The secondary data we used in this study are from the national censuses (every 10 years for the population and every 5 years for agriculture), as well as major national surveys carried out occasionally on smaller samples: the NSS (National Sample Survey) and the IHDS (India Human Development Survey). But even in India where we find one of the best data collection systems on labor issues among developing countries, they have limitations especially when it comes to agricultural wage labor. In large household surveys, Oya and Pontara (2015) highlighted for example: (i) the problematic notion of 'main job-holding' and reference periods in contexts of multiple, irregular and highly seasonal occupations; (ii) the unclear distinction between wage employment and self-employment; (iii) under-reporting of stigmatized jobs (such as physical work for higher castes or women in many regions of India); (iv) inadequate translations into vernacular language of statistical categories and concepts. In addition, there is lack of technical data on the more or less labor-intensive nature of the diverse cropping and livestock farming systems and the distribution of agricultural work over the year. To overcome these limits, we therefore had to rely primarily on our own fieldwork.

The main aim of our fieldwork was twofold: (1) to measure, at the farm level, the amount of work provided and its distribution between family members and wage laborers; (2) to understand why laborers are hired: is it a necessity or an opportunity?

To meet the first objective, the strategy was to reconstitute annual working calendars for the different crop and livestock activities in each farm surveyed, identifying the operations and tasks carried out as well as the workload associated with each of them (in hours or days, with a day representing 8 h of work) during each agricultural season. By comparing this calendar with the pool of family labor, it would then be possible to identify the periods when it was necessary to hire labor to cope with the workload.

To meet the second objective, the strategy was to assess and work with three economic indicators. The first is daily (or hourly) labor productivity, equal to the agricultural value added divided by the number of days (or hours) of total work required to create it (Aubron et al., 2009), regardless of whether this work is provided by family members or wage laborers. If this labor productivity is higher than the local laborer's daily or hourly wage, we can then assume that a wage worker produces more economic value than he/she costs, hence hiring him/her is more of an opportunity than a necessity. The second indicator is the gross margin (value added minus salaries paid to the laborers) per day of family work. The third indicator is the agricultural income per family member obtained within the farm. When the family running the farm also works as hired labor on other farms, locally or by migrating, the income from wage work was included in the total income related to agriculture. Here, we could not add wages from non-farming activities (this is one of the limitations of our research) but it is worth noting that nonfarm diversification was not systematically observed in our study, with many of the poorest households working solely in agriculture.

Surveys to meet the above objectives take time and without significant funding, can only be carried out on a limited number of farms, on a purposive sample. To do this, we carried out what is called an “agrarian diagnosis”, using the conceptual framework of comparative agriculture (Cochet, 2012, 2015), in 13 small study regions in India (Figure 1). These 13 small regions were gradually selected in the framework of a research project dealing with the role of livestock farming in Indian agriculture (IndiaMilk project, 2014–2019) so as to cover a range of contrasting situations in terms of biophysical environment, land tenure, historical development of irrigated agriculture and dairy farming. Thanks to a characterization of the biophysical milieu and a reconstitution of the recent agrarian history based on interviews with older farmers (20 to 50 in each small region), the agrarian diagnoses allowed us to rapidly grasp the diversity of the cultivated land and to capture the historical circumstances that led to the way land is distributed amongst farmers in each area (Cochet and Devienne, 2006; Lacoste et al., 2018). This allowed us to establish a preliminary typology of farms in each small study region.

In a second series of interviews with a purposive sample comprising 30 to 60 active farmers in each of the small regions, we carried out in-depth studies of the production systems in place, the associated workload and economic results, in such a way that farm types identified during the first phase were covered and gradually refined. The interviews dealt with the structure and functioning of each type of farm at both the technical and socio-economic levels. The structure of the farm refers to the cultivated area, equipment and buildings, and the labor force. The functioning of the farm is defined by cropping practices (choice of crops and cultivars, crop rotation and associations, calendar of technical operations for each crop, associated working calendar and workload, yields, etc.), and livestock farming practices (choice of breeds, breeding, feeding and animal health management, use of outputs, associated working calendar and workload, yields, etc.). It also includes flows of matter and energy between crops and livestock, as well as the work organization at the farm level. The social relationships considered involve land (e.g., land rental or sharecropping), work (e.g., employment of daily laborers to carry out certain tasks), capital (e.g., loan to dig a borewell), agricultural inputs and products (e.g. milk delivery to a cooperative). Data on prices, agricultural laborers' wages and equipment lifetime were collected through the same interviews to calculate the economic indicators.

