Connectivity and Savings Propensity among Odisha Tribals

Rural areas in India are underdeveloped relative to urban areas. One result of this has been a migration to urban areas; such migration, in turn, sharpens the problem of sustainability of rural communities. Even though there are some social benefits to urbanization, there are also disadvantages; and if urban migration is rapid and unplanned, social services in urban areas are likely to be strained. Given this, it is important to pursue strategies to develop rural areas. Such solutions can either involve external intervention with outside resources or alternatively, development using internal resources. The second solution is clearly more sustainable, as well as politically more feasible. An important part of such a self-reliant strategy involves rural saving. This paper uses the results of a survey to examine the factors affecting saving in a rural part of Odisha populated primarily by tribals. Our tentative findings are that savings propensity is determined partly by the extent to which individuals feel connected to the broader economy, and partly by cultural factors. One implication of these findings is that connecting rural areas to other, possibly urban, locations could elicit greater saving and this could lead to greater development, employment possibilities, economic betterment and all the consequent social welfare implications.


Introduction
Market access and connectivity are very important, particularly in developing economies. If a village lacks market access, then it will arguably be less able to develop its resource endowment. As a result, inhabitants may end up remaining in poverty, and consequently lack basic necessities such as water, electricity and schools. This may also spur emigration to locales that are better connected and where they can develop their human potential better. On the other hand, access to markets and connectivity to the broader economy may well allow local resources to be better utilized in accordance with demands for local products in the larger marketplace. This may spur changes in livelihood patterns, and also be a catalyst for greater savings since real rates of return on investment will go up and the value of savings will increase.
Many current development projects focus on providing information to farmers in areas with high incidences of poverty. For example, Harsha Trust's approach to poverty alleviation (as described on their website (http://harshatrust.org/strategy/) is to build human capital, provide suitable technology and investment, train the community in new skill sets and finally to link them to markets. Another example in the arena of financial services is the Indian Prime Minister's Jan Dhan Yojana, which focuses on providing access to financial services. What is common to these approaches is an interventionist mindset, though they are not both equally so (Harsha Trust could be considered more interventionist than the Jan Dhan Yojana). While not denying the need for such primary interventions, a second approach to development would focus on auxiliary interventions, viz. providing basic infrastructure, such as roads. The contention of this approach is that once they have roads and concomitantly, access to markets, residents will discover which livelihoods are valuable to them and make their own choices.
At the same time, access to markets will also make larger scale investments more viable, since production need no more be for the local micro-market (e.g. Malkangiri, the south Odisha location studied in this paper), but rather for the larger macro-market, consisting either of nearby large towns or perhaps even the larger national or global market, since access to a large local market (such as Jagdalpur or Bhubaneshwar that are farther away from Malkangiri) will in turn provide connectedness to larger selling arenas. Once larger investments make more sense, there will be a concomitant greater need for capital, ahead of production, leading to borrowing; more frequent cashflows, leading to savings and banking needs; and finally, a need to manage risk, leading to use of various insurance products.
Since roads are a public good, it is more difficult for residents of these regions to get together to build the roads themselves. Thus, intervention in the form of government road construction is necessary. Once connectivity is provided, though, markets will be set up by interested parties, or existing broader area markets will be utilized by local rural residents, who previously Electronic copy available at: https://ssrn.com/abstract=3556583 had no access to these markets. 2 This is an argument that starts with government investment in connectivity and results in greater economic development. This paper seeks to discover the extent to which such a narrative is supported by the data on savings and financial service use. Broadly speaking, the hypothesis that is sought to be tested is that regions with greater connectedness to markets are likely to be more involved in production for the market, and hence they are more likely to use financial services -savings, credit and insurance. The contention is not that direct intervention in terms of providing education, livelihood skills and changing of traditional mindsets is unnecessary, but that these perhaps need to be preceded by or, at the very least, accompanied by increased market access. 3,4 Most work on savings behavior at the micro-level has focused on areas that are relatively wellconnected to the larger economy. Savings behavior in such areas may be qualitatively different from that in less developed areas. As Polanyi (1957) has shown, the economy is always embedded in the larger society; as such, it is unwarranted to make inferences from studies that look at savings behavior in culturally different locations for the purpose of determining public policy for tribal areas. With this in view, we focus on savings behavior in economically backward areas that are arguably different in a cultural sense from the broader society. Specifically, this paper looks at of how savings propensity is determined in tribal areas in Malkangiri district in Southern Odisha. If savings propensity is indeed greater in areas with greater connectivity, this would constitute support for government investment in connectivity. 5

