Document(s) 7 de 16

Background

The growth of agricultural output decelerated in the post-reform period. According to the National Account Statistics the trend rate of growth of GDP at factor cost in agriculture, which was 3.3 percent in the period 1981/1982 to 1990/1991 and 3.1 percent over the years 1991/1992 to 1995/1996, decelerated to 2.4 percent during 1996/1997 to 2000/2001. Much higher growth rates were registered by manufacturing and particularly the services sector. Since agriculture is still the largest labor-intensive sector in the Indian economy, the slow-down in output growth in this sector has raised concerns if the possibility of economically viable labor absorption is reaching its limits in agriculture. It should be noted that a slow-down in output growth can also be expected to reduce the employment elasticity in this sector. As the returns to labor fall it will move out into alternative occupations, including schooling and off-farm employment, even if it does not significantly increase open unemployment. We shall now briefly review the various difficulties – both at the external and internal margins – which might have led to a marked slow-down in output and employment. We will see that problems in policy making might have had their share in the causes of the slowdown.

The external margin

The net sown area in agriculture marginally declined in India in the 1990s–from 143 million hectares in 1990–1991 to 141.23 million hectares in 1999–2000 (Statistical Abstract: India 2002). It can be seen from the data on 'patterns of land utilisation' that in the same period 'land not available for cultivation' increased from 40.48 million hectares to 42.41 million hectares. It indicates an increase of two million hectares of land for non-agricultural uses – not an insignificant shrinkage of the external margin for cultivation, at the annual rate of about 1.5 percent per decade. However, gross cropped area has increased from 185.78 million hectares to 189.74 million hectares in the same period. It appears that there exists little scope for labor absorption through extensive cultivation. We will now discuss why the possibility through intensive cultivation (raising cropping intensity) also appears to be limited.

Increasing labor absorption through irrigation

It is well known that in Indian agriculture, as in many other Asian economies, controlled water supply is the critical input which not only enhances land productivity but also increases opportunity for increasing the input of labor. In fact we have discussed in Chapter 6 that the Palmer-Jones–Sen model has made irrigation the centre of their interpretation of the pattern of rural growth in India. In recent years, however, there has been much discussion in the literature of the increasing difficulties and costs of providing controlled water supply to agriculture.

At the end of the nineties the total gross irrigated area (GIA) reached 39 percent of gross cropped area. In the last two decades of the century ground-water irrigation (through wells) increased much more rapidly. By the triennium ending 1998/1999 ground water accounted for 56.7 percent, canals for 31.2 percent and tanks for 5.5 percent of net irrigated area (World Bank 2005, p. 30).

It has been maintained that ground-water extraction through private pumps has reached its limit in most parts of India except eastern India. The subsidized power for agricultural use is an important factor that led to the decline of ground water resources. The remaining potential of ground-water resources largely lies in eastern India where it is hampered by the inadequate spread of electric power (Hanumantha Rao, 2004). However, as the World Bank reports: 'Capital Expenditure on major and medium surface irrigation schemes and flood control continue to account for the largest share of public expenditures in the agricultural sector… But future expansion of surface irrigation infrastructure will come at increasing cost'. (2005, p. 31).

Under-pricing of canal water is extensively practiced by state governments, who are responsible for the administration in this sector. The consequent financial crunch leads to a vicious circle of deterioration of the irrigation infrastructure, diminished water supply to farmers and their reduced capacity to meet even the subsidized costs. Further, the system is regressive. 'Small and marginal farmers who comprise about 82 percent of the farmers who use canal irrigation, cultivate about half of the area that is irrigated by canals' (ibid., p. 33, italics ours).

The impact of fertilizer price policy

Fertilizer subsidy has been one of the crucial elements in the package of policies introduced in the seventies to support the green revolution in agriculture. Domestic producers of urea are given a designated retention price, calculated on a cost-plus basis. The difference between this price and the administered farm-gate price, minus the distribution margin, is paid as subsidy to the producers. The amount of the subsidy increased continuously through the eighties and the nineties reaching a peak of 0.7 percent of GDP at the end of the century (ibid., Figure 4.1, p. 28).

At the state level the main beneficiaries of this large volume of subsidies have been the richer states, in which irrigation is also more extensive. Gulati and Naryanan (2002) estimated that between 1981/1982 and 1999/2000 the subsidy shares of farmers was 66.5 percent of the total, the remaining 33.4 percent accruing to the fertilizer industry.

