount ratio, are not so compactly placed but they are not distributed geographically in a random way either. Both groups are represented in fairly large blocks both in North and South India.

Map 6.1 NSS regions ranked by rural poverty 1972–1973 (source: Jain and Tendulkar (1988)).

We wanted to see how a similar poverty map looked from the NSS data of the 55th round in 1999–2000. The regions now numbering 58 as against 56 in 1972–1973 and were again broken down into four groups by the quartile values of the headcount ratio. The poverty map for 1999–2000 is presented in Map 6.2. The statistics comparable to these two years are presented in Table 6.1.¹

When we compare the two maps the first strong points which impress us are the very slight changes which have taken place in the spatial distribution of poverty incidence over the 30-year period. In particular the high-poverty region stretches from the East to the West across the heart of India, as it did in the early seventies, but it stops at the border of West Bengal. The low-poverty region is confined as before to the North-West. Assam (north-eastern India), which used to be a low-poverty region, now falls in a mid-poverty region. However, unlike in the early 1970s, in the late nineties one can discern low-poverty regions in patches spreading over northern and eastern India. The Table 6.1 does show the substantial decline in the headcount ratios that has taken place in the country over this period but the impact on the relative variations by regions is minor. The comparative stability of the inter-regional differences in poverty is surprising, because the period has seen some important changes in the rural economy – in particular the spread of the second wave of the green revolution to the rice-growing states.

Map 6.2 NSS Regions ranked by rural poverty 1999–2000 (source: Generated from unit level data of consumption schedule of 55th round).

Note
NSS regions are ranked in descending order in terms of rural poverty. The '0' shows regions not covered. Codes 1 to 4 reflect high to low level of poverty (HCR) in four quartiles.

Relationship between the growth of rural household incomes and poverty incidence

An econometric analysis of the cross-section data from the 1972–3 survey had revealed two points: (i) the incidence of rural poverty was explained mostly by the level of household income (as measured in the NSS by RAPCE-the rural per capita expenditure of households; (ii) the contribution of the inequality measure in household expenditure was significant but added only a small amount to the explanation of the variance (Mazumdar 1990). We wanted to see if this relationship held in the data set for 1999–2000, thirty years later.

The relationship between poverty incidence (as measured by the headcount ratio) and APCE in the rural areas is indeed close and non-linear. The non-linearity is to be expected. It shows that as rural income levels increase across regions its marginal impact on the headcount ratio diminishes as fewer people are below the poverty line.

We fitted a non-linear model to the two variables, and also added a measure of the inequality of the distribution of APCE in a second model. The model specification was as follows:

HCR = a*[exp(bRAPCE + cRAPCE²+ dGINI)]

The estimated results are given in Box 6.1. It is seen that adding inequality variable GINI to the equation improves the fit, but the R² increases only slightly by 8 percentage points. Evidently the degree of inequality matters but is of minor importance compared to RAPCE in explaining the inter-regional variation in poverty incidence. This result may come as a surprise since the NSS regions in our sample of observations vary a good deal in the structure of land distribution and off-farm activities, as well as other sociological factors (like caste composition, the proportion of agricultural laborers, etc.) which have an impact on the degree of inequality in the distribution of RAPCE. We conclude that the evidence shows that the variation in mean expenditure across regions is much more than that of the degree of inequality – and it is the former which is the more significant determinant of rural poverty. It should also be emphasized that as indicated the situation in rural India in this respect has not changed over the thirty years.

Box 6.1 Regression estimates of rural poverty ratio in 1999–2000 across NSS regions.

The close relationship between RAPCE and poverty incidence means that in studying inter-regional variations in poverty we can concentrate on the determinants of the former. This is what we do in succeeding sections of this chapter.

