Many Rivers to Cross: Social Identity,Cognition,and Labor Mobility in Rural India

Data and Methodology

Methodology

The detection of labor market mobility

We measure income mobility in absolute and relative terms. Absolute income mobility is measured by the logged value of the difference of income between 2010 and 2016 (after controlling for inflation). Relative income mobility is detected by a variable measuring the number of percentiles of mobility (percentile rank change) that a worker experienced across the distribution of annual wages between the 2010 and 2016–2017 waves. The variable can take the values [–100; 100].

Cognitive skills and personality traits

Individual endowments, such as educational background, are known to be important determinants of labor opportunities. However, the effects of other individual differences on the labor market are much more complex to assess. Our study identifies cognitive skills using three dimensions: literacy tests (reading and writing basic sentences); numeracy tests (four basic calculation tests); and the Raven Colored Progressive Matrices (CPM), meant to capture “fluid intelligence.” ³ The Raven CPM consists of three sets of twelve questions of increasing difficulty, which are cognitive, visual, nonverbal tests that do not require any level of formal education. It captures the ability to think and make sense of complex data and logical reasoning. The Raven CPM has been previously used in economics for cognitive skills assessment in low-literacy populations in developing countries (e.g., Serneels 2008). The score of the respondent to each dimension provides a refined measure of individual cognitive abilities.

In addition to these usual dimensions of individual endowments, we use the Long Big Five Inventory, which is a taxonomy that refers to five dimensions commonly used to describe human personality traits: openness to experience, conscientiousness, extraversion, agreeableness, and emotional stability. Practically, a set of forty-two questions (seven for each dimension) has been asked to the respondent to capture the five personality traits. ⁴ We implement an alternative and gender-specific personality traits factor analysis. The universality of the Big Five Factor Model (FFM) has indeed been questioned in several economic and anthropological studies. Laajaj et al. (2019) show that their validity outside of Western, educated, industrialized, rich, and democratic (WEIRD) populations is limited because of a risk of misinterpreting the Big Five survey. Moreover, by testing the FFM among forager-farmers in the Bolivian Amazon, Gurven et al. (2012) show that the FFM is not universal. They do not find strong support for the FFM but consistency among factors relating to prosociality and industriousness. They argue that further research is needed on how lower-rank personality traits assemble into higher-order personality traits. Following the intuition of these studies, we implement our own factor analysis of the forty-two questions on male and female subsamples separately. We extract the first five factors, then proceed to a promax rotation and analyze the content of each factor (see online Appendix 8). We implement checks of internal factor validity using the Cronbach’s alpha measure and compare the results of the two approaches: (i) and (ii).

Estimating the determinants of income mobility

To analyze income mobility in an econometric setup, we use the two aforementioned income mobility variables (i.e., absolute income mobility and relative income mobility) as dependent variables. The independent variables of interest are the cognitive skills C o g _ v a r (i.e., Raven score, numeracy, and literacy) and the personality traits P e r s o _ v a r . C o n t r o l _ v a r represents a vector of sociodemographic characteristics.

I n c o m e _ M o b i l i t y = β 0 + β 1 C o g _ v a r + β 2 P e r s o _ v a r + β 3 C o n t r o l _ v a r

(1)

The estimation of equation (1) by ordinary least squares (OLS) is likely to yield biased estimates because our variables of interest were only collected in the second wave of the dataset and because of sample selection issues pertaining to the nature of our research questions, which included only active occupied individuals. To deal with these issues, we restrict our sample to individuals older than 30 years old, ⁵ ensuring the stability of cognitive skills and personality traits across the two waves (Cobb-Clark, Deborah, and Schurer 2012), and we implement a Heckman selection correction method, which is commonly used in labor market analyses. Further information on these issues and the methods used are available in online Appendix 2.

