When a government implements a policy, there is often a question about how it will interact and/or displace existing informal practices. For example, awhile back there was a lot of discussion around how government provided insurance would displace (or not) informal risk sharing arrangements that may have been doing a good job of protecting some people from risk.
But it’s hard to address this kind of question deeply in a randomized control trial. A lot of the informal arrangements may be grounded in customary practices that are spread across a wide geographic area and they could move in a general equilibrium sense that the average RCT would miss. Into this gap comes an interesting new paper by Natalie Bau.
Bau is looking at the introductions of formal pensions in developing countries. And she finds that they shift behavior in some surprising ways – changing how parents treat their kids that they might expect to provide support for them in their dotage. She does this in a neat way, utilizing variation in customary practices in Indonesia and Ghana as well as census data and administrative information on the roll-out of pensions. Let’s take a look at some of her findings (and a bit on how she gets there).
One of the things about Indonesia and Ghana is that they have folks who practice both matrilocality and patrilocality. Matrilocality is where you go and live with your wife’s parents – often even in the same house. And yes, patrilocality is where you go and live with your husband’s parents (the paper has a nice short discussion of the literature on why these practices may have evolved).
So to start off with, Bau shows us that in the censuses (and in the more detailed data from the Indonesia Family Life Survey) that if you are from an ethnicity that customarily practices matrilocality (patrilocality) you are significantly more likely to have daughters (sons) living with their parents. And, by implication (with a couple of cites) if you are living with your parents, you are providing more support.
Bau then takes us across a whole bunch of countries to look at how these practices correlate with gender outcomes. The percent of the population that is of an ethnicity with patrilocal practices is significantly correlated with a bigger gender gap in economic and education outcomes (results for the percent matrilocal are similar for education but not significant). So, the potential logic here is that if your parents know you are going to end up living with them (and supporting them – not living in their basement and eating their food), they’ll invest more in your education.
Diving deeper, Bau looks at these relationships within country – looking at siblings within the same households in Ghana and Indonesia. And yes, after controlling for a whole mess of things, if you are a girl in a household from an ethnic group that practices matrilocality in Indonesia or a boy from a patrilocal ethnic group in Ghana you have more schooling than your siblings of the opposite sex.
Bau builds a model giving us some predictions on how the introduction of a mandatory pension program could shift things. One of the key points is that not only does the pension program take away income in the working years (making it more expensive to invest more in that kid who is supposed to live with you later) but the income coming back in retirement means you don’t need them anymore.
Now Bau gets even more creative as she shows us how the model plays out in Indonesia and Ghana. She uses (unrelated) geographic variation in the roll out of a big national pension program in Indonesia, combined with 2010 census data on ethnicity and different cohorts of ages to run a triple-difference regression. On average, fully treated women experience a 6.7 percentage point (7.6 percent) drop in the likelihood of completing primary school, a 3.3 percentage point (10 percent) drop in secondary, and a 1.1 percentage point (20 percent) drop in attending university. And they are 2.1 percentage points (27 percent) less likely to live with their parents.
Bau then moves over to Ghana (observing that “it is unlikely that the same omitted factor would drive the results across these two countries”) and looks at the impacts of their national social security/pension scheme. Here there is no data on geographic variation so she uses a double-difference approach (except for education where she can use gender for a triple diff). The effects are in line with Indonesia: men who were born to patrilocally inclined parents exposed to the pension program are 3 percentage points less likely to complete primary school and 1.7 percentage points less likely to live with their parents.
The triple-difference in Indonesia and, to a slightly less extent, the double-difference in Ghana are already controlling for a bunch of things (e.g. there are ethnicity by birth year fixed effects to control for overall trends in matrilocality in Indonesia). But she goes the extra mile and shows us robustness through a whole slew of approaches within the existing regressions and also in showing us that other explanations (matri/patrilocality being associated with greater returns to education, the potential role of other gender biased behaviors and whether remittances could be happening) aren’t driving the story.