Another important by product has been that, together with the other subsidies in the overall agricultural package of policies, this expenditure has been a major factor reducing the availability of finance for agriculture extension and research and development. The longer-run impact of this policy on the growth of the sector has been substantial though difficult to quantify.

Diversification of output

Diversification of the product mix is an important way of increasing markets, including exports and also of increasing the labor use in the sector. Figure 7.1 reproduces the chart for labor use in selected agricultural products from the World Bank Report.

The substantially higher use of labor per unit of land in non-cereal products is striking. Sectors outside vegetables, like livestock and fisheries, have also been important in providing both extra employment and high value added to agriculture in many developing countries.

Recent growth in Indian agriculture has indeed seen evidence of significant development of the non-cereal sector. The share of food grains in total value of output in the crop sector declined from 48 percent at the beginning of the 1980s to 40 percent at the end of the 1990s (ibid., Table 2.5, p. 7). There has also been significant growth in meat and fish output, including exports. But some aspects of agricultural sector policies have been a drag on the process of diversification.

India has liberalized the regime of controls in agricultural pricing and trade, both in the wake of the reform process of the nineties and in response to the responsibilities under the WTO agreements. But the late nineties saw an increase in the nominal protection coefficients (NPCs–the ratios of domestic to world prices). Of most significance in the present context is the increase in NPCS for rice and wheat. This increase was driven by the maintenance of minimum support prices in the domestic market in the face of a rapidly declining world prices.

Figure 7.1 Average labor use for selected crops (days/ha/season) (source: World Bank (2005)).

The government's food-grain policy was meant to achieve two objectives: provide adequate income to farmers, and to ensure an adequate supply of food grain at reasonable prices. The major elements of the policies are: procurement of grains at the minimum support price from the farmers; distribution through the public system at subsidized prices; and a variety of restrictions on private trade in these commodities. With the downward trend in the world prices of rice and wheat since the mid-nineties, and the limited opportunity for exports, the volume and cost of buffer stocks in the government distribution system have increased. The effective subsidies associated with this system have benefited disproportionately the states growing the bulk of these commodities – which happen to be the richer states – and the richer farmers within them. Along with the regressive nature of the subsidies, this price policy has been a major element in slowing down diversification to non-cereals in the agricultural sector.

Investment in agriculture

A persistent criticism of the agricultural policies in India has been that the financial burden of the elaborate system of subsidies, quite apart from the impact on efficiency and equity, has produced a financial crunch which has inevitably reduced the funds potentially available to support public investment and research on R&D. Even India's elaborate extension services, which had played such a crucial role in the green revolution, is said to be starved of funds. The lack of productive research has meant that there is no major breakthrough in agricultural technology on the scale of the green revolution insight. The prospect for a high rate of growth of output and consequent growth in labor absorption on this sector does not look all that good at the beginning of the new century.

Land productivity continues to be at a low level in India relative to comparator countries (Table 7.1). The low level of land productivity is a major reason for the low incomes of households' dependent on the sector – both in absolute and relative terms. Increase in land productivity creates the virtuous circle of higher agricultural income, higher off-farm employment, and further income growth per worker in agriculture as 'surplus' labor pulled away from the sector (see 'Diversity of activities in agriculture' section below).

Employment elasticity in agriculture

We now turn to a discussion of employment elasticity in agriculture. Tables 7.2 and 7.3 combine the NSS data on employment in this sector with our own estimates of the index of agricultural; output used in Chapter 6 to provide estimates of employment elasticity over the two NSS periods – the 1980s covering the period between the 38th and the 50th rounds (1983/1984 to 1993/1994) and the 50th and the 55th rounds (1993/1994 to 1999/2000).

It should be noted that the estimates of employment elasticity presented in the tables above are different from the ones given in Chapter 3. The earlier estimates are based on NSS employment data combined with National Account estimates of agricultural growth. The latter of course are not available for the broad regions given in Tables 7.2 and 7.3.

The overall figures for employment elasticity for all-India confirm its decline in this sector which has been mentioned in Chapter 3. We have discussed earlier the problems on the supply side of labor. The withdrawal of labor for education in the younger age-groups for example pulls down the numbers reported to be employed in the NSS counts. To minimize this problem we have defined employment as comprising only those reported in the prime 15–59 age group. Agriculture of course has a high incidence of self-employment. Thus the problems mentioned in Chapter 3 about the count of employment would be particularly strong in this sector – and would affect the UPSS estimates more than the UPS ones. It is seen that the decline in employment elasticity in the 1990s is only marginal based on the UPS estimates and is substantial for the UPSS count. This is as it should be. The upsurge in the demand for secondary female labor during the second green revolution in Eastern India, and its decline in the subsequent period, has already been discussed in Chapter 4.

