The distributional impact of structural transformation in rural India: case-study evidence and model-based simulation

The North Indian village of Palanpur has been the subject of close study over a period of six decades from 1957/8 to 2015. Himanshu et al. (2018) document the evolution of the village economy over this period and point to two distinct drivers of growth and distribution of income. An early period of agricultural intensification associated with the green revolution saw an expansion of irrigation and the introduction of new agricultural technologies, leading to rising incomes accompanied by falling poverty and fairly stable, or even declining, income inequality. From about the mid-1970s onwards, a cumulative process of non-farm diversification took hold, and was associated with further growth and poverty decline but also a significant rise in income inequality. Such a process of structural transformation has been observed more widely in rural India. We construct a simple model of a village economy that captures several of the salient features of the Palanpur economy and society, and that is able to reproduce the distributional outcomes observed in the village. Our analysis suggests that while non-farm diversification occurred alongside rising inequality, the counterfactual of no diversification would in fact be associated with an even greater increase. We suggest therefore that non-farm diversification has in fact helped to contain growth in inequality, and has played a particularly pronounced role in reducing poverty. To the extent that other villages in India share features similar to Palanpur, our findings may also hold elsewhere.


Introduction
Rural India is home to 70% of the nation's population and about the same proportion of poor people in the country.This rural population resides mainly in villages -the 2011 Census reports roughly 800 million people living in more than 600,000 villages.Most of rural India's workforce (60%) remains primarily involved in agriculture, but in recent decades this sector's growth has lagged other sectors in the economy.The deceleration in agricultural growth has been offset by the emergence and growth of the non-farm sector; in 2011/12 nonfarm workers accounted for 40% of the workforce, nearly double that observed only ten years earlier (Himanshu et al. 2018).
Martin Ravallion had a long-standing interest in tracking, and studying, the evolution of poverty and inequality in India.His contributions in this domain were important and influential.A series of papers, many written in collaboration with Gaurav Datt, drew on nationally representative household survey data to highlight the critical role of both agricultural and non-farm growth in governing the participation of India's poor in its overall economic development process (Datt and Ravallion 1998, 2002, 2011;Ravallion 2000).An important recent paper examined the process of structural transformation in India, highlighting the connection between rural living standards and growth in urban areas.The paper suggested that rural wellbeing was responsive to growth in small towns, to a much greater extent than to growth in India's large cities (Gibson et al. 2017).
The aim of this paper is to build on this study of structural transformation in India by examining this process at the village level.In particular, our focus is on the impact of this process on village-level income inequality.Recent years have seen growing interest and analysis of income inequality in India, with, in particular, concerns being voiced about a significant increase in inequality over time (Chancel and Piketty 2017;Crabtree 2018).Himanshu (2019) surveys the evidence from secondary data and suggests that although there are clear signs of rising inequality in certain dimensions, this is not universally the case.Notably, while estimates of income inequality show a clear rise, the evidence from India's NSS consumption surveys is less clear-cut.Common to most of the discussions of inequality in the literature, is a focus on inequality at the national level.This is a natural place to look and is clearly relevant when comparing India internationally.However, aggregate inequality can readily mask inequality outcomes, and trends, at the sub-national level.At the village level, inequality could be rising while national level inequality was stable or even falling.This could occur if the underlying processes were leading to rising within village dispersion of income, but accompanied by convergence across villages of average income.Recent analysis by Mukhopadhyay and Garcés Urzainqui (2018) points to the possibility that such a process is indeed underway in parts of rural India.
Conventional survey data are not able to monitor inequality trends at the village level in India; their sampling design does not permit such a fine level of disaggregation.It is therefore difficult to monitor inequality in a given village, much less to make broad statements about village level trends (although see Mukhopadhyay and Garcés Urzainqui 2018).And yet there are grounds for interest in such local inequality outcomes.In rural areas, the village population likely serves as a reference group against which villagers compare themselves.Attitudes about the magnitude and direction of change in inequality are likely to be heavily influenced by village level trends.In general, rising inequality is likely to put pressure on village solidarity and the functioning of village institutions (Himanshu et al. 2018). 1  In this paper we report on the findings from a study of long-term economic development in a single village, Palanpur, located in western Uttar Pradesh.This village study points to two key drivers of economic development over the six decades between 1957/8-2008/9: agricultural intensification and rural non-farm diversification.The agricultural intensification process is often referred to as the "green revolution" and was associated, in India, with the introduction of new high yielding seed varieties (mainly wheat and rice), the expanded application of fertilizers and pesticides, and the spread of irrigation.Non-farm diversification refers to growing employment opportunities and incomes from non-agricultural sources.These drivers have both combined to generate rising per capita incomes and falling poverty rates, notwithstanding steady population growth and falling per capita landholdings.Their impact on income inequality appears to have been distinct, however.Agricultural intensification, which began in the 1960s, was associated with the expansion of irrigation and the adoption of new farming technologies and techniques.Income inequality in Palanpur was broadly stable and even showed some decline during the period up to the mid-1970s.In the subsequent decades, rural non-farm diversification increasingly took hold in the village, alongside further agricultural intensification.Although per capita income continued to grow, and poverty fell, income inequality rose significantly.
We borrow elements of Palanpur's economic structure and society to construct a simple simulation model of a small village economy.We impose on it the exogenous forces of demographic change, technological change, and non-farm diversification to which Palanpur has been exposed over time.We show that allowing this simple model to run over a five decade period returns a time series of inequality change and poverty decline that mimics reasonably closely that which was observed in Palanpur.We suggest therefore that other Indian villages -with similar economic and social structures -may be also be experiencing rising inequality alongside a general rise in living standards.
We study further how inequality in a village such as Palanpur might have evolved if the non-farm sector had not emerged.Conventional inequality decomposition analysis suggests that non-farm incomes have become the main contributor to overall inequality in Palanpur.Yet such decomposition techniques do not capture well the counterfactual of how inequality might have evolved in the absence of diversification.Our simulation model indicates that if the village had remained an agricultural economy, the rise in income inequality during the latter half of the study period would have been more pronounced: rather than exacerbating inequality, non-farm diversification has thus acted to attenuate a far sharper underlying trend of rising inequality. 2  In the next section we offer an abridged description of economic development and distributional change in Palanpur.We base this on the detailed analysis presented in Himanshu et al. (2018).We then present our simulation model and results in Section 3. Section 4 offers concluding remarks.
1 Luttmer (2005) documents that in the United States, controlling for income, subjective welfare of individuals is lower in more unequal neighborhoods.Similarly, Lentz (2017) finds that in Ghana, subjective welfare falls when neighbors become richer.Araujo et al. (2008) document, further, that in rural Ecuador, "elite capture" of community driven development projects is more likely in communities with high inequality. 2These findings also resonate with the classic study of two villages in Karnataka conducted by Scarlett Epstein (1973).