All in all, this is a neat paper. It gives us a clear picture of how well established and undisputed government policy choices can have unintended and wide-reaching effects. And it shows us how existing data and some creative econometric approaches can help tell this kind of story.
But it’s hard to address this kind of question deeply in a randomized control trial. A lot of the informal arrangements may be grounded in customary practices that are spread across a wide geographic area and they could move in a general equilibrium sense that the average RCT would miss. Into this gap comes an interesting new paper by Natalie Bau.
Bau is looking at the introductions of formal pensions in developing countries. And she finds that they shift behavior in some surprising ways – changing how parents treat their kids that they might expect to provide support for them in their dotage. She does this in a neat way, utilizing variation in customary practices in Indonesia and Ghana as well as census data and administrative information on the roll-out of pensions. Let’s take a look at some of her findings (and a bit on how she gets there).
One of the things about Indonesia and Ghana is that they have folks who practice both matrilocality and patrilocality. Matrilocality is where you go and live with your wife’s parents – often even in the same house. And yes, patrilocality is where you go and live with your husband’s parents (the paper has a nice short discussion of the literature on why these practices may have evolved).
So to start off with, Bau shows us that in the censuses (and in the more detailed data from the Indonesia Family Life Survey) that if you are from an ethnicity that customarily practices matrilocality (patrilocality) you are significantly more likely to have daughters (sons) living with their parents. And, by implication (with a couple of cites) if you are living with your parents, you are providing more support.
Bau then takes us across a whole bunch of countries to look at how these practices correlate with gender outcomes. The percent of the population that is of an ethnicity with patrilocal practices is significantly correlated with a bigger gender gap in economic and education outcomes (results for the percent matrilocal are similar for education but not significant). So, the potential logic here is that if your parents know you are going to end up living with them (and supporting them – not living in their basement and eating their food), they’ll invest more in your education.
Diving deeper, Bau looks at these relationships within country – looking at siblings within the same households in Ghana and Indonesia. And yes, after controlling for a whole mess of things, if you are a girl in a household from an ethnic group that practices matrilocality in Indonesia or a boy from a patrilocal ethnic group in Ghana you have more schooling than your siblings of the opposite sex.
Bau builds a model giving us some predictions on how the introduction of a mandatory pension program could shift things. One of the key points is that not only does the pension program take away income in the working years (making it more expensive to invest more in that kid who is supposed to live with you later) but the income coming back in retirement means you don’t need them anymore.
Now Bau gets even more creative as she shows us how the model plays out in Indonesia and Ghana. She uses (unrelated) geographic variation in the roll out of a big national pension program in Indonesia, combined with 2010 census data on ethnicity and different cohorts of ages to run a triple-difference regression. On average, fully treated women experience a 6.7 percentage point (7.6 percent) drop in the likelihood of completing primary school, a 3.3 percentage point (10 percent) drop in secondary, and a 1.1 percentage point (20 percent) drop in attending university. And they are 2.1 percentage points (27 percent) less likely to live with their parents.
Bau then moves over to Ghana (observing that “it is unlikely that the same omitted factor would drive the results across these two countries”) and looks at the impacts of their national social security/pension scheme. Here there is no data on geographic variation so she uses a double-difference approach (except for education where she can use gender for a triple diff). The effects are in line with Indonesia: men who were born to patrilocally inclined parents exposed to the pension program are 3 percentage points less likely to complete primary school and 1.7 percentage points less likely to live with their parents.
The triple-difference in Indonesia and, to a slightly less extent, the double-difference in Ghana are already controlling for a bunch of things (e.g. there are ethnicity by birth year fixed effects to control for overall trends in matrilocality in Indonesia). But she goes the extra mile and shows us robustness through a whole slew of approaches within the existing regressions and also in showing us that other explanations (matri/patrilocality being associated with greater returns to education, the potential role of other gender biased behaviors and whether remittances could be happening) aren’t driving the story.
All in all, this is a neat paper. It gives us a clear picture of how well established and undisputed government policy choices can have unintended and wide-reaching effects. And it shows us how existing data and some creative econometric approaches can help tell this kind of story.
Join the Conversation