The results for the broad regions show some large variations. But before we can discuss the possible reasons for these differences we need to have a digression on a conceptual issue about the value of employment elasticity in agrarian economies.

The determinants of employment elasticity in peasant agriculture

What determines the volume of employment in an agricultural economy of the Indian type – in which self-employed farmers provide the majority of labor input, and hired wage labor is only a part of the labor force? There are two different approaches to the question. These might be called the 'production function approach' and the 'disguised unemployment approach'. In the production function model the amount of labor used is determined by a profit-maximizing farmer much like in an industrial firm. The level of employment is then determined by the volume of output, the use of co-operant factors like capital, and the relative price of labor (wage rate). The labor used in this model is of standard efficiency. If the supply of workers in the region is larger than the demand the excess moves away to other occupations or regions or is openly unemployed.

In peasant agriculture dominated by family farms employment is determined more by work sharing than profit maximization. If supply of labor exceeds demand, and the opportunities of off-farm employment are limited, workers are not wholly unemployed but are absorbed in farm activities at a lower level of work intensity.

The disguised unemployment model has two important predictions for the level of employment per unit of land (or output), and hence on the value of the elasticity of employment with respect to output over time. First, poorer regions with low land productivity can be expected to be beyond the point of labor absorption at which a reservoir of 'disguised unemployment' has begun to accumulate. If then the labor-force growth exceeds the growth of agricultural output we can expect to see a further accumulation of surplus labor. The resultant high elasticity of employment would then just be reflecting the growing volume of 'disguised unemployment'. Second, the elasticity of employment in agriculture will be higher in regions in which the opportunities for non-farm employment are less. If we find (as will be discussed in the next section) there is a positive relationship between off-farm employment and land productivity, the two conclusions will reinforce each other.

Differences between 'broad regions' in employment elasticity

It will be recalled from the material presented in Chapter 6 that of the seven broad regions distinguished in our classification, regions 1 and 5 are the ones with high land productivity and low incidence of poverty. Regions 2 and 7 are the high-poverty regions, while the other three occupy intermediate positions on terms of income levels and poverty incidence.

The contrast in employment elasticity between the low-poverty region 1 and the high-poverty region 2 is striking (see Table 7.3). The former had a high rate of growth of agricultural output in both periods, actually increasing to the highest among all regions in the 1990s. But labor absorption in agriculture was quite low, turning negative in the 1990s. Most of the growing labor force was absorbed outside agriculture, partly due to rising wages and mechanization, and partly due to the high growth of off-farm employment. In region 2, on the other hand, the labor force had to be absorbed in the agricultural sector itself. When output growth declined in the 1990s to a very low rate, the increase in 'disguised unemployment' in the sector was reflected in a massive increase in employment elasticity.

Region 5 is the other low-poverty region which has succeeded in finding productive employment for its growing labor force with the highest rate of growth of urban employment in the 1983–1993 period. In the post-reform period the growth rate of urban employment slowed down significantly (see Chapter 5). Agriculture was called upon to absorb a larger proportion of the growth in the labor force – well in excess of the moderate growth in farm output. It is likely that this is the major reason for the jump in the value of employment elasticity in agriculture in this region in the 1993–1999 period.

Two conclusions follow from these examples. First, relatively high employment elasticity in agriculture could result, not so much from a higher rate of demand for labor with agricultural growth (as the production function approach would suggest), but rather from the fact that this sector serves as the reservoir for labor unable to find more productivity employment in other sectors (as the disguised unemployment hypothesis stresses). Second, there is some suggestion from the inter-regional variations given in Tables 7.2 and 7.3 that relatively higher employment elasticities are found in low-income regions in which opportunities for non-agricultural development have been small or has grown weaker over time.

It is not, however, possible to prove this suggestion conclusively with regression models using unit level data because our observations are for three single years separated by time, and as such subject to large variations caused by random factors.

Our tentative conclusion is that, with the existing pattern of development of the agricultural sector, the prospects for gainful absorption of labor in agriculture is not all that great. In fact increase in land productivity, and the resulting increase in income per worker in agriculture, is more likely to increase labor absorption through non-farm development which might be induced. It is to this topic that we turn in the next section.