The Palmer-Jones–Sen model

A paper by Richard Palmer-Jones and Kunal Sen (2003) attempts to explain the spatial stability of poverty incidence in rural India in terms of initial ecological conditions in the 80-odd NSS regions. They used the 43rd and the 50th rounds of the NSS to calculate the average headcount ratio (HCR) for each of the NSS regions for 1987–1988 and 1993–1994. The authors then used a simple linear relationship to explain the inter-regional variation in HCR by the agricultural growth rates (measured by gross output per hectare – aggregated from available district level data into NSS regions). An initial level of HCR for the only year available 1973 is used as a control variable, and some socio-economic factors are added to 'allow for social factors and agrarian structure' (Equation 1). The strong negative effect of agricultural growth on poverty incidence remains even after allowing for the other variables (Table 3), and vindicates the importance of the relationship between poverty reduction and agricultural growth at a fairly disaggregated cross-section level.

The authors then work out in some detail the proximate determinates of agricultural growth. Their empirical results are based on two propositions: (i) a positive relationship between irrigation and agricultural growth – worked out in a time series production function form (equation 2 and table 4); and (ii) a positive relationship between initial agro-ecological conditions and irrigation – worked out in an empirical relationship between the level of irrigation in the district and the proportion of the district included in each one of 15 'agro-ecological' zones, the latter capturing the best conditions for irrigation (equation 3 and table 5).

The message seems to be that initial agro-climatic conditions have driven the process of agricultural growth and poverty reduction in India. These are the conditions which have defined the potential for irrigation, the 'fundamental variable' for growth in land productivity in South Asian conditions. They have been "conducive to agricultural growth given the emerging technologies and public investment, and which once set off, induces through political administrative pathways, further investments, growth and poverty reduction" (ibid., p. 5). The model is then in the genre of 'ecological fundamentalism'. 'The initial conditions are unchangeable and unmodifiable and hence truly exogenous to policy, while variables such as irrigation, literacy, and rural infrastructure would be regarded as outcomes of "policies", past and present, and, of course, private actions through markets' (ibid.).

In the empirical work the authors use the relatively homogeneous agro-ecological zones (AEZs) defined by National Bureau of Soil Sciences and Land Use Planning sponsored by the Indian Council of Agricultural Research (ICAR 1992) (see Palmer-Jones and Sen Map 1, p. 14). Each AEZ would comprise NSS regions (or States or districts) in different proportions – the proportions could be ascertained by overlying the map of the unit of analysis (state, NSS region or district) over the AEZ map.

The amount of detailed work in piecing together different sources of data is impressive. But at the end of it one is left with some points of enquiry:

1. The role of factors other than agricultural productivity growth, e.g., the role of the non-farm sector, is not addressed in the exercise.
   
   
2. Any change in the process, e.g., in the post-reform period, is not discussed.
   
   
3. The analysis speaks to the determinants of poverty for the country as a whole. It does not throw much light on possible differences in the trajectories of development in different regions. Dividing the whole of India just into two groups of the ecologically 'favored' and 'non-favored' regions is' illuminating for a limited purpose only. Even then the authors were unable to evaluate the marginal returns to public investment in the two regions in terms of the limited objective of poverty reduction. The only strong result was about region 13 which 'seems to have favorable agro-ecological conditions and moderate irrigation levels but has the highest poverty ratio in India'. Evidently this region was an outlier in the model and cried out for better attention to its potential for poverty reduction through agricultural productivity growth. But as we shall see even for this region more recent developments point to better conditions than what was diagnosed in the P–S analysis.

Broad regions

We decided to divide the country up into a limited number of 'broad regions', grouped from the available 60-odd NSS regions, on the basis of three principles: (i) the average incidence of poverty (as measured by the headcount ratio over the three rounds of the NSS – 83, 93 and 99; (ii) the agro-climatic zones into which the NSS regions fell; and (iii) geographical contiguity. After some experimentation 7 (seven) regions were distinguished. They are reported in Table 6.2.

(See Palmer-Jones and Sen, Map 2 and Table 2, for the definition of the agroclmatic zones.)