Descriptive Evidence of Labor Market Mobility

The labor market structure in our study area has experienced multiple changes between 2010 and 2016–2017. As online Appendix 3 shows, as elsewhere in Tamil Nadu and more generally in India, agricultural employment has declined drastically, especially for agricultural casual laborers. Moreover, employment outside of agriculture has simultaneously risen sharply for regular, nonqualified workers; and declined for casual workers. These evolutions suggest a trend of regularization of nonagricultural employment explained by the rapid development of the service sector in rural Tamil areas over the last decade. We also note that the Mahatma Gandhi National Rural Employment Guarantee Act (hereafter NREGA) scheme, which is a national employment program aimed at alleviating rural poverty, has become the main occupation for 11 percent of the occupied active population in 2016–2017.

Looking at income mobility reveals additional information on how the labor market evolution has reinforced caste and gender inequalities in terms of income. Table 1 shows that, overall, average annual incomes rose by 5,000 rupees (around USD [U.S. dollars] 75 in 2016) over the six-year period, but this amount masks strong heterogeneity. The first striking feature is that disadvantaged groups (i.e., women and Dalits) have, on average, lost around 700 rupees (USD 10.50) between 2010 and 2016–2017. Conversely, men and non-Dalits seem to have benefited from the labor market evolution, and we observe a significant rise in annual incomes for these social groups (up to USD 247 for upper castes).

OPEN IN VIEWER

Table 1

Absolute and Relative Income Mobility by Gender and Caste Group

Absolute Income	Observations	Mean Change in Annual Income (INR)	SD	Median Change in Annual Income (INR)
Whole sample	422	5,037	35,932	–496
Men	261	8,626	42,941	2,240
Women	161	–781	18,597	–2,553
Dalit	232	–739	27,128	–2,462
Middle castes	157	10,663	42,722	1,132
Upper castes	61	18,875	46,707	12,773
Nonmigrants	352	876	35,652	–2,215
Migrants	118	45,098	64,840	32,775
Relative Income	Observations	Mean Mobility in the Distribution (in percentiles)	SD	Median Mobility in Distribution (in percentiles)
Whole sample	422	0.6	31.2	0
Men	261	–0.2	32	0
Women	161	1.8	30	1
Dalit	232	–2.1	29.5	–2
Middle castes	157	3.2	33.1	3
Upper castes	61	7.1	32.7	10
Nonmigrants	352	–5.8	28.6	–4
Migrants	118	20.6	27.7	15.5

SOURCE: Authors’ computations of RUME (2010) and NEEMSIS (2016–2017) data.

NOTE: Nominal values of 2016 deflated income using World Bank Measure of Consumer Price Index.

In terms of relative income mobility, overall, we observe little variation, indicating that despite important occupational mobility, the earnings gap is not dramatically changing over time. Upper castes, and to a lesser extent middle castes, have not only experienced a significant increase in their earnings but also an upward mobility in the income distribution. ⁶ In brief, we do not observe significant changes at the bottom of the distribution, where Dalits and women are overrepresented, but the earnings gap is increasing between the middle and the upper tail of the distribution.

Last, we observe that migration enables important income gains in both absolute and relative terms. Half of the migrants have seen their incomes rise by more than 32,000 rupees between the two periods, translating to an upward mobility of 13 percentiles on average in the income distribution. The migrants’ employment structure explains, in part, this higher income mobility. Sixty percent of migrants have regular employment, compared to only 20 percent for nonmigrants. However, these figures need to be tempered. Only a small share of migrants’ occupations are urban qualified jobs (around 10 percent). But in a context of high uncertainty and precariousness, having access to regular incomes throughout the year is an important asset in comparison to casual and seasonal activities available in the villages.

Overall, migrants are better educated and skilled than nonmigrants (Table 2), namely, regarding all the cognitive skills (Raven, numeracy, and literacy) and some personality traits, such as emotional stability and extraversion (the differences being significant at the 1 percent level). Migrants are also more likely to have completed higher education than nonmigrants. These specific individual characteristics of our migrants’ subsample could explain why migrants have experienced better income mobility between the two periods.