Diversity of activities in agriculture

We have so far discussed employment in agriculture on the basis of the UPS and UPSS classifications of the employed workers in agriculture. These concepts are used by the survey to classify the employed respondents to allocate the latter to different occupations/industry on the basis of their major activity. The data collected this way pays no attention to the time spent by the workers in different activities. The CDS concept attempts a partial accounting of the time budget. It gives the distribution of person-days in different types of work undertaken by members of rural households.

All activities relating to production of crops are included in "cultivation". They comprise six manual and one non-manual activity (Table 7.4a). However, in all rounds of the NSS a little more than 40 percent of all person-days are classified in 'other cultivation activities' and this proportion does not show any definite change over time. The next in importance is harvesting which accounts 21–22 percent of cultivation activities, followed by ploughing and weeding (10–12 percent each). Note that 'other cultivation' is different from 'other agriculture'. The latter as seen in Table 7.4b account for a sizable proportion of the total rural households' activities: the most important of this type is 'animal husbandry'. Nevertheless a substantial part of this type of activity is also not definitely specified in the NSS codes.¹

It will be seen from Table 7.4c that rather more than a quarter of the rural person-days of work are spent in 'non-agricultural activities'. This fraction does not change much over time (not presented here), but there are interesting variations over the broad regions – which also do not vary much over time. The more prosperous regions 1, 4 and 5 have a larger share of time devoted to these activities. So has broad region 7–a high-poverty region which has a high man–land ratio and limited opportunities in agriculture (see Chapter 6). The regions of relatively high poverty incidence – regions 2, 3 and 6 have a relatively smaller proportion of time devoted to non-agriculture. We conclude that for rural households, diversification to non-agriculture is significant, and, across the 'broad regions', its relative importance in terms of labor time spent on such activities is inversely related to the incidence of poverty.

One would like to know what levels of income are generated by the labor time spent on such activities, and how they compare with income originating in agriculture. Unfortunately the 'thick' rounds of the NSS (on which much of the work on this book is based) do not collect data on the components of household income. But there was special survey of the NSS, the so-called 59th round which collected data on this topic as part of a general survey of farmers' economic conditions. The share of household income of farmers derived from off-farm activities is given in Table 7.5.

It is seen that the relative income in off-farm activities of farmers' households is quite low. This is as one expects since the off-farm activities recorded in this survey are marginal for the households concerned. Nevertheless, the substantive point remains that, in contrast to some other Asian agricultural economies, most notable Taiwan from its early stage of development, both the total income and income per unit of labor time generated in off-farm activities of farming households have been quite low in India. Even the more prosperous regions in India like regions 1 and 5 do not seem to have higher relative labor productivity in off-farm activities.

In the next section we consider the role of non-farm activities who 'specialize' in the non-farm sector in the rural economy. The 'thick' rounds of the NSS distinguish such households on the basis of occupation/industry of the principal earner in the household. We do not have income data but the welfare level of households can be approximated by the statistic of household expenditure per capita.

We need to be aware of the limitations of the main source of our data, the NSS, before proceeding further. First, a large share of employment in India is in the 'self-employed' category. There is an inherent difficulty of allocating income accruing from self-employment when more than one earner from the same household is in income-earning activity. Households from different self-employed activities by different members of the household would be typically pooled together. There is no way of distinguishing the individual contributions of individual earners. Hence the income we can deal with is household income, and we can normalize for the size of the household. Further, it is generally accepted that figures on expenditure given by the respondent in the household is more reliable than that of income. Thus we use the measure of household welfare as given by mean expenditure per capita.

When we are comparing levels of household welfare across sectors we need to identify the principal occupation of the household. This poses problems both conceptually and in terms of execution. The conceptual problem arises from the fact that a significant number of households will have more than one earner, and not all earners will be in the same category of occupation. The secondary earners might not be all wage earners. If they are working in the self-employed sector, they will be pooling their earnings with other earners of the household to create the household's pot of earnings. By assigning all the household income effectively to the principal occupation of the household we might be exaggerating the income – and the expenditure which it sustains – originating from this occupation.

Rural off-farm employment

It might appear at first sight that the pressure of excess labor supply on land and the incidence of 'disguised unemployment' in poor regions would be partly relieved if off-farm employment were to develop in a significant way in theses areas. We saw in the last chapter that while in the northern 'broad regions' off-farm growth seemed to add to the process of regional difference in growth over time, there was some suggestion that the southern regions, particularly in the 1983–1993 period, were able to compensate for the slow growth in land productivity through a more vigorous growth of the off-farm (and urban) sectors. What is the evidence on this point from the pattern of development of off-farm employment in rural India, taking all 75-odd NSS regions together?