Literature Survey
The two main theories of household savings behavior depend upon the life-cycle theory (Ando and Modigliani, 1963) and the permanent income theory (Friedman, 1957) respectively. The first emphasizes the life-stage of the saver, while the second emphasizes expected future income rather than current income; hence an increase in expected future income could decrease current savings. The hypothesis that we consider in this paper could be considered a test of the permanent income hypothesis to the extent that we believe that greater connectivity implies larger investment opportunity sets and higher permanent income. Theories of saving more pertinent to developing economies look at other factors, such as access 2 While schemes such as the Jan Dhan Yojana do depend somewhat more on the agency of the individuals involved, they may be putting the cart before the horse, in that the need for such accounts may not be perceived by the local population. The whole program seems to be part of a government push for transparency and computerization of banking operations, which make sense mainly in terms of the larger picture, rather than in terms of an immediate benefit for putative users of these services. 3 If there are reasons why economic agents will not be able to respond on their own to increased market connectivity, a limited primary intervention may still be warranted. Nevertheless, targets of the intervention should be provided with a connected environment where the new skills can be profitably employed. 4 Viswanath (2015) argues for educational and skill development initiatives in the context of an evaluation of microfinance programs. to credit markets. Deaton (1992) suggests that financial savings can be inhibited due to lack of access to credit markets; instead individuals might prefer to invest in real assets (Rosenzweig and Wolpin, 1993) or rely on self-insurance particularly when macro-shocks like droughts are likely (Kazianga and Udry, 2006). Kulikov, Paabut and Staehr (2007) suggest that ownership of non-income producing assets, such as dishwashers may influence savings as well, if they are considered by their owners as wealth. Hubbard et al. (1995) suggest that state-sponsored social insurance programs (like Medicaid in the US) might be a substitute for savings; the same is likely to be true of implicit social insurance in close-knit societies (Karlan et al., 2014).
We now look at studies of savings behavior in rural South Asian environments -these are particularly useful in that they are likely to provide indications of cultural determinants of savings behavior that need to be taken into account in our study. Goedecke et al. (2016) look at savings practices in coastal and central Tamil Nadu, centering on Villupuram and Cuddalore districts. They find that caste membership is important; dalits are more likely to use gold as a savings vehicle, as opposed to land which they were traditionally not allowed to own. In contrast to some other work (Carpenter and Jensen, 2002), they do not find substitution effects between informal savings and bank savings. Cheema et al. (2018) find that in Pakistan, savings propensities are higher in rural areas and among educated, wealthier families possessing livestock. Naik (2013) looks at the question of savings in Sundergarh district in north-western Odisha. Although this area is also quite backward, most of the households that were surveyed are landless laborers. As such, the results may be different from those in Malkangiri. Naik's sample was half Christian, which made a big difference; this, too, suggests that the behavior of her sample is probably quite different. Gedela (2012) finds that savings increases with income among rural and tribal households in Vishakhapatnam district. Savings increases with age of household head, but at a lower rate. He also found that dependency ratio affected the amount of savings (the higher the dependency ratio, the lower the savings). He also found that male households save more than female-headed households (adjusted for income, age of head of household, dependency ratio, et cetera).
We find that connectivity is indeed related to savings propensity. As noted above, this can be viewed as support for Friedman's permanent income hypothesis to the extent that greater connectivity implies larger investment opportunity sets and higher permanent income. Our findings that savings propensity is related to TV and mobile phone ownership is contrary to the findings of Kulikov et al. (2007), but consistent with the observations of Goedecke et al. (2016) that economic behavior is affected by cultural norms and beliefs. Cultural norms may also be explanation for higher savings propensities amongst families that rely exclusively on agricultural revenue. We also see that households with older heads also save more, just like Gedela (2012) and Carpenter and Jensen (2002). Our findings regarding the relationship between food purchasing behavior and savings behavior is an original contribution to the literature. In particular, we find that households that save more tend to purchase meat less frequently (though not sugar, milk and fruit). We interpret this to mean that meat is seen as a luxury and dispensable. Alternatively, it may reflect conservatism in an area that is traditionally poor and less given to meat consumption for reasons of poverty.