OPEN IN VIEWER

Table 2

Descriptive Statistics on Cognitive and Personality Traits for Migrants

	Nonmigrants (n = 352)		Migrants (n = 118)
	Mean	SD	Mean	SD	t-Value
Personality traits
Openness	0.18	0.46	0.13	0.44	1
Conscientiousness	–0.11	0.29	–0.14	0.26	0.85
Extraversion	–0.21	0.35	–0.08	0.26	–3.7*
Agreeableness	0.13	0.31	0.03	0.28	2.9*
Emotional stability	0	0.42	0.20	0.38	–4.5*
Raven score	11.7	8.1	17.3	11.31	–5.8*
Literacy	0.49	0.49	0.65	0.48	–3.9*
Numeracy score	1.52	1.26	1.96	1.28	–3.2*
Education (%)
No education/below primary	50.6	50.1	29.7	45.9
Completed primary	21.6	41.2	21.2	41
Completed middle school	13.4	34.1	21.2	41
Completed high school	10.8	31.1	14.4	35.3
Higher secondary school	02.6	15.8	5.1	22.1
Bachelor’s	1.1	10.6	8.5	28

SOURCE: Authors’ computations of RUME (2010) and NEEMSIS (2016–2017) data.

*

p < .01.

The Determinants of Labor Market Mobility

Discussion

Using a unique panel dataset allowing us to study the role of cognitive skills and personality traits on labor market outcomes in rural India, we have identified the extent to which individual differences overcome social inequalities in the labor market. Our analysis of intragenerational mobility uses measures of cognitive and psychological differences that are scarcely used in the case of developing countries, especially India, and deepen the understanding of the dynamics of income mobility in the rural area of Tamil Nadu. This analysis highlights the existence of a plurality of mechanisms that reinforce caste and gender inequalities, which we discuss in what follows.

Importantly, this analysis has joined together disciplinary approaches. Behavioral economics, which includes recent and fruitful advances in psychology economics, has provided new evidence that cognitive and socioemotional skills are likely to have direct and indirect impacts on individual choices and outcomes in the labor markets, but it is too often disconnected from the analysis of social structures in which individuals are embedded. By contrast, structuralist approaches recognize that individuals cannot be considered outside of the social relations that make up the collective structure (Polanyi 1944). Labor markets, therefore, appear as a place of negotiation and social interaction, where complex forms of power and domination are encompassed in relations of rivalry, exclusion, and cooperation. While both approaches are often presented as incompatible, they appear to have numerous points of convergence. Most behaviorists pay attention to the role of social norms and interactions but overlook their inherent nature, while many structuralists emphasize the structural origin of cognition and emotion but disregard the extent to which they shape individual preferences and choices. We recognize here that both views are meaningful and need to be articulated, and our findings illustrate the complementarity of both approaches.

Bearing on migration, men in rural Tamil Nadu have experienced a regularization of labor and an exit out of agriculture toward construction, education, or the growing service sector, whereas women have experienced opposite trends (Djurfeldt et al. 2008). Women’s culturally mandated seclusion limits access to employment and education; therefore, women depend on employment within the village, where opportunities are scarce, mostly agriculture oriented, low paid, and available only on a daily basis. We similarly observe a reinforcement of inequalities among castes.

Individual cognitive and psychological predispositions might appear as determinants of labor market mobility. All things being equal, however, cognitive skills play a limited role in income mobility, while personality traits seem to be significantly associated with it. We do observe a significant effect of literacy for women in absolute income mobility (see, for instance, Nordman, Sarr, and Sharma [2019] for similar findings in Bangladesh). This is especially true for Dalit women, since upper caste women are more likely to be involved in self-employed occupations, requiring a certain level of economic and social capital that lower castes rarely meet.