Map 6.3 should be read with the representation of poverty incidence in Map 6.2 above to get a fix on the demarcation of the broad regions in our analysis. It should be noted that in Map 6.3 we have divided the medium-poverty zones into two sub-groups – medium low and medium high. We have also distinguished geographically between sub-regions in the Northern and Southern parts of the country with similar incidence of poverty. Thus we end up with seven 'broad regions' in our subsequent analysis.

Defining the broad regions

Region 1 is the most clearly demarcated – not only did it have the lowest incidence of poverty in 1999 (less than 6 percent) but also the steepest decline over the period considered. It stretches from the Western Plain, Kutch and part of Kathiwaar peninsular into the Northern Plain and central highlands, and further into the fertile irrigated areas of Punjab and Haryana.

Region 2 is the 'heart' of the poverty belt, which had been identified as early as the early 1970s (Jain and Tendulkar) accounting for substantial part of the rural poor in 1999. It covers the area of the Eastern (Chattisgarh) plateau and Eastern Ghats and extending into the central highlands and part of the Deccan plateau. This is a hot semi-arid region with limited scope for irrigation.

Region 3 is the medium-poverty region extending over Eastern UP, Bihar and into the Central Highlands. It had more potential for irrigation than Region 2 though the soil is less favorable for staple agriculture.

Map 6.3 Broad regions of India (source: Generated through GIS software).

Region 4 is a more heterogeneous one stretching along the east coast of India. It includes the hot sub-humid to humid plains of Bengal and Assam and stretches north-east to include the area of the Eastern Himalayas, and further south into the semi-arid perhumid area of the Eastern coastal plain.

Region 5 is the Western Gnats and Coastal Plain with red laterite and alluvium derived soils and humid to perhumid ecological conditions.

Region 6 is the arid region of the Deccan, including parts of Telengana and the Eastern Ghats with red and black soil.

Region 7 is the Eastern Ghats and Tamil Nadu uplands the Karnataka Deccan plateau with red loamy soil.

Table 6.3 presents the cropping pattern in seven broad regions. In terms of cropping pattern broad region 5 clearly stands out.

Characteristics of broad regions

Incidence of poverty

Table 6.4 gives the headcount ratio (% poor) by broad regions for different NSS rounds. The method of calculating the HCR is as follows:

1. The HCR is calculated by using a state-specific poverty line built up by Deaton for all the three years. Therefore, the portion of each broad region falling in different state would have different poverty line.
   
   
2. rpov87 and rpov93u are based on APCE (average per capita consumption expenditure) of uniform reference period (URP), given in the NSS unit-level data of the year 1987–1988 and 1993–1994. The rpov99 was based on APCE of mixed reference period (MRP). To make 1993 poverty estimate comparable with that of 1999, we have converted APCE (URP) of 1993–1994 to APCE (MRP) by following the procedure of Sundaram and Tendulkar (2003a). Note that given the comparability problems posed by the change in reference periods, it is pertinent to compare poverty incidence between 1987 and 1993 on the basis of the URP, and the change between 1993 and 1999 on the basis of the MRP estimates.

Figure 6.1 graphs the HCR by broad regions for the different NSS rounds. It is seen that the reduction in poverty is more uniform across regions in the first period 1983–1993 than in the subsequent post-reform years. The second line from the top in the graph (showing poverty incidence in 1993) has shifted down in a roughly parallel way, except for region 2 (slightly less poverty reduction) and region 7 (slightly larger poverty reduction). The change in the incidence of poverty in the post-reform years 1993–1999 varies more as between the broad regions. The three regions 3, 4 and 6 have rather similar incidence of poverty in the 1999–2000 round but regions 3 and 6 had substantially higher poverty incidence than region 4 in earlier years. That is to say 3 and 6 had a steeper decline in poverty than region 4. Region 2 – the high poverty region of the North, managed little poverty reduction in the second period, while region 7, the high poverty region of the South, actually saw an increase in the headcount ratio. The two low-poverty regions, 1 in the North and 5 in the South, continued to reduce the incidence of poverty at much the same rate.