The village setting
Palanpur is a small village located in Moradabad District in western Uttar Pradesh.It has been the subject of close study since 1957-8, when it was first surveyed by the Agricultural Economics Research Centre (AERC) of the University of Delhi.A subsequent survey was fielded by the AERC in 1962/3.In 1974/75, Christopher Bliss and Nicholas Stern spent nearly a year in the village collecting data; their fieldwork culminated in the publication of a book on the Palanpur economy (Bliss and Stern 1982).The village was next surveyed in 1983/4 by Jean Drèze and Naresh Sharma, in collaboration with Nicholas Stern, and they visited the village again in 1993 for a quick resurvey.Findings from these data collection efforts were reported in Lanjouw and Stern (1998).In 2008/10, Himanshu and collaborators conducted fieldwork over two consecutive years and then returned again in 2015 for a quick resurvey.Analysis of the complete dataset, covering all of the survey years, was reported in Himanshu et al. (2018).Enquiring into the evolution and determinants of distributional outcomes has been a central theme in both Lanjouw and Stern (1998) and Himanshu et al. (2018).Scrutiny of trends in income inequality and of mobility patterns is possible with the Palanpur data because the data collection efforts covered the entire village population -not just a sample. 3he village is located in the plains of the Ganges river near the town of Chandausi (13kms), and the large city of Moradabad (36kms), which is also the district headquarter.A railway line connects the village to these urban centres, as well as to Delhi, some 220 km away.This railway line has been the primary means of access to the outside world over the study period, although in recent years, road access has also improved.Palanpur's population density, and proximity to urban centers, is not atypical in this part of northern India.Access to non-farm jobs often occurs through commuting; although migration rates have risen in recent years, it remains a relatively uncommon occurrence.
In early 2008 (the year for which the most complete data are available) Palanpur had a population of 1,255 persons, divided into 233 households (Table 1).Overall population growth was slightly above that for India in the 1950s and 1960s but substantially below the national average during the last 25 years.However, following adjustment for out-migration, the population growth rate in Palanpur is found to be very similar to India for the same 25-year period.
Although there are 8 caste groups in the village, and a few additional individual caste households, the three main castes in the village are the Thakurs, Muraos and Jatabs.In 2015, the Thakurs and Muraos each represented roughly a quarter of the village population while the Jatabs represented another 15 percent.The remaining third of the village was comprised of smaller caste groups, and included as well two Muslim groups together representing another 15 percent of the population.Thakurs occupy the top of the village hierarchy in terms of status.They continue to be powerful economically and politically.They were the first to move into the non-farm sector in a major way but have now been joined by other castes.Muraos, ranked just below the Thakurs, are a cultivating caste and take pride in their agricultural skills.Jatabs, at the bottom of the village hierarchy, remained economically and socially marginalized until around 2005, but have seen a rise in population share over time.Their economic conditions have also improved over the years.The Muslims in the village have become increasingly active in non-agricultural self-employment activities.
One of the key findings emerging from the 2008/9 survey data is that the circumstances of the Jatabs is showing significant improvement, both absolutely and relative to the rest of the village.This process is paralleled by a clearly discernable expansion of non-farm employment in the village economy.What is key is that Jatabs appear now to be enjoying greater access to non-farm opportunities than in the past, and this is translating into rising per capita incomes and upward mobility.We provide a brief documentation of these trends further below.