Archetypes of each production systems can then be modeled. Similar to Weberian ideal-types, they do not correspond to real farms nor to the averages of the three to eight farms surveyed by farm type. They are constructed models (Aubron et al., 2016). Considering all the information gathered during each interview, the aim is to identify what structurally characterizes the type studied, differentiating it from what is provoked by a specific conjuncture (Perrot and Landais, 1993), such as the illness of one of the family members in the past year. Step by step we thus carried out an “informed reduction of the observed variation” (Lacoste et al., 2018). The data presented in this article (Tables 1, 2 and Figures 6, 7) derives from this modeling process, and each farm mentioned corresponds to one of the archetypes we modeled.

These archetypes are contrasting cases showing the diversity of farms in each small region even when one of them is largely dominant. Their use in this article does not seek to statistically represent Indian farms but to capture, document and understand in depth main typical cases dealing with our research question even if some of these are statistically marginal. For example, our archetypes are not accurate representations of rainfed agriculture because “patronal agriculture” that we have tried to document and understand is mainly present in irrigated areas. Moreover, for lack of space, we were unable to present all the results obtained in the 13 small regions studied, but the reports for each of them are available on request from the authors.

In this preliminary section, based on the literature and our own investigations, we characterize the structure of the working population and the labor requirements in Indian agriculture, as well as the way they have evolved over time driven by the agrarian reforms and the agricultural Green and White Revolutions.

The size of farms clearly plays a role in structuring labor requirements. In India, with an average size of 1.08 ha in 2015, farms are extremely small in size (Figure 2). This is a well-known feature of Indian agriculture, which is reinforced over time as a result of inheritance and land division between generations. The other key characteristic of the Indian land structure, which is less often highlighted, is a result of the unequal distribution of land amongst agricultural workers. While the agrarian reforms introduced in the decades following India's independence undoubtedly allowed farmers and former tenants to formally own private property, and ended the taxation system that existed earlier, there was no large-scale distribution of land, and numerous families, mostly from lower castes, remained landless. Cultivated land is hence very unequally distributed among cultivators (Figure 2), alongside a large contingent of landless farming households accounting for about 40% of India's rural households (Rawal, 2008, based on NSS data). In the 2011 Census, for the first time, out of a total of 263 million farmers (including “marginal” workers), the number of agricultural laborers (144 million) surpassed the number of cultivators (119 million) (Dorin and Aubron, 2016: Figure 3).

The Green wheat and rice Revolution, boosted by proactive public policies since the mid-1960s (Dorin and Landy, 2009), had a profound effect on agricultural work. Irrigated areas have more than doubled since 1970, mainly thanks to groundwater lifted with diesel or electric pumps. This irrigation, combined with short-cycle seed varieties, made it possible to increase crop intensity (number of crop cycles per year on the same plot) from one to two or even three. The average was about 140% in 2017 India-wide (Figure 4). The multiplication of crop cycles combined with the increase in yields owing to subsidized inputs (especially chemical fertilizers and energy to pump groundwater) has increased production and the time dedicated to harvest and post-harvest activities which remained little mechanized (see Konduru et al., 2013 for cotton). Indeed, since the 1990s, in irrigated zones the tractor has become a common sight on the largest farms—which rent its services to neighboring farms—(Gulati and Juneja, 2020), but its usage is often limited to soil preparation and transportation operations. Moreover, on certain irrigated farms, cereals, sugarcane or even cotton were replaced in the 2000s by market gardening. The association and rapid succession of several vegetables crops on a same plot make these new cropping systems even more labor intensive in terms of manual work. The Green Revolution thus represented a vast process of labor intensification, alongside the high intensification in inputs (lab-seeds, chemical fertilizers, pesticides…) and irrigation equipment. According to the data we collected in three regions (Table 1), the number of workdays required per hectare per year to cultivate irrigated crops is between 176 and 940, as compared to only 112 to 269 for rainfed crops.