Figure 6.1 Trends of HCR across broad regions.

We now turn to the relative importance behind the inter-regional difference in the headcount ratio. The major elements are: i) the levels of land productivity relative to the pressure of population of land (the land-man ratio); (ii) the relative importance of off-farm rural employment; and (iii) the relative importance of urban development. We discuss each of these elements in the equation individually before bringing them together in the last two sub-sections.

Land productivity

Land productivity is obtained by dividing Value of output (at constant 1993–1994 all-India prices) by net sown area.2 Its variations across the broad regions and over the three years 1980, 1990 and 1999 are portrayed in Figure 6.2.

The two low poverty regions, 1 and 5, have consistently maintained and improved upon their land productivity. But high land productivity had been achieved by region 4 as well, particularly in the last period after the second green revolution, and by regions 3 and 6 to a smaller extent. Evidently, in the case of these other regions greater pressure of population on land has depressed household welfare.

Land–man ratio and land productivity

Figure 6.3 maps the position of the seven broad regions at two dates – 1983 and 1999 in the land–man ratio and land-productivity space.

Figure 6.2 Land productivity across region.

Ishikawa (1978) suggested that in Asian peasant agriculture, as the land–man ratio deteriorates due to the pressure of population on land, the agricultural economy adjusts by increasing land productivity – that is the movement in the space of Figure 6.3 would be in the direction of the south-east or the fourth quadrant. A second part of the Ishikawa hypothesis was that the points in the graph will lie along a rectangular hyperbola. The area under this hyperbola remains constant, signifying that the productivity per man remains roughly constant. In other words more intensive cultivation increases land productivity but only just compensates for the deterioration of the land–man ratio. Technical progress or the availability of co-operant inputs like capital can of course shift the curve upwards and to the right, thus increasing land productivity by more than the hypothetical level.

The following points can be made by looking at the scatter in Figure 6.3:

1. The movement of all regions over time has been in the Ishikawa direction – to the South East.
   
   
2. Region 1 – the low poverty region of the north-west have a position all of its own lying above and to the right of the other regions. It shows the importance of the higher level of technology – based presumably on irrigation – which enable it to attain a higher level of land productivity for all levels of land–man ratio relative to the other regions. Note that both in 1983 and 1999 it had a lower land–man ratio than region 2 and only slightly higher than region 6, but substantially higher land productivity.
   
   
3. Regions 2 and 6 are distinguished by having a steep slope of the curve connecting the two variables. It signifies a sharper decline in land–man ratio, relative to the increase in land productivity, than the other regions, The increase in the pressure of population on land for these regions has been of critical importance in depressing household welfare.

Figure 6.3 Land–man ratio and land productivity in agriculture: 1983 data points connected to 1999 points by arrows, across broad NSS region.

Notes
Land Productivity is Value of Output per hectare of Land (in Rs.) at 1993–1994 constant prices.
Land–Man Ratio – Hectare of land per person.
1 to 7 show changes in broad regions 1 to 7 and A is the figure for all-India (combined 1 to 7). Y-axis is land–man ratio.

4. All the other regions lie close together along a downward sloping non-linear curve. But it should be apparent that at most points of this hypothetical curve the elasticity would seem to be more than unity That is to say these regions over time have been able to increase its land-productivity at a higher rate than what would just compensate for the deterioration of the land–man ratio. This is the basis for the increase of productivity per man in these regions – though obviously at different rates (This point is pursued further below.)
   
   
5. Turning now to inter-regional comparisons it is interesting to note that outside of region 1, as we move down the scale of the land–man, each region, at both dates, lies to the south-east of the one before it. The only exception would seem to be region 7 – whose position at both dates is left or south-west of the region. Evidently this southern high-poverty region is burdened by particularly low potential for raising land productivity to compensate for its relatively low land–man ratio.