Agriculture
Throughout the survey period, the economy of Palanpur has essentially been one of smallscale farmers.The proportion of landless households is relatively small by Indian standards and there are no clearly outstanding large-scale farmers.The bulk of economic activity is in agriculture, and since the late 1950s the village has seen agricultural practices transformed in connection with the spread of irrigation, the introduction of new seed varieties, fertilizers and pesticides, the emergence of rental markets for agricultural equipment, and the introduction of new crops.
In India as a whole, agricultural GDP grew at an average rate of 2.8 percent since the early 1950s, with rates above 3 percent per annum after the 1980s (Himanshu et al. 2018).Agricultural growth rates in Palanpur have mirrored the national income growth rates for most of the period post-Independence.Agriculture remains of great importance to village livelihoods, notwithstanding the growth of non-farm activities in recent decades.Eightyfour percent of Palanpur's households reported income from agriculture in 2008/9 although only 23 percent were dependent on agriculture alone.Key to the agricultural development process over the survey period has been the expansion of irrigation from around half of village land at the beginning of the survey period to 100 percent by the 1974/5 survey years, as well intensification of farm capital in the form of farm mechanization that has been both land-augmenting and labor-saving.While farm mechanization has raised agricultural productivity, it has also played a role in enabling the release of labor to nonfarm activities.Additional forces of agricultural change have been the shift of cropping patterns towards higher value crops, such as mentha (spearmint), as well as improvements in farming practices.Wheat and rice yields have increased markedly, especially between 1957/8 and 1974/5, although also continuing thereafter.Although costs of cultivation also rose over time, the rising yields have been sufficient to ensure that village cultivation income in Palanpur grew over time in real terms.

