World Bank Blogs
http://blogs.worldbank.org/planet.xml
IBRD and IDA: Working for a World Free of Poverty.enTrouble with pre-analysis plans? Try these three weird tricks.
https://blogs.worldbank.org/impactevaluations/trouble-pre-analysis-plans-try-these-three-weird-tricks
Pre-analysis plans increase the chances that published results are true by restricting researchers’ ability to data-mine. Unfortunately, writing a pre-analysis plan isn’t easy, nor is it without costs, as discussed in recent work by <a href="http://economics.mit.edu/files/10654" rel="nofollow">Olken</a> and <a href="http://web.stanford.edu/~niederle/Coffman.Niederle.PAP.JEP.pdf" rel="nofollow">Coffman and Niederle</a>. Two recent working papers - “<a href="http://www.nber.org/papers/w23544" rel="nofollow">Split-Sample Strategies for Avoiding False Discoveries</a>,” by Michael L. Anderson and Jeremy Magruder (<a href="https://are.berkeley.edu/~jmagruder/split-sample.pdf" rel="nofollow">ungated here</a>) and “<a href="http://www.nber.org/papers/w21842" rel="nofollow">Using Split Samples to Improve Inference on Causal Effects</a>,” by Marcel Fafchamps and Julien Labonne (<a href="https://julienlabonne.files.wordpress.com/2017/06/sample_split_simulations_web.pdf" rel="nofollow">ungated and updated here</a>) - propose some very clever refinements to address some of the challenges inherent in pre-analysis plans.<br />
<br />
Two of the big problems are that (a) it is hard to formulate the best way to test a hypothesis without looking at an associated dataset, and (b) even if one knew the best way to test a hypothesis, most papers perform a series of tests, each associated with the outcome of a previous test. Coffman and Niederle’s discussion of the first problem suggests that if an experiment is inexpensive enough that it can be replicated, then the first round of exploratory work can always be replicated by a second round of confirmatory work. Olken refers to the second problem as “pre-specifying the entire ‘analysis tree,’” the combinatorics of which quickly become intractably onerous without the ability to know some of the patterns of results in advance. The two new papers basically wed these insights, formalize them statistically, and present a solution: split your experimental dataset; use the first piece of the dataset for exploratory work, choosing which hypotheses to test and how to test them; refine a pre-analysis plan using this exploratory work; and, plan in hand, use the second piece to actually perform the tests, in essence performing a replication of the exploratory work.<br />
<br />
But, having split the sample, a smaller testing dataset will surely reduce statistical power – the chance that you’ll actually detect an effect if it is there. Or will it? Both papers have a common departure point. Statistical power is the probability of detecting a nonzero effect, conditional on a coefficient truly being nonzero. However, the probability that you – the researcher – successfully detect such an effect also depends on something else: the probability that the coefficient you’ve decided to estimate (true beta, not beta-hat) is actually nonzero. Since that isn’t a sure thing, split-sample methods can increase the odds that you succeed at rejecting a false null hypothesis – and score points for science.<br />
<br />
Bear with me, here comes the math. (Choose your own adventure: if you aren’t sufficiently caffeinated for the math right now, just skip down a few paragraphs.)<br />
<br />
<strong>Fafchamps and Labonne</strong><br />
Fafchamps and Labonne take an approach involving a parameter, <em>psi</em> - the likelihood that, when writing a pre-analysis plan uninformed by actual experimental data, a researcher tests a hypothesis for which the null is indeed not true (i.e. where there is truly a non-zero coefficient).<br />
<br />
Here’s the basic idea. Consider what would happen if a researcher correctly worked out that the statistical power for a test was 0.99. Very nice. However, what if there was only a 50 percent chance (<em>psi</em>) that the test was an interesting one to perform? The other half the time, the researcher tests a hypothesis for which the null is true and there is no effect to find. That means the probability of detecting a true effect is only 0.5*0.99 = 0.495. A great dataset with slim odds of a discovery: a loss for science.<br />
<br />
Fafchamps and Labonne’s suggestion is to split the sample, and test the intended hypothesis, in all the forms one can think of, in the first half. Power for this search process is lower, since the sample is smaller: 0.86. But that means that if the researcher tries all relevant hypotheses, there is an 86 percent chance of detecting it, conditional on stumbling on the right one. Then, whatever hypothesis the researcher picks in the first round, she tests in the second round, having written a more-informed pre-analysis plan. Power? 0.86 again in the second half. Take the product of those two numbers to find the probability of detecting the right formulation in the first half and then having it pass the test in the second half: 0.74. Voila: 74 percent power instead of 49.5 percent power. Progress!<br />
<br />
<strong>Anderson and Magruder</strong><br />