Rural non-farm sector

EMPLOYMENT

The welfare levels of rural households depend on the development of the non-farm sector along with the level of land productivity. Regions with low land productivity or unfavorable land–man ratio might be able to pull up their income levels with active participation in either the rural off-farm sector or employment in the urban areas. The role of the urban sector is portrayed in the next subsection. Figure 6.4 presents the percentage employment in rural non-farm activities in the seven regions of our study.

Figure 6.4 Share of non-farm employment across region.

Employment in the rural non-farm sector (NFS) can respond to two different types of developments. High growth in the farm sector creates demand for non-farm products (including services) and 'pulls' labor into it. On the other hand, limited opportunity for increase in land productivity together with pressure of population on land could 'push' labor into the off-farm sector.

The 'pull' effect seems to have been important in the prosperous low poverty region 1 – particularly in the development over time. The percentage of employment in NFS was relatively low in 1983 (NSS 38th round) but grew 30 percent over the period until 1999 as the farm economy prospered. Although NFS has increased somewhat over time in other regions, the rate of growth has been quite limited in all the regions – with the possible exception of region 7.

Looking across the seven regions it is clear that it is the pressure of population on land (as represented by the land–man ratio) that seems to be critical in determining the relative size of NFS. It is striking that the lowest levels of NFS outside region 1 are to be found in the regions with a relatively high land–man ratio: regions 2, 3 and 6 (see Figure 6.4). Since the regions differ a lot in terms of their incidence of poverty and hence levels of income, the conclusion suggested by this evidence is that it is the pressure of population on land, rather than the level of income, that is the dominant influence on the size of the NFS.

Both regions 5 and 7 are low land–man regions. The NFS sector in region 5 has been at the highest level in India for the entire period of our study, while the sector in region 7 has had a growth rate almost as high as that of the low-poverty region 1. But the two regions differ in terms of poverty incidence. Region 5 can clearly point to the successful development of its NFS sector as an instrument in its achievement of a low incidence of poverty in spite of the unfavorable land–man ratio. But region 7 continues to be a high-poverty region despite its relatively high growth rate of NFS.

LABOR PRODUCTIVITY

The proportion of rural income generated in the non-farm sector does not depend only on the proportion of employment in this sector. The other variable is the relative level of labor productivity. It is not possible to determine a priori how the latter will vary with the prosperity of the region. On the one hand, we would expect that in a relatively poor region there will be good deal of labor 'pushed' into off-farm activity for lack of opportunities in cultivation and related activities – and this will tend to depress the relative productivity in non-farm sectors. On the other hand, we would expect the agricultural sector to be less integrated with the non-farm economy in poorer regions. The enhanced 'dualism' in such regions would tend to make the productivity in non-farm to be relatively higher. We do not know which of these two influences would prevail in an inter-regional comparison. The empirical data presented in Table 6.4 suggests that in fact the latter is the more dominant influence. High-income regions (like 1, 3 and 5) have a lower productivity gap, while the highest productivity gap is found in the poorest regions 2 and 7.

Rate of urban-employment growth

How far does creation of employment outside the rural sector provide an additional element to the pattern of inter-regional differences? The data for the different rounds on this point are portrayed in Figure 6.5. There is a clear correlation between the incidence of poverty and the rate of urbanization across regions. As with NFS, region 1 again stands apart from the others in not only having a higher than average proportion of employment in the urban areas all along, but also in experiencing a faster growth of this sector than the other regions. The low poverty region 5 shows the highest urban rate, which increases by a third between the 38th and the 55th rounds. Clearly urban employment played as much of a role in poverty reduction as the NFS in this region. The lowest urban rates are found in the two high-poverty regions of the Central-West and the South (regions 2 and 7). The medium poverty regions 3, 4 and 6 have intermediate levels of urbanization and show only small gains over the period.