Off-farm rural employment is a heterogeneous sector. Kijima and Lanjouw (2004), drawing on the evidence from a host of village studies, distinguish three major categories of NFS: (i) regular employment (generally salaried); (ii) casual employment (daily wages); and self-employed enterprise activities. The first category, often related to public-sector jobs created by rural development programs, are generally the most sought after as it not only offers higher earnings but more importantly stability of employment. 'Casual non-farm employment is generally thought to be less demeaning to a worker than agricultural wage labor, but returns may be only marginally higher'. Finally, the self-employed consist both a group of low income earners who are pushed into the sector, and higher-income workers with business activity. Kijima and Lanjouw report that analysis of the NSS for the last three 'thick' rounds shows that the overall employment share of the non-farm sector as a whole has hovered around 25–30 percent for all-India, with no evidence of any growth over time. Casual labor has been in the neighborhood of 6 percent, regular wage workers constituted 7–8 percent and 12–14 percent was the self-employed. The three components obviously have different distributional impact – regular workers and a portion of the self-employed in particular would tend to be recruited from the better-off economic classes.

The significant question about off-farm employment relieving the pressure of population on land is its relationship to the level of productivity (or income) in agriculture. There are two different hypotheses in the literature about this relationship.

The Johnston–Mellor hypothesis

In the traditional view, associated with the work of Johnston and Mellor, off-farm activity develops in response to the prior development of agriculture. High land productivity, such as was achieved in selected regions due to the green revolution, increases demand for off-farm goods and services, both in the rural areas and smaller towns. The growth of farm productivity and off-farm activity constitute a virtuous cycle of mutually supported development.

This model has also an important implication for relative productivities in the farm and non-farm sectors at different levels of rural welfare across regions. As already mentioned, the existence of an excess supply of labor in traditional agriculture is not compensated adequately by off-farm employment, and does not take the form of open unemployment. Agriculture is the 'residual' sector for the population which cannot move to other occupations or regions. Since there is no floor to self-employed income in this sector one sees a fall in the income of households' dependent on agriculture. In off-farm employment on the other hand, the level of wage earnings or business income will have a floor determined either by the reservation price of labor or the opportunity cost of capital. Thus we would expect to see regions with a low absolute level of income in agricultural households would also show a relatively lower ratio of agricultural to non-agricultural incomes. That is to say, the hypothesis is that in a cross-section sample of NSS regions the relative income in agriculture would be positively related to the absolute level of agricultural income.

The Foster–Rosensweig hypothesis

The contrary view has been most elaborately developed by Foster and Rosensweig (2004). They distinguish between 'traded' and 'non-traded' types of off-farm activities. While the latter could be a function of the development of the local rural economy and hence would be sensitive to the growth of agricultural income in the region, the 'traded' part is not necessarily tied to local development. Further, Foster and Rosensweig suggest that writers have over-emphasized the self-employed part of off-farm employment to the exclusion of wage earners. The development of business activity in the rural economy is expected to be a function of the growth of capital from outside the local economy seeking out labor at affordable cost. Thus low-wage regions with low land productivity would have a preferential pull on such investments. The proportion of employment in off-farm activities in such regions would accordingly be higher. Clearly this interpretation of the development processes in the rural economy outside agriculture emphasizes the importance of outside capital rather than capital generated by internal savings of rich farmers.

It is useful to note that the Foster–Rosenweig hypothesis has no particular prediction about the relative incomes in the farm and off-farm sectors. While we do have the scenario of capital migrating to less prosperous regions, presumably with lower agricultural incomes, we cannot expect to see any particular changes in the wage or income differences between the two sectors in the regions concerned without more specific indication about labor market dynamics.

Testing with NSS data

The analysis in Chapter 6 for broad regions revealed that the evidence on the basis of the seven regions distinguished leaned towards supporting the predictions of the Johnston–Mellor rather than the Foster–Rosensweig model. It was the pressure of population of land which seemed to be critical in the determination of the share of employment in the non-farm rural sector. Since a higher man – land ratio was generally associated with a lower level of per capita income and higher incidence of poverty, there was some positive relationship between income levels and the share of non-farm employment. Also the poorer regions tended to have a larger gap between the income per worker in the non-farm sector relative to the farm sector. It is the purpose of this section to go beyond the level of aggregation involved in the discussion of Chapter 6. We shall try to test the hypotheses in a more detailed and rigorous way with the help of all the observations available from the 70-odd NSS regions.