Sample selection and data collection
Rationale: An important reason for our choice of location of Malkangiri for our study is that it is one of the least developed districts in Odisha. Malkangiri is a border district of Odisha and touches Andhra Pradesh on its south and Chhatisgarh on its west. It has a very high proportion of scheduled castes (57.85% of the population) and scheduled tribes (22.77%) as per the 2011 census. As the Odisha District Gazetteer for Malkangiri notes, "the geography of Malkangiri district is marked by different hill terrains, far-flung cut off areas and dense forest," with more than 90% of the population living in rural areas. According to the last census, the literacy rate was below 50%. According to the 2013 State of the Forest Survey Report, 40% of the total geographical area of the district is forested, though given the prevailing rate of deforestation, the present proportion is likely to be lower (Pattanaik, Reddy and Reddy, 2011). According to the District Gazetteer, 57.8% of the population is tribal (2011 census), with 97.8% of these living in rural areas. Our sampling strategy, as outlined below, attempts to select respondents from locations that provide sufficient variation in terms of connectivity, but also rules out major urban areas (such as block headquarters). This results in a greater proportion of tribals than for Malkangiri district, overall.
Method: A master list of villages for Malkangiri district in the state of Odisha was obtained in March 2017 from an Indian Government website. 6 The names of 996 distinct villages were obtained through this procedure. According to this site, are were seven distinct blocks -Kalimela, Khairput, Korkunda, Kudumulugumma, Malkangiri, Mathili and Podia, in which there were 108 different gram panchayats. 7 We first chose 36 villages which were sampled from the 996 distinct villages. The 36 villages were chosen in the following manner. Malkangiri has seven blocks, consisting of 108 Gram Panchayats (Village Governing Zones; GPs). Field workers from an NGO, WASSAN (Watershed Support Services and Actitivities Network), familiar with Malkangiri district were asked to assign each GP to one of seven baskets based on the following subjectively understood criteria: Gram Panchayats are not close to markets, they are far from block HQs, roads are bad (60% damaged) and distant from NHs. 6 Gram Panchayats are far from the district HQ and far from NHs, the roads are in bad condition, areas are hilly, but it is possible to travel on foot, although these roads are not passable in the rainy season; locations are far from markets. 7 Gram Panchayats have no roads, any rivers that exist are non-navigable, they are far from the national highways and they are also far from large cities.
There were eleven different field workers, who engaged in this exercise. Through open discussion, agreement was reached on the assigning of Gram Panchayats to seven blocks; the characteristics of the different baskets are provided in Table 1. This procedure yielded 11 Gram Panchayats in basket 1, the most connected; 31 in basket 2; 27 in basket 3; 15 in basket 4; 8 in basket 5; 10 in basket 6; and 6 in basket 7. The field workers were asked to describe Gram Panchayats in each basket and a rubric was generated using this description.
Subsequently, the same workers were asked to classify Gram Panchayats as either too dangerous for data collection due to Naxalite activities or not dangerous. All the Gram Panchayats falling in basket number 7, seven Gram Panchayats in basket 6 and three Gram Panchayats in basket 5 were classified as dangerous, for a total of sixteen Gram Panchayats. We also decided not to include gram panchayats where block headquarters are located. Such areas are likely to be more urban and otherwise as well, quite different from rural areas.
Occupations that are not found in other areas may be found only in such urban/quasi-urban areas. Thus these areas are not easy to compare with other more rural areas. Six Gram Panchayats in basket 1 were excluded following this procedure. From the remaining Gram Panchayats in six baskets, three Gram Panchayats were chosen randomly without replacement from each basket, for a total of 18 Gram Panchayats. Next, two villages were chosen randomly without replacement from the villages in each of the chosen 18 Gram Panchayats, for a total of 36 villages. Unfortunately, there were difficulties in obtaining household lists and ultimately, interviews were conducted in only 13 villages by four different interviewers, based on household voter lists that were obtained by the workers. Ultimately, 39 households were interviewed by Jatishmaya Biswas; 15 households by Kanhu Charan Sahani; 8 by Subhalaxmi Das in Parkhanmalla village in Parkanamala Gram Panchayat in Kudumulugumma Block; 24 by Mamta Mahapatra for a total of 86 households (see Table 2 for details). Survey: Interviewers were provided by the WASSAN Foundation. Local personnel were needed to interview the respondents since many of the respondents were tribals who did not necessary speak even the state language, Odiya. Most of the interviews were, in fact, conducted in Odiya and Kui; interviewers were conversant in Kui and Odiya, with a reasonable knowledge of Hindi and some understanding of English. All interviewers underwent training in order to acquaint them with the meaning of the questions that they were to ask. Even though the questionnaire itself was not translated into Hindi or English, interviewers were provided, during the training, with explanations of the questions in Hindi by the author, supplemented by Odiya explanations from WASSAN personnel, who knew both Hindi and English.
The connectivity measure was obtained by asking the question "How connected do you feel your village is?" for current connectivity and "How connected do you feel your village was, three years ago?" for past connectivity. The additional information was provided to interviewers to aid them in the posing of this question:

