24 December 2019

Vox's Twelve Best Social Science Papers Of The Decade

1. “Free Distribution or Cost-Sharing?” (2010) by Jessica Cohen and Pascaline Dupas 
giving away anti-malaria bednets for free dramatically increases their usage relative to charging a small, nominal fee. 
2. ”Using the Results from Rigorous Multisite Evaluations to Inform Local Policy Decisions” (2019) by Larry Orr, Robert Olsen, Stephen Bell, Ian Schmid, Azim Shivji, and Elizabeth Stuart 
you can’t just take average results and expect that the same effect will hold in your specific case. 
3. “Understanding the Average Impact of Microcredit Expansions: A Bayesian Hierarchical Analysis of Seven Randomized Experiments” (2019) by Rachael Meager 
Meager wrote a more accessible summary here 
Does a study conducted in one place generalize to other places? Does say, distributing bednets for free work well just in the parts of Kenya where Cohen and Dupas did their experiment, or does it work in all malaria-affected countries? Will a charter school chain that appears to deliver higher test scores in Boston work in Houston? This is exactly the problem that paper number two above found to be so serious in education policy: results don’t always generalize.
Meager’s paper is groundbreaking because it offers a way to predict how well study results will generalize. “The relevant question is not whether the effects vary across settings but by how much they vary,” she writes in her summary.
So Meager uses techniques from Bayesian statistics to measure how much the results of a specific intervention — microcredit or microfinance programs for the global poor, of the kind offered by groups like Grameen or Kiva — vary from study to study. She doesn’t have a huge number of studies to go on (only seven) but she’s able to use this method to find that the effectiveness of microcredit varies a bit, but not a huge amount, from place to place. That suggests our evidence on microcredit is reasonably externally valid: The results in a new location are likely to resemble the results in past locations pretty closely, if hardly perfectly. 
4. “Does School Spending Matter? The New Literature on an Old Question” (2018) by Kirabo Jackson 
In this review (ably summarized here for folks without NBER access), Jackson walks through 13 recent papers, many coauthored by Jackson himself, that use highly rigorous near-random methods to measure the influence of money on school outcomes. . . . does pouring more money into public schools improve outcomes? — and the answer, Jackson finds in the research base, is yes. I previously thought per-student funding didn’t matter much; I now think it matters a great deal. 
5. White Backlash: Immigration, Race, and American Politics (2015) by Marisa Abrajano and Zoltan Hajnal 
White Backlash is one of those books published before the 2016 elections that started to feel sharply prophetic as Donald Trump won the Republican nomination and then the presidency. Three years after the defeat of Mitt Romney led to speculation of a new durable demographic majority for Democratic presidents, Abrajano and Hajnal presented a detailed, quantitatively rich counterargument. 
6. “Democracy for Idealists” (2016) by Niko Kolodny 
It’s easy to construct a narrative in which democracy in the United States is eroding. Kolodny, one of the leading political philosophers currently working on questions of democratic theory, quietly posted an article a couple of years ago walking through this literature and trying to determine what, exactly, should trouble us about it and what shouldn’t. Voter ignorance is not a dire threat to democracy, he argues, nor is a lack of “responsiveness,” which he convincingly suggests is an incoherent ideal. What worries him most are concerns about the distribution of political influence: the fact that some Americans’ access to political influence is far greater than that of other Americans. This concern agitates toward an expansion of suffrage and toward resisting efforts to suppress the vote. But it makes our concern with practices like gerrymandering harder to articulate. 
7. “The Coalition Merchants” (2012) by Hans Noel 
If public opinion doesn’t determine the future of public policy, as the studies limned by Kolodny suggest, what does? Noel tells a compelling story that places “coalition merchants” — party activists, sympathetic journalists, and other ideologues — at the center, deciding “what goes with what” and what it means to be a conservative or a liberal. intellectuals like William F. Buckley and groups like Americans for Democratic Action were crucial in identifying support for government services with support for civil rights. 
8. “Valuing the Vote: The Redistribution of Voting Rights and State Funds following the Voting Rights Act of 1965” (2014) by Elizabeth Cascio and Ebonya Washington 