Expansion of non-farm employment
Despite intensification and technological change, and given that overall land availability is fixed (the village lands cover roughly 400 acres -or about 160 hectares), agriculture is increasingly unable to support a steadily growing village population.Consequently, a growing share of village income comes from non-agricultural sources.By 2015, non-farm activities represented roughly two-thirds of total primary employment in Palanpur and accounted for nearly 60 percent of average household income in 2008/9 (see Fig. 1 and Table 2).This compares to less than 10 percent of employment and 20 percent of income in 1957/58.In Palanpur, the increase in the population, the decline in per capita landholdings, and the release of agricultural labour through mechanization, has facilitated, incentivized, and accompanied a process of non-farm diversification.Better access to towns and cities via improvements in railways and communications infrastructure, particularly mobile phones, has helped villagers to find jobs and has led to a growing number traveling outside Palanpur for their employment.Over time, there has been substantial change in the range and nature of the non-farm jobs available.While these jobs were restricted mainly to traditional caste-based jajmani services and a few regular jobs in the railways during the first two survey years of the 1950s and 1960s, over time there has been a significant expansion beyond the network of traditional services.Non-farm employment is now found in a range of establishments such as the cotton factory, a sugar mill in Bilari, a paper factory in Nagalia, marble polishing units in Chandausi, and casual labour in brick kilns.The jobs in the non-farm sector can largely be categorized into two kinds: i) low-paying casual and menial activities; and ii) more attractive and higher-income opportunities.Casual wage employment would fall under the first category while the latter includes betterpaid regular jobs (often government provided) as well as some profitable self-employment activities.At the same time, the lower-paying non-farm jobs remain more remunerative than agricultural labour, and many allow for more frequency of employment whilst retaining the flexibility to work in agriculture.Some activities, like marble polishing, lie between the two ends of the spectrum.
The casual non-farm sector in Palanpur has registered the highest growth in employment in recent decades, notably in activities related to the construction sector.The rate of growth in casual employment has been followed by self-employment in employment terms.Self-employment has seen the fastest income growth in Palanpur by a substantial margin.Entrepreneurship has been striking.Regular wage jobs have declined both relatively and absolutely-there has been very little growth in the number of these jobs after the early 1990s.
In the face of declining per capita landholdings and lower cultivation incomes households have pursued a broad spectrum of livelihood strategies.Himanshu et al. (2018) indicate that participation in the non-farm sector varies across households depending on the size of their landholdings and caste affiliation.Casual non-farm employment is the primary source of employment for the landless and the near landless as the rewards from agricultural labour become less attractive relative to the opportunities outside.For households with landholdings above 30 bighas (slightly less than 5 acres) non-farm participation is less common, although there are a few engaged in relatively high-earning, regular employment or particularly remunerative self-employed enterprises.
A majority of households in Palanpur now belong to the small and marginal land farm category and, for these, non-farm activities are playing an important role both in increasing income and diversifying risk.Dependence on cultivation has declined and the constraints on income related to land have loosened.Although households rarely exit fully from cultivation, the extent of participation in it is determined by opportunities in the non-farm sector.Those with well-paid regular jobs reduce their involvement in cultivation significantly while those with casual sector jobs remain more dependent on cultivation in order to maintain levels of income and to manage variability.While land endowment plays a role in the basic occupational decision concerning participation or not in non-farm activities, access to these jobs, especially regular jobs, is also influenced by the caste affiliation of an individual.Access to regular jobs is often determined by an ability to pay bribes, as well as by influence, contacts, and networks based on caste and kinship.As a result, such jobs tend to be held by relatively advantaged caste groups such as the Thakurs.The historically disadvantaged Jatabs, on the other hand, are poorly placed to find regular non-farm employment.Caste networks also play a role in some casual non-farm jobs, as for example, for work in the Moradabad railyards, where Thakurs are also predominant.
Recent decades have seen a marked increase in participation by Jatabs in the casual wage non-farm labour force, where there are few barriers to entry.This process has led to improvements in their economic circumstances.The traditionally cultivating caste of Muraos originally displayed some resistance to non-farm participation-preferring to focus efforts and attention on cultivation.But declining agricultural incomes over time (due to declining per capita landholdings) have led some of them to belatedly pursue opportunities in the non-farm sector.
There is little evidence to show that education plays a key role in determining access to non-farm jobs.Most of the jobs available in the non-farm sector are unskilled in nature and formal education does not appear to be essential.Women are greatly under-represented in the non-farm labour force of Palanpur, and indeed in the entire labour force of the village.There are a few examples where women do work alongside their husbands in nonfarm activities-for example, as part of a family group producing bricks in brick kilnsbut overall it appears that social restrictions on women working for wages as non-farm labourers, salaried employees (except for some government jobs), or entrepreneurs, continue to hold.Economic activities in agriculture within the village are also circumscribed for women, although there is some indication of increasing cultivation activity for Murao women as participation in non-farm activities by Murao men starts to increase.
While full migration from Palanpur is not particularly common and is not yet increasing as a proportion of households, the related practice of villagers commuting from Palanpur on a daily basis, or for periods of short duration, is both common and increasing over time.
Commuting permits villagers to continue to reside in Palanpur, and maintain some involvement in cultivation, while they access an ever wider range of non-farm job opportunities in the surrounding area and nearby towns and cities.Migration seems increasingly associated with declining land per household.It can be facilitated by caste and family networks.