Model and Hypotheses:
The theory, once again is the following. Access to input and output markets leads to scaled-up production and capital intensive production that makes optimal use of local resources. This, in turn, creates a demand for financial resources, both savings and credit, as well as tools to manage risk, i.e insurance. In the absence of market access, it is optimal to keep to a subsistence economy. If markets are available, then it makes sense to produce for a larger set of consumers.

Results:
Our unit of analysis is the family. While a large majority of heads of households noted agriculture as their primary occupation (86%), an important proportion (39.74%) listed Nontimber forest products INTFP) as their secondary source of income (Table 3). Of the 62 individuals who answered the question, 77.42% identified themselves as members of a scheduled tribe, 16.13% identified as members of a scheduled caste, 3.23% as members of another backward caste and a further 3.23% as "other." We have data available for various aspects of the family's financial condition, both the balance sheet and the income statement, as well as other characteristics of the family's living and working environment that are useful for our analysis. From the balance sheet, we have information about the various kinds of assets that they possess. From Table 4, we see that the propensity to save varies according to different respondent characteristics. In particular, we can see that the propensity to save is increasing in connectivity. The more the respondent feels that his/her village is connected to the broader world, the more they tend to save out of their income. Although there is a stronger correlation between the propensity to save and current perceived connectivity, compared to past connectivity, there is no significant relationship between the propensity to save and the change in perceived connectivity. -0.0815 83 tlg1lo Note: Proportion of income saved by respondent, specifically respondent's bank balance as a proportion of totinc1 (total income of all family members)  From Table 5, we see that the propensity to save is higher for those who consume meat less frequently, those who purchase vegetables less frequently, those who do not have mobile phones, and those who do not have TVs or motor vehicles. On the other hand, people who own radios do seem to save more. Finally, people who possess life, health, crop or other insurance also seem to save more as a proportion of their income. One way of understanding this is to say that this is an expression of conservatism. We note that people who are less likely to consume high-quality and high-priced goods and those who purchase insurance are more likely to save, while those who save tend to purchase insurance. Both these characteristics can be reasonably identified with conservatism.