The partial correlation of RAPCE with selected variables

We first examine the relative importance of different variables affecting rural incomes (approximated by RAPCE), taking one variable at a time. The correlation matrices for the variables enable us to do so. The definitions of the key variables are as follows:

rapce_ci – Rural average monthly per capita consumption expenditure at constant prices adjusted for inter-state difference in prices.

uapce_ci – Urban average monthly per capita consumption expenditure at constant prices adjusted for inter-state difference in prices.

Rapce – Rural average monthly per capita consumption expenditure at current prices.

lnpro – Land productivity obtained by dividing value of output of crops at constant 1993–1994 prices divided by net sown area.

hn_ag – Ratio of income in the non-farm relative to the farm sectors. It is proxied by ratio of average monthly household mean consumption expenditure per capita of non-farm to farm households.

tur – urbanization ratio obtained as the share of urban UPS workers to total UPS workers.

tnfups – share of UPS non-farm labor to rural labor.

cul_nsa – net sown area per UPS worker involved in cultivation.

We define the variables in logs so that it is easy to examine the relative elasticity of RAPCE with respect to each of the variables from the regression models to follow. The correlation matrix is given in Table 7.6.

The more important conclusions are as follows:

1. The correlation of RAPCE with land productivity is quite high, showing an elasticity of 0.48. In fact it increased quite dramatically between 1983 and 1993, before falling off somewhat in 1999 (not shown in the table). Some of the reason for the low correlation in the 1983 round is the problem with inter-region price conversions in that round in particular. But even when we look at the results without these price corrections, the correlation coefficient of these two variables for 1983 at 0.39 is much lower than for the later dates. Evidently the importance of land productivity in the determination of the inter-regional variation in rural household welfare becomes substantially stronger after the second green revolution of the 1980s.
   
   
2. Non-farm employment is positively correlated with land productivity – supporting the Johnston–Mellor rather than the Foster–Rosensweig hypothesis. In fact the correlation co-efficient between tnfus and lnpro is at 0.5667 the highest in the matrix of Table 7.6.
   
   
3. As is to be expected from the last two results, the non-farm employment variable tnfups is positively correlated with RAPCE and the correlation value increased as much as that of yield between 1983 and 1993 and continued to increase somewhat in 1999. But the ratio of income per capita (as proxied by household expenditure) of non-farm to farm households is negatively correlated with RAPCE. The obvious inference is that in higher RAPCE areas the productivity per worker in non-agriculture falls relative to that in agriculture. A plausible interpretation is that in a cross-section view of the rural NSS regions, as non-farm employment becomes a source of increasing importance, the 'dualism' between farm and non-farm activities decreases.
   
   This is a second important finding of relevance to the Foster–Rosensweig thesis. Part of the reason why non-farm employment seems to be of more importance in poorer, low land-productivity regions is now seen to be because its relative productivity is higher in such regions due to a stronger incidence of 'dualism'–and not because a greater proportion of non-farm employment is found in them.
   
   
4. Both the urbanization variables tur and uapce increased their correlation coefficients with RAPCE dramatically between 1983 and 1993, specially the former. The former in fact increased marginally also between 1993 and 1999, while the latter fell slightly. All this can be interpreted in terms of a greater integration of the urban and rural economies, particularly the development of small towns which has been noticed as an important aspect of development since 1983.
   
   
5. The correlation of cul_nsa (the net sown area per cultivator) with RAPCE also increased steeply from 1983 to 1993 and further to 1999. Thus the impact of the farm sector on the rural expenditure also increased along with the bigger role of urbanization. All this contributed to a very large increase in the explanatory power of these variables in the regression to determine RAPCE.

The elasticity of RAPCE with respect to the key variables

What are the relative strengths of the variables studied above – particularly land productivity and non-farm employment on rural income levels? Regression models with the relevant variables put together were tried in order to decipher their joint impact on RAPCE. We tried the regressions both with the dependent variables rapce and rapce_ci (that is to say, both without and with price deflation at the regional level). While the values of the coefficients are not that different, slightly better fits were obtained for the former set. We therefore report and discuss the results from only this set.

The more important conclusions are summarized below.

1. The elasticities of RAPCE with respect to tnfs for the different rounds are given in Table 7.7. The elasticities with respect to land productivity and the cultivated area per worker are also included in the table. It is apparent that the elasticity of RAPCE with respect to farm income is much more than that of non-farm employment. (Note that the elasticity of farm income would be the sum of elasticities of land productivity and cultivated area per worker.)
   