Figure 6.5 Share of urban UPS workers in All UPS workers.

Components of RAPCE across broad regions

There is a close relationship between the rural household welfare levels as measured by the average rural household expenditure per capita (RAPCE) and the incidence of poverty as measured by the headcount ratio (see section 1). We therefore tried to look at the different components of RAPCE which contribute to its differences across regions.

We make use of the identity:

Where Yr = total rural household income (expenditure)

P = total (rural + urban) population in region

Pr = rural population

Ya = total income (expenditure) of agriculture households

N = net sown area

Then Yr/Pr is RAPCE

Ya/N is land productivity

N/P is the land–man ratio

Yr/Ya is the ratio of total rural income to agricultural income (an index of the relative importance of the rural non-farm sector)

P/Pr is the inverse of the proportion of the population in the rural sector

Note that the Yr/Pr as given in the identity will not correspond exactly to the actual RAPCE obtained from the unit level data of the NSS. There is the issue of missing crops, and there is also the problem of the difference between household income and expenditure due to household savings among other things. Furthermore, a critical element missing from equation (1) is that of net value added per unit of gross output since detailed data on this point for the NSS regions is not available. Nevertheless, we can treat the Yr/Pr in equation (1) as a reasonably close index of the actual RAPCE.

Taking logs of all the terms equation (1) the percentage difference of all the variables in any region with respect to the base region – say region 1 – can be calculated. Thus the percentage difference in Yr/Pr between region 1 and every other region can be expressed as a sum of the percentage differences of the variables included in the RHS of equation (1). We can then form some notion about the relative quantitative importance of the latter in accounting for the difference in the hypothetical Yr/Pr.

Table 6.5 sets out the calculations for the 55th round of the NSS. We also include in the second column the actual value of RAPCE for this round (at 1993–1994 prices). It is seen that the signs of the differences of the actual values agree fully with those of the hypothetical values entered in the last column as the sum of the components in columns 3 through 6. It is, however, seen that the differences in the hypothetical values are exaggerated in all the regions except 4 and 7.

The following interesting conclusions emerge from the values of the components in relation to the sum:

1. Difference in land productivity is of overwhelming importance in the lower value of Yr/Pr in regions 2 and 6. It is also a significant factor in the lower level of rural income in regions 3 and 7. Only in two regions 4 and 5 land productivity seems to be higher than that of region 1, but in both these regions the adverse land–man ratio overwhelms the land-productivity advantage.
   
   

2. The more unfavorable land–man ratio plays a bigger role in the regions 3–5 and 7.
   
   

3. The important role played by the rural off-farm sector and urbanization (columns 5 and 6) in the southern regions 5 and 7 are striking. Region 5 is able to lift itself to a low-poverty region, in spite of a very unfavorable land–man ratio, through major developments in these activities, while region 7 mitigates its unfavorable land productivity and land–man ratio to some extent.
   
   

4. Off-farm activity and urban development are much less important in the more northern regions where the pressure of population on land is not as great (regions 1, 2 and 6). Regions 2 and 3 are in fact poorer because of the lower level of performance of these sectors compared with region 1, but the dominant factor behind the difference is lower land productivity – as indeed it is so in region 6 as well.

Dynamics of the broad regions

Using equation (1) the growth rate of RAPCE can be decomposed into the algebraic sum of the growth rates of the variables on the RHS. Note it is expected that N/P will all be negative. Yr/Ya is an index of the growth of the non-farm sector in the rural areas, and as we have seen will be positive. P/Pr shows the effect of urbanization and also will be positive. The decomposition exercise helps us to quantify the relative importance of the different variables in the identity in the growth rate of RAPCE in the seven regions. The results are given in Table 6.6.