The role of livestock and the work associated with it have also been greatly transformed over the last decades. In irrigated zones, livestock has to face the challenge of reduced spaces and grazing periods as a result of the development of irrigated crops and the replacement of its two historical roles (draught and manure) by equipment or inputs (Aubron et al., 2019). This situation has led certain irrigated farms to abandon livestock and to specialize in crop production. On the contrary, smaller farms, or farms that did not have access to water, maintained or even developed their livestock activity to take advantage of its traditional roles (draught and manure) as well as to produce milk, for which a market now exists, particularly via the dairy cooperatives created by the Indian White Revolution from the 1970s onwards (Dorin and Landy, 2009; Scholten, 2010). In some regions, this dairy farming even involves agricultural laborers with no access to cultivated land, who thus develop landless livestock farms. Estimates vary greatly, but between 40 and 90 million families were involved in dairy farming in India in the early 2010s (30% of Indian households according to Dorin et al., 2019), with an average herd of 3 cows or buffaloes, including 1 to 1.3 milk-producing animals (Government of India, 2012; Dorin et al., 2019). 70% of India's cows and female buffaloes are bred on farms with access to cultivated land of less than one hectare (NSSO, 2006). The work time for breeding activities depends on the availability of fodder resources. A larger amount of time is required when farms have limited access to land and water, i.e., to straws and cultivated fodder. This makes them largely dependent on natural vegetation, which involves time-consuming gathering and grazing activities, even for small herds (Aubron et al., 2019).

In the 13 study regions, the use of hired workers is noteworthy (see Table 2 for five small regions). The vast majority of farms employ workers, most often day laborers from lower castes, to carry out a part, or even over half the agricultural work, alongside the family. Production systems where all the work is carried out by family labor are actually quite rare, and while a few cases correspond to farms that function solely on the basis of hired labor, with the family making no contribution in terms of work, such cases are even rarer in our sample.

The results obtained in all the study regions reveal the following trends:

The households surveyed across the country by the IHDS in 2011–12 (IHDS, 2011), largely confirm these results. Among the 16,409 households shown in Figure 5A, we note that, with a few exceptions, the use of hired labor remains high, even on farms smaller than 1 ha. The general (unweighted) average is 30 days/year, or 32 days/ha (for 1.15 ha/household on average). However, on holdings smaller than 0.4 ha, this average falls to 7 days/year (37 days/ha for 0.19 ha/ household on average), while with 5 ha or more, it rises to 182 days/year (<21 days/ha for 9.27 ha/household on average). Hence the more the farm size increases, the more employees there are per farm. However, we also note that the intensity of wage labor per hectare decreases when the size of the farm increases (Figure 5B) which can be explained by less labor-intensive crop choices or a higher rate of mechanization in large farms.

The Indian countryside is hence filled with lessons on how agricultural wage labor can help both identify and explain farming models. Numerous Indian farms are clearly family farms. They correspond to families who invest their labor to earn an income, but wage labor is nonetheless present. This wage labor ensure that the family workforce is fully occupied and paid for its work during the year, either by hiring out its labor during fallow periods in the calendar when the means of production are lacking (land, animals, irrigation), or by hiring in laborers when family labor is insufficient during peak periods.

But there are also numerous patronal farms in India. As they have greater means available to them, particularly good access to irrigation, they can implement cropping systems that are more productive, despite being more labor intensive. They hire laborers not only to cope with work peaks, but also to generate more income for the family workers. While they are clearly different to family farms, these patronal farms have little in common with “corporate farms”. In the latter case, the investors look for a profit from their capital, and not payment for work they do not provide.

We have thus shown that Indian agriculture cannot be reduced to a homogenous series of family farms. The latter differ in terms of surface area, but also greatly in the role they attribute to hired labor. These conclusions are in line with those of several authors (Bélières et al., 2014; Garner and de la Campos, 2014) who stress that farm size is not a sufficient criterion for differentiating family farms from other farming models and like us, they make labor and the use of hired labor central to the analysis. Our technical and economic approach to agricultural labor in India, however, goes further and make the logics at work in the different farming models more explicit.

By characterizing the structure and technical functioning of the farms and the resulting demand for labor over the course of the year, we better understood the place of wage labor in family farms. Hiring of laborers during peak periods, highlighted with our work calendars, is a common practice in family farms in other parts of the world (see for example Zhang, 2015 for China). The same is true of the sale of labor to other farms by family farmers lacking the resources (land especially) to meet their subsistence needs on their own farms, as described for example in Rwanda (Petit and Rizzo, 2015).

Our economic assessment (labor productivity to be compared with daily wages, family gross margin, agricultural income) then clarified the contours of the farming models. It leads to a definition of patronal farming—a farm model in which the use of wage labor makes it possible to increase the earnings of family workers—that clearly distinguishes patronal farming from both family and corporate farming. Family and corporate farms may or do employ wage workers, but with a different rationale from that of patronal farming. This is a step forward from the literature that often presents patronal farm as an intermediate category (Bélières et al., 2014; Marzin et al., 2015; Sourisseau, 2018). This article proposes to consider it as a separate category, with its own logic regarding the hiring of wage workers: neither to cope with work peaks, nor to draw a profit from their capital, but to increase the earnings of family workers.