Poverty and inequality
The richness of data covering all households in Palanpur for a span of many decades permits an analysis of the dynamics of poverty and inequality at a level of detail not normally available from secondary data sources.Poverty in Palanpur was very extensive in the early survey years-over 80 per cent of the population was classed as poor during the first two rounds (Table 3).The growth in incomes associated with expanding irrigation in the late 1950s and the 1960s, and the green revolution technologies and methods in the late 1960s and early 1970s, led to a sharp decline in poverty, with the headcount ratio falling to less than 60 per cent by 1974/5, and remaining at roughly that level in 1983/4.Poverty then fell again sharply after 1983, with non-farm employment playing an important role in improving the fortunes of many, including those at the bottom of the distribution.A similar picture emerges with mean consumption expenditure estimates which changed only moderately in the first two decades but had increased strongly by 2008/9.
Table 4 presents various measures of inequality calculated on the basis of this indicator, for each of the survey years.Between 1957/8 and 1962/3, the Gini coefficient rose from 0.336 to 0.353 and then fell back sharply by 1974/5 to 0.272.The remarkable decline between 1962/3 and 1974/5 was the likely consequence of three principal factors.First, with the investment in irrigation in the 1960s and the advent of green revolution methods in the late 1960s and the 1970s there was a significant expansion in the use of modern agricultural technologies.The distributional 'incidence' of the expansion of irrigation was particularly progressive.Whereas previously only a few, better-off, farmers had been in a position to irrigate their land (using 'Persian wheel' lifting technologies which required digging and maintaining a large well and complementary draught animal power), this period saw the expansion of irrigation to all farmers.By 1974/5 all village land was irrigated.Second, 1974/5 was also a particularly good agricultural year in terms of harvest quality in Palanpur.As a result, those who had spent less on inputs were less at risk from lower or negative incomes in the face of a bad harvest.And 'errant' farming practices (late sowing, poor weeding, etc.) tended to be less severely penalized.
The third factor to contribute towards an equalization of income in 1974/5 was that the distribution of land cultivated in Palanpur was more equal in that year than in other years.This was mainly a result of a fall in the proportion of land owned by Thakurs due to a few land sales.Moreover, during this year tenancy and sharecropping practices displayed a clear pattern of large landowners leasing-out their land to those with smaller landholdings.In subsequent years that pattern tended to be more mixed, including the more frequent observation of cases of 'reverse tenancy', where households with small holdings lease-out to those with more land.
Between 1974/5 and 1983/4 inequality increased but remained lower than in 1957/8 and 1962/3.A combination of factors helps to explain the rise.With the ongoing intensification of agriculture, the Muraos as a group, already with the Thakurs amongst the more prosperous groups, experienced improved relative prosperity due to higher returns from cultivation. 4By 1983/4 the Muraos had even surpassed the Thakurs in terms of per capita income.In addition, in 1983/4, new non-farm employment opportunities were becoming increasingly available, and were taken up mostly by villagers from economically better-off backgrounds.Due to a disappointing harvest in 1983/4, the income gaps were further widened between those who derived some earnings from outside and those who were entirely dependent on agriculture.And there was wider dispersion within cultivator incomes, influenced by spending on inputs and the relative impact of the poor harvest.
In the most recent survey for which income data are available, conducted in 2008/9, the Gini Index, at 0.379, was at its highest level compared to all other survey years.The sharp increase in inequality between 1983/4 and 2008/9 merits further examination.While changes in inequality across the early survey rounds can be understood in terms of the impact of expanding irrigation and green revolution technologies and methods on agricultural incomes, as well as the varying efforts or abilities of Palanpur households to create improvements in agricultural productivity, the distribution of income in later survey rounds would appear to be more significantly influenced by the expanding rural non-farm sector.This follows straightforwardly from income decompositions that assess the contribution of different sources of income to overall income inequality.
Table 5 reports results from a decomposition of the Gini by income sources. 5The share of income from non-farm sources in total income has increased over the survey period from 13 percent in 1957/8, to 46 percent in 2008/9.In contrast, the share of cultivation income has declined from almost 50 percent in 1983/4 to 30 percent in 2008/9.In addition, there has been increasing divergence in non-farm incomes which can be observed from the rise in the non-farm income source Gini (G k ).It increased from 0.59 in 1983/4 to 0.64 in 2008/9.These two factors, along with the Gini correlation coefficient, give the contribution of various income sources to overall inequality.In 2008/9, the contribution of cultivation income was 20 percent while that of non-farm income was 58 percent.The corresponding figures in 1983/4 were 63 percent and 22 percent respectively.The rise in the contribution of non-farm income to inequality, as suggested by this decomposition, is dramatic in those 25 years.
While this analysis is suggestive, it is important to note that these decompositions are essentially accounting exercises.They do not tell us what would have been the evolution of poverty and inequality absent the emergence of non-farm employment opportunities.While non-farm employment has clearly shaped the distribution of income in Palanpur we should recognize that in the absence of such non-farm incomes, a different development dynamic would have taken hold in village.We explore the distributional consequences of this counterfactual in the next section.