The data presented in Table 6.6 help us to throw some light on the question: does the difference in land productivity – which was seen to be of such importance in the lower level of rural welfare in most of the regions relative to region 1 – a result of differential growth over the 1983–1999 period? Considering the period as a whole the growth rate of land productivity (Lnpro) was indeed higher in region 1 – with the exception of regions 3 and 4. But looking at the two shorter periods 1983–1993 and 1993–1999 separately, the striking fact emerges that the differential growth rate is largely due to developments in the 1993–1999 period. Over the 1983–1993 period, the growth rate of land productivity was significantly lower than that of three of the other region and exceeded the growth rate only in regions 4 to 6. This changed in the post-reform period 1993–1999. The growth rate of land productivity in region 1 shot up, while it became low or negative in three of the other regions. Even the four regions which had positive growth rate – the growth rates fell far short of the one attained by land productivity in region 1 with sole exception of region 4. The point underlines the problem of uneven regional development in agriculture in the immediate post-reform years which have been emphasized by many commentators.

The second point pertains to the role of the rural non-farm and the urban sectors. We had noticed the difference in 1999–2000 between region 1 and the other northern regions on the one hand, and the southern regions as a group, on the other. It is now seen that theses differences had indeed gathered momentum over the 1983–1999 period. It was the result of the differential patterns of growth over the entire period. The low-poverty region of the North (region 1) has maintained its difference in RAPCE (and poverty incidence) or pulled away from the others partly because of its high growth rate of land productivity, but also partly (with respect to the northern regions 2 and 3 in particular) because of higher growth rate of the rural non-farm and the urban sectors.

The importance of the rural non-farm and urban sectors were seen to be more important in the Southern regions in the 55th round. It is now seen that this is due to the relatively high growth rates of these sectors over the preceding twenty years. They grew at a relatively high rate not because of, but to compensate for, the low growth of land productivity.

Conclusions

The spatial stability in the inter-regional variation in rural poverty is impressive, The Palmer-Jones–Sen model is a very useful contribution in suggesting that the stability can be traced to the initial agro-ecological conditions of different regions of India which determined the effectiveness of infrastructure investments, particularly irrigation, and the subsequent growth of land-augmenting technical progress in agriculture. In this chapter we have tried to see if this broad interpretation is too restrictive, and if the aggregate picture might not hide important variations in poverty incidence and of factors other than land productivity in explaining the inter-regional variations.

A first attempt has been made to divide India into 'broad regions', grouping the NSS regions into seven clusters determined partly by agro-ecological conditions. Since it was shown that the relationship between RAPCE and the head-count ratio is very close in rural areas, we tried to concentrate on the determinants of the variations in RAPCE across our broad regions. It was seen that while the variations in land productivity is indeed of major importance, we need to take account of other factors to have a fuller explanation. Most important are: (i) the variations in land–man ratio; (ii) the relative importance of rural off-farm employment; and (iii) the degree of urban development. The decomposition model given in this chapter seeks to highlight the comparative importance of these factors in accounting for the variations in RAPCE across the broad regions. To mention one result in particular: the more important role played by the latter two factors in the southern regions of 5 and 7 is striking. The dynamic extension of the decomposition model helped us to unravel some of the interesting differences in trends across the regions. It was seen that the low poverty region of the North-West (region 1) was in fact losing in advantage over the other regions in terms of the growth of land productivity in the period 1983–1993, but that this equalizing trend has been reversed in the post-reform years of 1993–1999. Over the entire 1983–1999 period the maintenance of the leading role of region 1 in poverty reduction has not been entirely due to differential growth in land productivity (relative to the offsetting trend in land-man ratio). The growth of off-farm employment, both in the rural and the urban areas of this region, has contributed at least half of the differential growth in RAPCE relative to regions 2 and 3. The rural non-farm and urban sectors played a larger role in determining the level of RAPCE in the southern regions in the 55th round. These sectors grew at a relatively high rate over our period not because of, but to compensate for, the low growth of land productivity.

Document(s) 6 of 16