This Indian case-study also provides food for thought on the links between farming models on the one hand and the national development path and employment on the other. The structural importance of patronal farms in India may be a result of its caste system (where physical work is degrading for many upper castes, Bardhan, 1982), but also of the country's history. India experienced an incomplete agrarian reform, which left a large part of the rural workforce landless, and a Green Revolution, which beyond the intensive use of “modern agricultural inputs” was largely based on an intensification of the use of labor, partly provided by hired workers. The working conditions of laborers are clearly better than in the past, when various forms of bonded labor still prevailed (Prakash, 1990; Breman, 2007), but inequalities remain immense and laborers' incomes very low, with a tendency to deteriorate in recent decades (Jha, 2015).

While patronal farming has thus not allowed a large part of the rural Indian population to shift out of poverty, it does employ people and has made it possible to retain a large proportion of the labor force in the countryside. Maintaining or providing employment in agriculture (still 45% of the working population in India according to NSO, 2020), even if many of these jobs are poorly paid, is all the more valuable as, despite an average growth of 7% for 20 years, India's economy is marked by secondary and tertiary sectors that generate few jobs (Dorin, 2021). In such a context, the Indian Union developed a coherent agricultural policy through the Green Revolution, by stabilizing agricultural costs and prices for patronal producers, while subsidizing food through the Public Distribution System (PDS) for the poorest (Dorin et al., 2004). As the latter are often employed on patronal farms, the PDS has also made it possible to keep agricultural salaries low. The White Revolution reinforced this edifice, by giving smallholders, and even landless households, access to a market and thus to a supplementary income through milk (Aubron et al., 2019). This buffering capacity of the Indian agricultural model in terms of employment was highlighted during the COVID-19 crisis, when several millions of workers joined the agricultural sector in the third quarter of 2020 following the lockdown (Dang et al., 2021). Even if these jobs are low-paid—leading the observers to speak of disguised employment—the absorption of such a large flow of workers is notable and would be unthinkable in many other countries. India thus illustrates the capacity of public policies to structure and stabilize the farming models that exist in the country and to use them to orientate not only production but also employment (Bélières et al., 2014; Marzin et al., 2015).

The recent history of Indian agriculture nonetheless questions the sustainability of this patronal farming model. Rural salaries are a key issue today, and changes, however timid, seem to be taking place in agricultural production systems. The Mahatma Gandhi National Rural Employment Guarantee Act (MNREGA), adopted in August 2005, follows on from numerous programmes that since 1970, have sought to create employment for the poor by developing rural infrastructure. But for the first time, this was a legally binding law: the State commits to providing at least 100 days of paid employment to every rural household engaged in unqualified manual labor. The candidates are entitled to unemployment benefits if the local administration does not provide them work within 15 days of their requesting it. This particularly restrictive context seems to have borne fruit, and may even have led to raising salaries (Das and Usami, 2017; Berg et al., 2018; Misra, 2019).

The progress of salaries is probably not unrelated to the more widespread use of motorized mechanization (Bhargava, 2013) which replaces manual work or draught power and thus reduces the workload. In the regions studied here, apart from irrigation equipment, mechanization has long remained marginal. Tractors nonetheless appeared in the countryside in the 1990s, and mechanized cereal harvesting is increasingly common in certain States such as Punjab. Less pronounced, but more problematic, is the growing use of chemical herbicides. The Indian consumption level had remained exceptionally low until present (unlike chemical pesticides). Manual weeding was done by laborers, for whom it was a precious source of fodder. However, over recent years there has been a clear increase in the use of herbicides (Gupta et al., 2017). Soil preparation that is increasingly mechanized is not a very time-consuming activity, but this is not true of weeding and harvesting, which mobilize numerous workers who could gradually be replaced by equipment and synthetic inputs.

In parallel, in several Indian states, land regulations have recently been modified to facilitate the introduction of large corporate holdings and numerous authors question the effects of such measures on rural employment (e.g., Singh, 2006; Swain et al., 2012). More recently, three Farm Bills challenged the public procurement of cereals system and price guarantees for producers in the concerned regions. The latter, often owners of patronal farms where motorized mechanization is underway, in fact strongly contested these bills during protests in Delhi and elsewhere (Jodhka, 2021; Kumar, 2021). These protests lasted more than a year until the laws were repealed in December 2021. The question which then arises is whether the Indian patronal farming model, which provides employment, but is not capable of reducing poverty, is compatible with further liberalization of the Indian economy. If this is not the case and these transformations continue, the type of farming practiced in the Indian subcontinent could change, along with its economic development model.