A simulation model for tracking distributional outcomes
The distribution of income, however defined and assessed, is a highly derived concept, with multiple determinants.Conversely, the distribution of welfare is likely to affect many socioeconomic variables.In this section we build a relatively simple model of a village economy that highlights some of the determinants of inequality, while ignoring the reverse impact of inequality on those determinants.Our objective is a model that allows the impact of drivers of inequality to be studied in isolation.
We describe the building blocks of the model and its calibration to data from Palanpur.Inspiration for the model comes from the Lewis model, which can be seen as a simple explanation of a trend in inequality that follows the Kuznets hypothesis.As described in Section 2 above, powerful forces of change have shaped the distribution of income in the village.Technological change in agriculture, the expansion of non-farm employment opportunities, and demographic change have been influential, but largely exogenous to the village.We are not only interested to scrutinize further how the distribution of welfare in Palanpur has been shaped by these factors.We also want to gauge how welfare might have evolved in their absence by examining counterfactual scenarios.
We conclude by assessing how our model-based simulations have informed our understanding of economic development in Palanpur.Pointing to the likely general applicability of the model's structure to other villages in this part of Uttar Pradesh, and perhaps beyond, we suggest that our findings may well be of broader relevance.

Model description and calibration
The model has an annual recursive structure.The unit of time is a year, which simplifies calibration of the model.

i) Population dynamics
We assume people live for exactly 70 years.They live in single-person households, so no distinction is made between individuals and households.Between ages 15 through 50 they individually produce offspring with constant probability per year.Persons over age 14 earn an income.For implementation we generate a population with an age distribution that is more or less in equilibrium, while at the same time producing the population growth rate observed in Palanpur 1958Palanpur -2009 (i. (i.e. 2% per year). 6The model allows for three 'castes' representing the three largest castes in Palanpur: Thakurs, Muraos and Jatabs; the population dynamics for all three castes are the same. 7The model thus ignores decisions on household size and the impact of such decisions on economic outcomes.

ii) Economic dynamics
Economic dynamics in the model are a combination of occupational and income dynamics.For income earners (adults between 15 and 70 years of age) we assume that incomes evolve according to an autoregressive process: where it is a standard normal random variable, independent across time and individuals.The parameters of this process ( , , ) are specific for caste, occupation and epoch, as , further explained below.The year before children reach the age of 15 we initiate them with an income that is related to their parent's income according to, dropping individual index i: Children then earn an income from the age of 15.The full model is depicted in Fig. 2. Figure 2 indicates that the model generates individuals from the three castes as well as their offspring.For each individual the model further simulates a sequence of occupations and incomes.Thus the model can be used to simulate the interplay of population, occupation and income dynamics as well as derivatives of those, such as inequality.However, the stochastic processes underlying the simulations are assumed to be independent across individuals, effectively ignoring any interaction between them.Model parameters are estimated and set to reproduce certain statistics from Palanpur's recent past, thereby providing a more or less plausible explanation of these statistics.
In the implementation below we take the parameters of population, occupation and income dynamics as exogenous.For a small village, such extreme assumptions seem justifiable.As noted above, there is no feedback of inequality or other model outcomes to the model's parameters.

iii) Calibration
Many steps in the model are partly determined by random variables.For most of the analysis we have fixed these underlying random variables by setting the 'seed' of the random number generators.Therefore, unless stated otherwise, model outcomes should not be looked at as 'expected' or average model outcomes; they are the outcome of a particular simulation path.

Population
As mentioned above, the model is calibrated to data from Palanpur.A population growth rate of 2% is assumed (see footnote 8) for all three castes ('J', 'M', 'T') starting from an age distribution that is approximately in equilibrium.In this paper we do not explore the possible effect of differential population growth on inequality.

