Earlier this week, we introduced the SuperFreakonomics Virtual Book Club, wherein we’ll regularly invite readers to chat with some of the researchers and other characters we wrote about in SuperFreakonomics.
Our first guest was University of Chicago economist Emily Oster, whose research, co-authored with Robert Jensen, formed the basis of the section where we discuss how the introduction of television turned out to be an unlikely boon for rural Indian women. (I should have also mentioned that we cite Emily’s fascinating research on how women were regularly put to death for centuries on charges of witchcraft.)
Here are Emily’s answers to your fine questions. Thanks to all for participating. Next up, we’ll feature Sudhir Venkatesh, talking about his and Levitt’s research on street prostitution in Chicago.
SuperFreakonomics Book Club
Emily Oster Responses:
Thanks, everyone, for your questions. Rob and I have done our best to answer them below. Interested in more technical details? You can read the paper for yourself on my webpage.
Have you proven there is a causal link between getting TV and lower birthrate, or just a correlation? It could be that social attitudes were undergoing changes at the same time Indian wages were rising and allowing [some people] to afford TV, and in fact it was the rising economic status that caused women’s lot to improve. Do you know if TV was causal? If so, how have you tested for it? — Jon
This is the central issue in the paper, and the central challenge. The simplest version of this concern is that the type of women who have TV’s are different than the type who do not. For example, television ownership rates in Delhi are much higher than in rural Bihar (a poor state), but we would hardly want to attribute differences between those areas to TV.
Our data are well-suited to address this concern, since we estimate effects using changes in television availability. Put simply, we surveyed 2,500 women in 180 villages in 2001. In that year, 64 of the villages had access to cable TV. We returned to talk to the same women in the same villages in 2002. Between the surveys, 11 of the villages had newly received access to cable. When we returned in 2003, another 10 villages had gotten access; 90 of the villages remained without cable during this period.
To estimate the effect of cable we compared changes in the outcomes of interest over time in villages that changed their cable status versus villages that did not change their status. If cable had a causal effect on the variable we are interested in, we expected to see changes in these outcomes in villages that changed cable access and not in villages that did not change access (those that always had cable or never got it). Further, we expected to see changes between 2001 and 2002 for the villages that got cable in 2002 and between 2002 and 2003 for villages that got cable in 2003. This is exactly what we observed. If you look at the paper, you can see this very clearly in the simple graphs.
This deals with the most basic concern. But getting access to cable is not random. You might still be concerned that there are other changes going on in these villages which get access to cable and those other changes (village is getting richer or more “modern”) are driving the changes in women’s access.
We do several things to address this concern.
First, we do observe income and education. We control extensively for changes in income over time, and for education, age, and other demographics. Second, and perhaps more important, we look at whether there are pre-trends in these outcomes. We ask whether it looks like the status of women is changing in areas that get cable before they get cable. We find this is not the case. In fact, getting cable in the future is not at all predictive of changes in the status of women or fertility. We argue that this addresses the concern that our results are driven by attitudes changing for other reasons.
Do you think the content on TV has a particular effect? Were families who watched a certain type of program more apt to exhibit the change in behavior?
If the content did play a role, do you think there are other applications for this effect? For example, could you affect the rate of STD transmission by providing a specific type of content that resonates with the target population? — AJ
In our data, we do not directly observe the programs people are watching. We can make general statements about content. For example, we know that people typically watch Indian-produced television, and the three most popular genres are soap operas, game shows, and sports. Beyond this, however, we cannot be very specific about what type of television content is driving this.
There actually are programs — not as much in India, but more in Africa — that attempt to promote a social message like the one you describe. For example, there are soaps in Africa that promote an anti-HIV message. I don’t know of much work evaluating their impact on behaviors, although it would clearly be interesting.
One very nice example — unfortunately, not by us! — of a paper that is able to look specifically at content is by Eliana La Ferrara and coauthors (the paper can be accessed here). They explore the impact of soap operas in Brazil and find, among other things, evidence that after the introduction of a particular soap opera, people begin to name their children after characters on the show!
What kind of data did you use in this work? Was it a government-sponsored survey? Did you buy it from some private company? — Felipe Araú
Rob collected much of this data himself, as part of a larger survey. The administrative data on schooling that we use at the end of the paper were purchased from the government and supplemented with another survey that we organized and ran.
Do your data reveal any differences based on either the women’s economic status or caste? What about Muslim versus Hindu? — PragmaticC`ynic
We did not have enough religious variation in the data to estimate differences for Muslim versus Hindu women. We did explore differences by socioeconomic status (education, income) and age. Our data suggest slightly larger effects for older women, but limited differences across education or income.
How were the findings received in India? Second, is there a next logical (but possibly counterintuitive) step that would further help women in rural India? Last, were any negatives associated with the expansion of cable? I read the book and thought this was pretty interesting. We like to think about how we have such control over our decisions, but it’s clear that none of us are so independent as we think. This isn’t limited to India. — charles
As far as we know, the reaction in India has been mostly positive. We’ve been in some media outlets there and not gotten much negative feedback. Having said that, I’m not sure how much effect we are having on policy, either!
In terms of negatives, the short answer is that in our data we do not observe any negative effects; that doesn’t mean that there are not some that we cannot observe.
Have you or has anyone else documented a similar relationship in the U.S.? Can this also go too far and give individuals an unrealistic expectation? — Kris
As far as I know there has not been work on this in the U.S., although economists have used the introduction of television in the U.S. to look at other outcomes, like the effects on children’s test scores. [Editor’s note: a later section in SuperFreakonomics examines the relationship between crime and the introduction of TV in the U.S.; we’ll run a virtual book club on that section when we get to it.] In terms of going too far, I’m not sure. We certainly wouldn’t argue that television is the solution for all the world’s problems with gender inequality, but the effects we find are large enough that they seem worth taking seriously from a policy perspective.