   
2. The elasticity of RAPCE with respect to income generated in the non-farm sector is probably a more relevant variable to compare with the elasticity with respect to farm income. As already indicated the labor productivity gap between the non-farm and farm sectors narrows with increase in RAPCE.

The elasticity of RAPCE with this variable hn_ag is highest in the 50th round at–.213 in the multiple regression framework. Since this value is well below unity, it can be easily be demonstrated algebraically that the elasticity of the income ratio of non-farm to farm with respect to RAPCE will be positive but below the value of the employment ratio (tnfs). In other words the positive association of the proportion of employment in non-farm and the rural APCE is moderated to some extent by the narrowing of the productivity gap between the two sub-sectors because of the diminishing 'dualism' between them as regional rural income increases.

A surprising finding of our regressions is that the elasticity of RAPCE is very high with respect to the urbanization variables, particularly UAPCE. Table 7.8 reports the elasticity value for the two such variables used in our regressions. The relationship seems to be especially strong in the 50th and the 55th rounds and the value of the elasticities well exceed those of farm income and rural non-farm employment. The importance of urban development – particularly the development of small towns – for rural incomes in recent decades is evidently an important part of the changing rural economic scenario.

The results give unequivocal support to the model of a 'cumulative' process of development in the rural sector. Rural incomes are propelled by increased land productivity, and off-farm employment adds to the virtuous circle by responding to it positively. The gap in labor productivity between farm and off-farm sectors is reduced in this process.

The impact of liberalization on marginal farmers

We have seen that liberalization in the agricultural sector has been more on the external front in the post-reform years with limited effort to dismantle the regime of controls and subsidies in the internal economy. The impact on agricultural output growth has not been very impressive. At the same time several authors have raised the issue of adverse effect of post-reform developments on equity in this sector (see, for example, Chandrasekhar and Jayati Ghosh 1999; Sheila Bhalla 2005). Some evidence emerged in the analysis presented in Chapter 6 above that the post-reform growth process in the rural sector favored the more prosperous regions. There has also been an undercurrent of concern that important changes are taking place – which affect particular sections of the population adversely – and which are not captured by aggregate statistics. One of the issues is the impact of changes introduced by liberalization on small and marginal farmers.

'Distress inducing' growth

Liberalization has allowed competition from foreign countries even as world prices of some key agricultural commodities had a substantial downward trend in the nineties. The impact of these developments on distribution in the agricultural sector has been significant in some areas. A notable example is the case of Telengana in Andhra Pradesh. This case study has been analyzed by Vakulbrahmanam (2005) who has sought to generalize the case of Telengana as an instance of 'distress' of small farmers in the growth process fueled by liberalization.

Two crops, rice and cotton, account for almost 50 percent of the gross cropped area in Telengana. World prices of both have taken a dive while the domestic prices in Telengana have remained stagnant (ibid., Figures 2 and 3, p. 977). Indian manufacturers have begun to import cotton lint in response to its downward trend. In spite of this increased competition the area under cotton has continued to increase at a high rate. Between 1985 and 2001 the area under cotton in Andhra Pradesh increased at an annual rate of 17.2 percent, while the area under rice increased at only 3.3 percent, and the area under a number of coarse grains actually decreased (ibid., Table 16). This is because cotton is a high-value crop and also provides a higher level of employment per acre.

It is possible to provide a dynamic model in which with a large enough differential in output per acre between the commercial and food-crop sectors, the rate of shift of acreage to the former would be continuing even if the gap is reduced over time. But a reading of the article by Vakulbrahmanam reveals that there might be several supplementary factors at work. First, marginal farmers are more specialized in non-food crops because they do not have access to irrigated fields which is necessary for cultivation of rice. They are net buyers of food. So with the increase in the relative price of food their welfare declines and the response is to increase work on crop cultivation at the expense of leisure – a process emphasized by Chaynov (1966). Second, the reform process saw an increase in input prices – of power, credit and fertilizer, in particular, which squeezes the "net surplus" further. This effect is likely to be more important for marginal farmers, both absolutely and relatively, and might elicit the Chaynov type of response even further.

Vakulbrahmanam's data from the NSS showed that the average per capita expenditure for large and medium farmers increased significantly in the pre-liberalization decade of 1983–1993 but was nearly stagnant in the post-liberalization period of 1993/1994–1999/2000. The contrast was much sharper for marginal farmers and agricultural laborers – who actually suffered a significant decline in welfare in the second period. Consistent with this poverty decline was arrested in the post-liberalization period, and the agricultural growth rate in real terms was stagnant contrasted with its robust growth in excess of 4 percent per annum in the eighties. (ibid., Tables 2, 4 and 6).