Occupation dynamics
Individuals from each of the three castes earn either exclusively agricultural or non-agricultural incomes.With three castes this would amount to 3 × 2 = 6 occupations, but we have pooled the agricultural and non-agricultural occupations of the 'M' and 'T' castes, resulting a total of 4 occupations.To obtain transition probabilities between occupations, households' total incomes in a particular survey year were classified as agricultural or not depending on whether the total from agricultural income components exceeds that from the nonagricultural income components.All income components that are directly related to agriculture have been classified as agricultural, including a number of categories termed 'non-cultivation' in Table 2 above.This then allows estimation of occupation transition probabilities.For instance, Table 6 describes transitions for the 'J' caste in the period 1974-1983 in absolute numbers and as annual transition probabilities.In a few cases the raw transition numbers had to be adjusted to allow derivation of annualized transition probabilities.8

Income dynamics
The first step in setting the parameters of the income generating process is a regression, per caste-occupation combination, of a household's log per capita income in a particular survey round to log per capita income in the previous round, linking households through their ancestor's household number.This amounts to 16 regressions, reflecting 4 transition periods and 4 caste-occupation combinations.After this step the results of the regressions are annualized.It can be verified that parameter estimates α, β > 0, σ2 based on a period covering k years can be annualized to parameters , , 2 as follows: Muraos and Thakurs, 1974-83, agriculture 1974-1983  We the procedure Table 7 for the period 1974-83.Complications arise if the regression coefficient on lagged log income is negative.In those cases we have set the coefficient to zero, amounting to annual independent lognormal income draws for all years in a period.Table 8 shows that negative autoregressive parameters are found in half of the regressions, suggesting that inter-and intragenerational income mobility is considerable.
After having thus obtained annualized regression parameters, in the next step we have calibrated and and 2 to better match the overall Gini and the caste occupation-specific income distributions. 9The resulting fit of the model can be gleaned from the figures the Appendix.The good match is no surprise: it only shows that the model is flexible enough to fit the income data from the Palanpur surveys well.For instance, the model has enough degrees of freedom to exactly replicate the Gini coefficients from the survey in the baseline scenario.
Recall that model simulations are outcomes depending on a particular, fixed, set of draws for the underlying random variables.To gauge the importance of random drivers of outcomes in the model, Table 9 below gives the mean and standard deviation of Gini coefficients, across multiple independent baseline simulations. 10They suggest that for such a small population as Palanpur, changes in the Gini between 1958 and 1983 could very well reflect random instead of structural fluctuations.On the other hand, the increase in inequality between 1983 and 2009 clearly stands out as a significant change compared to the period before 1983.
Table 10 shows the analogous results for the poverty headcount rates.The close match between baseline headcount rates and the data is the result of calibrating each year's poverty line to reproduce the poverty head count data (up to rounding error due to finite size).Data model outcomes suggest that poverty decline was fastest in the 1964-1974 period, at almost three percentage points per year.These and subsequent tables of poverty and inequality rates are based on simulated adult incomes.Children's income or their impact on household welfare have been disregarded.

Counterfactual Simulations
In this section we present simulations to interpret trends in the income distribution.In all simulations we have fixed the underlying random variables for better comparability.However, it should be kept in mind that small differences in the order of 1 to 2 percentage points could very well be the result of random fluctuations.

i) Occupational dynamics
In the first simulation households remain in their (or their ancestor's) occupation for the year 1957/8.This amounts to setting all occupations to 'agriculture' since only one of the 1957/8 households (a Thakur household) was classified as non-agricultural that year.As can be seen from the aggregate results in Table 11, this scenario almost reproduces the baseline Gini values, but it points to a somewhat larger increase in inequality between 1983/4 and 2008/9.The same conclusion holds for headcount rates except for the final year 2009, with considerably higher poverty.This suggests that moving out of agriculture has contributed to poverty reduction, especially after 1983.These overall trends are confirmed by the actual data observed for Palanpur, with a Gini coefficient among agricultural earners of 0.372 in 2009 compared to 0.338 for non-agricultural households.

ii) Technological change
In this scenario we simulate the effects of technological change.We present two scenarios.In the first scenario the income processes are kept at their 1974 specification, amounting to annual draws, specific for of the four occupation groups (Jatabs: agriculture and Muraos and Thakurs: agriculture and non-agriculture),11 while keeping occupational transition as in the baseline scenario.In the second scenario we fix agricultural incomes at their 1958 specification, while increasing the mean to reflect overall income growth.We therefore interpret this scenario as simulating 'neutral' technological change in agriculture.The findings are presented in Tables 12 and 13.The outcomes for the first scenario can be interpreted as the result of occupational mobility, net of changes in income processes.They suggest that, by itself, the increase in the number of households classified as non-agricultural has not led to major changes in inequality or poverty.This seems to contradict the previous results (Table 11), which suggested that staying in agriculture would have led to less poverty reduction over the period 1983-2009 as well as a somewhat larger increase in inequality.The reason is that in the current set of scenarios the parameters of the income process have been kept at their 1974 value, while in fact non-agricultural incomes have especially increased in the 1983-2009 period.For instance, the Jatabs' incomes (in current prices) earned in agriculture grew 5 percentage points slower than non-agricultural incomes.The outcomes of Table 12 therefore seem driven by 'price changes' (or rather productivity changes), not quantity changes.The second scenario, 'neutral agriculture' shows stronger poverty reduction and a more or less constant inequality over the simulation period.Strikingly, the increase in inequality between 1983 and 2009 is fully absent.It is therefore tempting to conclude that this increase is due to developments in agriculture.However, such a conclusion cannot be drawn from a simulation model.At best it provides a hypothesis for further investigation.