+
On the other hand, it we believe that better connected villages would be less conservative, then the conservatism hypothesis fails because better connected villages actually have greater savings propensity. Hence the conservatism that we identify here is not a cultural conservatism, but rather an economic conservatism. Table 6 shows that those who pursue farming as an occupation exclusively tend to have a greater savings propensity compared to those who are farmers primarily, but also tend to have other occupations. This may be explained as due to a greater need for precautionary savings caused by the lower level of income diversification. Alternatively, it may reflect a more conservative attitude to saving on the part of a population subgroup that is probably more conservative in its attitudes generally. It would also seem that the negative relationship between savings propensity and total income on the one hand and the positive relationship between savings propensity and perceived connectivity are independent. A multiple regression of savings propensity on total income and perceived connectivity shows the expected signs of the coefficient, but the slope coefficient on total income is no longer significant (Table 7; Model III). However, if we use revenue from rice (revric) as a measure of income, then the coefficient is now significant at the 10 percent level (Model II).
Similarly, a multiple regression of vh1ba, a dummy variable for frequency of purchase of vegetables, and on perceived connectivity shows both significant, the first at the 5% level and the second at the 10% level (Model IV). On the other hand, the measure of mobile phone ownership, which has a simple negative correlation with savings propensity, turns insignificant when included with a measure of connectivity (Model V). Thus, there is clearly evidence of the significance of connectivity as a positive influence on savings, while there is some support for an independent negative effect of vegetable/meat consumption.
It is possible to combine all of these different strands of evidence if we interpret connectivity as a belief that returns to investment are likely to be high. If so, then that would allow us to relate connectivity to a weaker propensity to consume "luxury" goods and a corresponding stronger propensity to save -and presumably invest.
We also regressed connectivity on mobile phone ownership, meat consumption and total income to see if the connectivity effect could be explained by other respondent characteristics. However, the meat consumption and mobile phone ownership coefficients in this regression are insignificant, while the total income coefficient is significant and positive. While this is not surprising, we see from Model III that it is connectivity that is a better explainer for savings propensity than total income.

Analysis of Results:
As mentioned above, our hypothesis could be considered a test of the permanent income hypothesis to the extent that greater connectivity implies higher future income. To that extent, our results could be considered support for the permanent income hypothesis.
We find a positive correlation between savings propensity and mobile phone ownership (in our regressions) on the one hand and TV ownership (in univariate analysis), on the other. This goes contrary to the hypothesis of Kulikov et al. (2007). However, radio ownership is negatively correlated with savings propensity. One way of reconciling these seemingly disparate results is to look at the implications of the mediating variables as measures of respondent expectations regarding the future. Radios are cheaper, on the one hand, and generally seen as backward relative to television sets and mobile sets. Hence a choice to buy radios may reflect the belief of the owner that future prospects are not good. On the other hand, purchases of TVs and mobile phones reflect membership in the modern world as well as investments in the future. From this point of view, these results complement our primary results regarding connectivity.
On the other hand, connectivity may be proxying for availability of savings vehicles. Hence, villages with greater connectivity may show greater savings simply because the possibility of savings vehicles exists (need to control for availability of MFIs).

Conclusion
Savings propensity is determined by at least two forces in our sample -those with a feeling greater connectivity, save more. This makes sense if connectivity is a proxy for the ability to use money better in the future. This is consistent with the finding that bigger savers also tend to be those who do not have mobile phones and television sets -all pointing to looking ahead and not consuming today.
The second possibility that savings propensity is at least partly cultural. This is backed by the finding that individuals who are farmers and are not involved in other activities (no alternate secondary occupation) tend to save more. Households with older heads also save more. These people also tend to purchase meat less frequently (though not sugar, milk and fruit). This also makes sense if meat is seen as a luxury and dispensable.