Our results also emphasize the crucial role of emotional stability for income mobility of Dalits and non-Dalits. We also observe that agreeableness is significantly correlated to relative income mobility for Dalits, but not to absolute income mobility. Dalits are more likely to be involved in casual occupations where relationships with the employer (oftentimes from higher castes) are crucial for job continuity. Also, for men, openness to experience, emotional stability, conscientiousness, and extraversion are significantly correlated with relative and absolute income mobility. In other words, in a highly competitive and narrow labor market, those being proactive, enthusiastic, and curious are more likely to be integrated into networks facilitating better job accessibility, enabling them to move up in the income distribution. Openness to experience, extraversion, and emotional stability are also known to increase access to broader and more diverse social networks (Pollet, Roberts, and Dunbar 2011). Here, we assume and partly verify (with our network data) that social ties may constitute a transmission channel in the relationship between some dimensions of cognition and labor mobility (see online Appendix 11 and previous findings for Bangladesh in Hilger, Nordman, and Sarr 2018).

Similarly, emotional stability is one of the main determinants of mobility into better-quality jobs. For women, openness to experience is an important determinant of exiting agricultural jobs, suggesting that this type of mobility requires being able to take risks and to challenge highly constraining social norms such as seclusion. None of the cognitive skills variables allow individuals to enter better-quality jobs, which points toward the value of soft skills in enabling access to better jobs.

Migrants do not reap the benefits of their advantageous cognitive and personality traits to improve their absolute and relative incomes. Indeed, although they appear to benefit from higher income mobility, cognitive skills of migrants, namely, numeracy, are not rewarded after migration. This result suggests that the jobs found by migrants in their new location do not require such skills, as these jobs may be of poor quality. Hence, while greater cognitive attributes might constitute a trigger for deciding to move out of their origin villages (Table 2), these skills may not be properly rewarded in the type of jobs they find in the new destination (mostly brick kiln industries). One potential limitation of this migration analysis, which is shared by most migration studies in general, is that some migrants are difficult to track. This is especially true for migrants who move to cities to access specific and qualified jobs. As a consequence, the migrants of our sample are certainly not representative, which may lead to underestimating the effect of cognitive and personality traits on income mobility.

Finally, our study provides suggestive evidence that in an effort to be universal, the FFM may blur the complexity of psychological traits of specific groups, leading to unclear evidence concerning their effects on labor mobility. Indeed, the universality of the Big Five may not be appropriate in the context of a rigid society where the role of men and women are strongly compartmentalized and socially constructed. We attempt to provide an alternative personality taxonomy suited to the rural Indian context by identifying new gender-specific factors. These factors provide interesting results for women and have a higher internal validity than the FFM (as suggested by higher Cronbach’s alphas presented in online Appendix 8.3). In line with previous studies (for instance, Laajaj et al. 2019), our results suggest that more research is needed on the relevance of FFM factors for analyzing the personality of non-WEIRD populations, especially with a gender dimension.

Conclusion

Our rural India case study shows that personality traits are important determinants of labor market mobility but also emphasizes a strong rigidity of the socioeconomic structure of the Indian labor market in terms of gender and caste, and its relative stillness over time. While for women, literacy, emotional stability, and openness to new experience appear to allow income gains, these benefits are limited by the labor market structure, maintaining them in low-skilled and casual occupations. This result also holds for individual rural migrants who have moved out of their origin villages for better job prospects, but mostly then access poor-quality jobs, with no clear reward for their specific skills. For Dalits, emotional stability and agreeableness seem to play an important role in relative income mobility. These interesting findings highlight the segmented nature of the Indian labor market, which is still strongly organized by diverse forms of domination. As shown in previous research, the caste system adapts and rearranges (Harriss-White 2003), mitigating the impact of any type of structural change to equalize livelihoods of individuals. Our article calls for further research to understand how personality traits are acquired and shaped and how they can be leveraged to allow disadvantaged groups to access better jobs and higher incomes. Our results also call for further exploration of the nature and the variety of socioeconomic barriers facing disadvantaged groups in rural India. Indeed, even if individuals have the required personality traits that enable an upward mobility pattern, having these traits might not be sufficient to transcend caste and gender labor segregation because of other factors, such as wage discrimination and nepotism.

Supplemental Material