Concluding remarks
The highly stylized model that we have constructed and calibrated on Palanpur data is able to track the true Palanpur data rather well.Nevertheless, its relative simplicity sacrifices many features that may play an important role in Palanpur's society.To mention just a few: household size decisions, mutual dependence of economic processes, technical change and migration are all ignored.Therefore we cannot claim to have provided a realistic model of Palanpur.Rather what we suggest is that a village which resembles Palanpur in certain basic ways, notably a breakdown of population into certain sub-groups that engage in agriculture and participate in the nonfarm sector in ways that are broadly similar to what is in might expect to see living standards like poverty inequality evolve in similar ways.It is obviously unreasonable to study a village like Palanpur closely and to then assert that the experiences and processes observed there will hold in many, or all, of India's villages.What our modelling exercise has shown however is that we need not insist on the exact duplication of all of Palanpur's characteristics for a similar evolution of outcomes to become plausible elsewhere.This finding lends support to, and possibly helps to interpret, the finding in Mukhopadhyay and Garcés Urzainqui (2018) that within-village inequality is rising in many parts of the country.
A further insight from the model relates to our findings from the counterfactual simulations.Decomposition exercises, such as that described in Section 2, might have been interpreted to indicate that rising non-farm incomes have been responsible for the rise in observed inequality in Palanpur between 1983/4 and 2008/9.What our counterfactual simulations suggest is that inequality and poverty might actually have been even higher in the absence of nonfarm diversification.These findings are not contradictory once one realizes that the inequality decomposition does not pose our counterfactual question of how inequality would have evolved in the absence of nonfarm diversification.Rather, the decomposition unpacks a given year's distribution of income and attributes a contribution to nonfarm earnings based on the observed share of nonfarm income in total village income, and its distribution in that year.In that sense, the decomposition is only an accounting exercise, and needs to be interpreted carefully.
An important implication of this particular counterfactual analysis is that it points to the possibility that increases in inequality in village India might be expected to be larger in those villages that have failed to see significant nonfarm diversification over time.The common perception in the literature is that non-farm income diversification is likely to be force for rising inequality in rural areas.Instead, we suggest that if the diversification process resembles that observed in Palanpur -with the poorer segments also gaining access to the non-farm occupations-then the diversification process may actually be acting as an important driver of poverty reduction, and a brake on further widening of inequality.

Table 1
Basic population indicators of Palanpur

Table 2
Income shares in Palanpur over time(%)

Table 3
Estimates of poverty headcount rates in Palanpur Mean income/consumption are at 2008/9 prices deflated by CPIAL for respective years.The poverty line employed here is based on the official poverty line for rural Uttar Pradesh from the Planning Commission (Planning Commission 2009).This line yields a consumption-poverty rate of 38.3% in Palanpur in 2008/9.To determine a comparable income-poverty line, the per capita income level associated with a poverty rate of 38.3% in 2008/9 was obtained, and then deflated back using CPIAL price indices to obtain the income poverty rates for the earlier years

Table 5
Inequality decomposition by factor componentsInequality in source of income k (Gini coefficient (G K ))

Table 9
Gini coefficients: data and baseline scenario outcomes 9Calibration consisted mainly in small adjustments of and 2 estimated by the income regressions.10Thestandarddeviationsaresimilarto those obtained from bootstrapping the Palanpur data.Note that the Gini coefficients from column 'Baseline data' in Table10differ from the Gini coefficients in Table4.This is because we have included only households from the Jatab, Murao and Thakur castes.

Table 11
Simulation results: baseline and no occupational change (NOC) scenarios

Table 13
Headcount rates: baseline and counterfactual scenarios