The scenario presented above is for one region or district and is consistent with qualitative evidence about distress among farmers including suicide due to economic pressures. Part of the pressures arises from the fluctuations in market prices for non-food crops and is clearly related to liberalization not accompanied by adequate measures for crop insurance. How does the experience generalize to the all-India picture?

The trends for all-India

A significant point emphasized by Vakulbrahmanam is that while agricultural output in Telengana had grown over the 15 years of the last century at the healthy rate of 4.7 percent per annum, the growth in the real wage rate for agricultural labor had slowed down over this period, becoming virtually stagnant over the period of 1994–2000 (ibid., Table 6). We analyzed the data on output growth in agriculture and the wage rate (daily average earnings) of agricultural worker for 57 NSS regions between the 50th and the 55th rounds. The statistics are set out in Table 7.9. The much higher inter-quartile range for output growth implies that the means of the two variables are fairly close, but the median growth rate of the wage rate is much lower.

We also studied the growth rates of household welfare as measured by the household expenditure per capita for the different classes of agricultural households as distinguished by Vakulabrahmann. Table 7.10 gives the statistics of growth rates calculated.

It is seen that, unlike in the Telengana case, there is no monotonically decreasing growth rate of household welfare as we go down the landholding classes. There is, however, a difference between large farmers on the one hand, and the marginal farmers and the agricultural laborers on the other. In spite of the bottom quarter of the large farmers having negative growth, the overall growth rate of this class – either in terms of the median or the mean – was substantially above that of marginal farmers or agricultural laborers. There is some evidence supporting the Telengana phenomenon for all-India.

Region-specific evidence

Doubts remain nevertheless about the validity of the above analysis of central tendencies based on all-India figures averaged over many regions when the inter-regional variance is so high. In an alternative exercise we looked at the question if any significant trends could be found in the period studied by looking at region-specific growth rates for the household welfare of different classes of farmers. This was done by taking each of the lower farming classes in turn and then regressing the region-specific growth rate of per capita expenditure (measuring household welfare) of each class on the same variable for large farmers. As an example, the graph showing the scatter is given for one pair of the classes distinguished in Figure 7.2–namely the growth rate of APCE for the 'marginal' farmers. plotted against the growth rate for 'large farmers'. The variance is large but a regression line could nevertheless be fitted to the scatter with a significant slope. Similar scatters for the growth rate of each of the other three classes plotted against the growth rate of large farmers also show a significant positive relationship (not shown here).

Figure 7.2 Growth rate of consumption of marginal farmers vis-à-vis large farmers.

Table 7.11 presents the linear equations of the growth rates of each of four classes in the agricultural sector regressed on the growth rates of large farmers. Agricultural labor households are defined as those whose main source of earnings is wage labor in agriculture, whether or not they are landless or cultivate a small piece of land. The other landholding classes are distinguished on the basis of the size of their operational holdings.²

In spite of the relatively low value of R2 (suggesting there are many other factors behind the large inter-regional variance in growth rates of APCE), all the coefficients of 'b' are significant at an acceptable level. We find that even for medium farmers the growth rate is only a third of the rate achieved by the large farmers. Further, there is indeed a gradual reduction in the slope co-efficient as we move from medium to marginal farmers and to agricultural laborers. There does not, however, seem to be any difference between the coefficients for small and marginal farmers.

The results give credence to an aspect of the hypothesis that post-reform developments in the agricultural sector have helped larger farmers more than the marginal ones. But it should be remembered that the period between 1993/1994 and 1999/2000 which we have considered has not been a particularly prosperous one for agriculture. We would like to see developments in subsequent periods when relevant data are available from further rounds of the NSS.

Conclusions

In conclusion we can recount the more important results from the detailed discussions in this chapter:

1. Policies affecting the agricultural sector continue to favor the more prosperous regions.
   
   
2. The objective of policy should not be viewed as maximizing employment elasticity in agriculture. There is some evidence to suggest that employment elasticity is higher in low-productivity regions simply because agriculture, as the residual sector dominated by family farms, is best able to absorb 'surplus' labor.
   
   
3. Off-farm employment, both in the rural and the urban sectors, seem to be more important in regions with higher agricultural income – supporting the hypothesis of 'cumulative causation'.
   
   
4. There is disturbing evidence of post-reform developments favoring larger farmers more than the marginal ones and the landless.

Document(s) 7 de 16