Washington’s paper with Cascio tells a more hopeful story, of what can happen when a disadvantaged ethnic group is finally given suffrage in an authoritarian regime. The Voting Rights Act of 1965 did a huge amount to break up the one-party states that prevailed in most Southern states after the end of Reconstruction, states which some scholars have likened to single-party dictatorships abroad. In doing so it gave black voters, and black communities as units, power over the provision of public goods that they lacked before. Cascio and Washington found that this shift produced meaningful changes, and in particular, a marked increase in government spending. They also offer some suggestive evidence that much of these transfers went to education spending, which (as the Jackson review above suggests) likely improved the quality of instruction for black students. 
9. “Race and Economic Opportunity in the United States: An Intergenerational Perspective” (2018) by Raj Chetty, Nathaniel Hendren, Maggie Jones, and Sonya Porter 
You’d have to actively try to avoid including a paper by Chetty, Hendren, and the rest of the Opportunity Insights lab at Harvard on a list like this, given how much they’ve taught us about economic opportunity, segregation, higher education, and more (I have a big soft spot for Chetty et. al. on Danish retirement savings accounts).
Some of the findings are depressing but unsurprising: Black and American Indian children born into upper- or upper-middle-class families are nearly as likely to fall to the bottom fifth of the income distribution as to stay in the top fifth. Upward mobility for children born into the bottom fifth of the distribution is markedly higher among whites than among black or American Indian children. 
Others are depressing but surprising; conditional on their parents’ income (a big conditional, to be sure) black women outperform white women in terms of their individual earnings. This does not mean there is no income gap between white and black women (black women’s parents, after all, make a lot less on average than white women’s parents) — but it does provide strong evidence against both family structure-based and genetic explanations of racial inequality in the United States. 
10. “Cluelessness” (2016) by Hilary Greaves 
The choices we make have unpredictable consequences that ripple out for centuries or millennia, by affecting life and death. Greaves does a great job of explaining cases where this kind of cluelessness is fine (where we can just make our best guess as to which action will work out best) and in which cases it’s really, really troubling. 
11. “Occupy Liberalism! Or, Ten Reasons Why Liberalism Cannot Be Retrieved for Radicalism (And Why They’re All Wrong)” (2012) by Charles Mills 
he does not, as some Marxists and other radicals do, reject the liberal tradition wholesale. While he acknowledges and emphasizes the explicit racism of figures like John Locke and Immanuel Kant, he nonetheless has tried to develop what he calls a “black radical liberalism” that can overcome these origins. . . . he provides a stirring defense of traditional liberal values — like protection from unnecessary state encroachment on individual liberty — as necessary for racial justice. “Liberalism’s failure to systematically address structural oppression in supposedly liberal-democratic societies is a contingent artifact of the group perspectives and group interests privileged by those structures, not an intrinsic feature of liberalism’s conceptual apparatus,” he writes. 
12. Automating Inequality: How High-Tech Tools Profile, Police, and Punish the Poor (2017) by Virginia Eubanks 
Eubanks studies three specific algorithmic systems currently used by state and county governments in the hopes of making service provision more efficient. The opening example, of Indiana’s botched eligibility system that wound up wrongfully denying access to Medicaid and food stamps to thousands of people, is pretty straightforwardly awful. But the examples of Pittsburgh’s algorithm for evaluating the severity of child abuse and neglect cases, and Los Angeles’s system for determining which homeless people should receive housing assistance, are subtler and in some ways more eye-opening. 
The LA system, for instance, seems to mostly work well — except that it masks the extent to which the city’s problem is structural (a lack of housing supply and crucially a lack of funding for permanent supportive housing) rather than an issue of rationing better through better algorithms. The Pittsburgh system helps remedy a very real problem of limited child and protective services resources for addressing cases of abuse, but because the algorithm is poorly designed and is predicting the wrong variable, it risks criminalizing poverty in certain cases. 
It’s an early example of the harms that misaligned AI can cause as deep learning becomes more and more capable in coming years, and a reminder of what can go awry when politicians mistake technical solutions for political solutions.
From Vox (omissions from original not all noted editorially).


Dave Barnes said...

Kirabo Jackson's paper

andrew said...

Thanks for the link to this very important paper. I've too have been a skeptic on the input contribution to outcomes in education (because the first order effect of student socioeconomic status is so overwhelming next to any other factor, and look forward to seeing a solid counterargument.