VoxEU Column Education Gender

How pension plans changed kinship traditions

Families’ attitudes towards educational investment and lifetime saving are underpinned by longstanding cultural attitudes that must be considered in policy design. This column shows that in Indonesia and Ghana – two culturally distinct societies – families historically invested in the education of those children who would look after parents in old age. The level of this investment declined after the introduction of pensions in both countries.

Cultural traits affect a diverse and important array of human behaviours, including fertility decisions, female labour force participation, and educational investment (Fernández 2011, World Bank 2015, Collier 2017, Ashraf et al. forthcoming). Though cultural traits are ‘sticky’ and often persist across generations, they may also evolve in response to the environments in which individuals live (Boyd and Richerson 1988). Large-scale development policies change these environments and may in turn change cultural practices. Yet, there is little evidence on the link between policies and cultural change. In a new paper, to provide evidence on this link, I examine the effects of large-scale social policies on cultural practices in two low-income countries (Bau 2019).

The impact of a government pension plan on matrilocality and investment in girls’ education in Indonesia

In the paper, I examine whether the introduction of government pension plans changed cultural practices in Indonesia and Ghana. I study a set of cultural traditions that determine whether daughters (matrilocal), sons (patrilocal), or neither gender (neolocal) continue to live with parents after marriage and care for them in their old age. In matrilocal ethnic groups, daughters stay with parents, while in patrilocal ethnic groups, sons stay with them. Traditionally, both practices are widespread, and therefore potentially important determinants of economic behaviour. In anthropological data on 1,265 ethnic groups around the world, 69% are traditionally patrilocal and 16% are traditionally matrilocal (Morduck 1967). Studying these practices provides me with an unusual opportunity to measure how cultural practices change in response to new policies. This is because, unlike the practice of many other customs, whether households practice matrilocality or patrilocality can be directly observed in most census or survey data.

I hypothesise that when households belong to matrilocal ethnic groups, parents are relatively more likely to invest in their daughters’ education. This is because they will share in the returns from that educational investment when their daughter supports them in their old age. Similarly, parents from patrilocal ethnic groups will be relatively more likely to invest in a son’s education. I further hypothesise that when governments introduce pension policies, parents – who no longer need as much old-age support – will be less likely to transmit matrilocal and patrilocal traditions to their children. As a result, both the educational investment incentivised by these traditions and the practice of these traditions themselves will decline. So, the new social policies that often accompany economic growth may lead to the decline of traditional cultural practices. 

My main analyses use census data from Indonesia. Indonesia provides an unusual laboratory for studying the effects of pension policies on matrilocal practices. First, within Indonesia, there is variation across ethnicities in terms of which gender traditionally lives with parents after marriage and whether children live with parents at all. To measure this variation in the data, I match the census data to data gathered by anthropologists on ethnic groups’ cultural traditions (Morduck 1967). This variation allows me to compare the outcomes of traditionally matrilocal and non-matrilocal ethnic groups. Second, Indonesia introduced a large-scale expansion to its pension system in 1977. This expansion generates variation across birth cohorts in terms of how old a child was when the plan was introduced. Third, there is geographic variation in the location of pension offices at beginning of the programme. Since evasion of the plan was initially high (Muliati 2013), locations with offices are likely to have better enforcement. Indeed, today, more retirees receive pension benefits in places that had more offices in the early 1980s. These three pieces of variation allow me to measure the differential effects of the pension plan on females from traditionally matrilocal ethnic groups who were born late enough to be exposed to the plan and in areas where the intensity of treatment was greater. My analysis allows me to separately account for ethnicity-level effects, time trends, and geographic differences between individuals.

The data from Indonesia are consistent with the paper’s key hypotheses. In contemporary data, daughters in traditionally matrilocal ethnic groups receive more educational investment (relative to their brothers) than daughters in non-matrilocal groups. In total, this difference leads to 0.22-0.36 more years of schooling for traditionally matrilocal females. Traditionally matrilocal women who were more exposed to the pension plan (based on both geographic and cohort variation) received relatively less education and were relatively less likely to practice matrilocality as adults. The institution of the plan also decreased the likelihood of being observed practicing matrilocality after marriage. 

Symmetrical findings in Ghana

To establish whether the Indonesia results apply in another, very different, context, I then turn to Ghana. Like Indonesia, Ghana instituted a pension plan in the 1970s. However, Ghana and Indonesia have very different cultural and religious make-ups. Indonesia is predominantly Muslim, and Ghana is predominantly Christian. While in Indonesia I compare traditionally matrilocal and non-matrilocal females, in Ghana the variation is between traditionally patrilocal and non-patrilocal males. Therefore, studying different genders and cultural traits in these very different countries provides further evidence that the results in Indonesia are not driven by an unobserved variable correlated specifically with matrilocality.

In Ghana, sons in traditionally patrilocal ethnic groups receive more educational investment (relative to their sisters) than sons in non-patrilocal groups, resulting in approximately 0.18 more years of schooling. The timing of the pension plan, which was introduced in 1972, allows me to evaluate whether traditionally patrilocal males who were more exposed to the plan as children behave differently. Indeed, these males were less likely to complete primary school and are less likely to practice patrilocality as adults. 

Altogether, the results from Indonesia and Ghana are symmetric. Matrilocality incentivises educational investment in females, while patrilocality incentivises it in males. The introduction of pension plans crowds out these educational investments and reduces traditional practices. So, though culture is often persistent, the introduction of new laws and policies can cause cultural change. 

While it may seem surprising that cultural traditions that have persisted over hundreds of years can suddenly change, modern savings and pension plans change the economic environment in new and unprecedented ways. Until modern times, parents had few ways to save for old-age besides investing in their children. Thus, as low-income countries grow and adopt new social policies, these new policies may also lead cultural practices to evolve. 

The results of my paper also emphasise the importance of taking culture into account when designing policies. By de-coupling old age support from parents’ human capital investments, the pension plans had the unintended consequence of reducing education. Understanding the costs and benefits of different policies requires understanding the cultural context.


Ashraf, N, N Bau, N Nunn, and A Voena (forthcoming), “Bride Price and Female Education”, Journal of Political Economy.

Bau, N (2019), “Can Policy Change Culture? Government Pension Plans and Traditional Kinship Practices”, CEPR Discussion Paper no. 13486.

Boyd, R, and P Richerson (1988), Culture and the Evolutionary Process, University of Chicago Press.

Collier, P (2017), “Culture, Politics and Economic Development”, Annual Review of Economics, 20, 111-125.

Fernández, R (2011), “Does Culture Matter?”, Handbook of Social Economics, vol. 1, 481-510.

Murdock, G P (1967), Ethnographic Atlas, University of Pittsburgh Press.

Muliati, I (2013), “Pension Reform Experience in Indonesia”, prepared for IMF conference “Designing Sustainable and Equitable Pension Systems in Asia in the Post Crisis World”.

World Bank (2015), World Development Report 2015: Mind, Society, and Behavior,  World Bank Publications.

2,730 Reads