Anderson and Magruder operationalize Olken’s distinction between “primary” and “secondary” hypotheses: a primary hypothesis is about a “key variable of interest,” while a secondary hypothesis is of lesser (or perhaps conditional) importance. Anderson and Magruder consider two parameters: u_h, the importance associated with a hypothesis; and p_h, the prior associated with whether that hypothesis is actually false. A hypothesis with large values of u_h and p_h is likely to be considered “primary.” If either of these is sufficiently small, however, it costs more in power than it yields in expectation to include the hypothesis in a pre-analysis plan.<br />
<br />
Here’s the basic idea. Consider that there is a main hypothesis tested with power 0.8 in the full dataset. Consider a second hypothesis, with the same statistical power on its own, but for which an accurate prior is that there is only a 10 percent chance that the null is false – that the underlying coefficient is actually nonzero. If the researcher tests both hypotheses, she should adjust for multiple testing. The Bonferroni correction to the p-value means that there is now only power 0.71 on each hypothesis. But since the second null hypothesis had only a 10 percent chance of being false, this means we have sacrificed 9 percent statistical power on the first hypothesis (0.80-0.71) while only gaining a 7 percent chance of an additional hypothesis being rejected (0.71*0.10). If the researcher’s objective function is expected total hypotheses rejected, this is a bad deal (0.78 instead of 0.80). A loss for science.<br />
<br />
Anderson and Magruder’s suggestion is to split the sample, but to then do something they call “hybrid.” Leave the main hypothesis alone: it will be tested in the full sample, regardless. It can be in the pre-analysis plan from the very beginning. But use a little bit of the data, perhaps 30 percent, to try out the second hypothesis. That’s a small sample, so be lenient: look for an absolute T statistic of 1.2, for example. Conditional on the 10 percent chance that there is an effect to detect, there is a 63 percent chance of detecting the second effect in this 30-percent sample. (Of course, conditional on the 90 percent chance that there is really nothing to detect, there is also a good chance of a false positive: 23 percent, under the null.) Now, if the secondary hypothesis doesn’t pass the threshold, the researcher just gets to do the one main test; this happens 0.1*(1-0.63) + 0.9*(1-0.23) = 72.9 percent of the time. So the nice feature of the hybrid approach is that, much of the time, the main hypothesis doesn’t need a multiple test correction. Its power ends up being 0.729*0.8 + (1-0.729)*0.71 = 77.6 percent.<br />
<br />
When the secondary hypothesis does pass the threshold, Anderson and Magruder have another suggestion: just do a one-sided test for it. After all, it is wildly unlikely that, if a real effect is at work, it would turn up with the right sign in the one sample split but with the opposite sign in the other. So: test for only the sign that appeared in the first split of the data. (This is a clever way to use a little bit of information from the first split of the data to increase the power of your test in the second split.) With 70 percent of the data remaining, a one-sided test with Bonferroni adjustment (since it is the second hypothesis) has power 0.65. How many hypotheses will be rejected in expectation? 0.776 + 0.1*0.63*0.65 = 0.817. If the researcher’s objective function is total hypotheses rejected, this is a better deal (0.817 instead of 0.800). Progress!<br />
<br />
The math is over. Now to wrap up.<br />
<br />
There were three pretty innovative tricks in these papers. The first is splitting the sample. Though Anderson and Magruder point out that splitting the sample has been used for various purposes in statistics for more than 80 years, this application is a new one. Split-sample approaches help a pre-analysis plan when, <em>ex ante</em>, you can’t precisely characterize the hypotheses you would like to test, or the exact weights you attach to the importance of testing them. They provide more power than guessing the hypotheses, but less power than if you had been sure of the hypotheses from the get-go. The two other tricks? Using a hybrid pre-analysis plan approach; and the one-sided test in the second split. This last trick—the one-sided test in the second slice of the data using the sign from the first slice of the data—improves statistical power, and is one of the very few situations I can think of in which a one-sided test in a pre-analysis plan both legitimately preserves test size and doesn’t risk missing unanticipated negative results – after all, the impacts of new interventions may surprise you! (Examples come to mind in <a href="https://publications.iadb.org/bitstream/handle/11319/6825/Challenges%20in%20Educational%20Reform%3A%20An%20Experiment%20on%20Active%20Learning%20in%20%20Mathematics.pdf?sequence=1" rel="nofollow">education </a>, <a href="https://www.aeaweb.org/conference/2017/preliminary/paper/Z9NksG6G" rel="nofollow">cash transfers</a>, and <a href="http://documents.worldbank.org/curated/en/781951467995662688/pdf/WPS7505.pdf" rel="nofollow">public works programs</a>, to name a few.)<br />
<br />
My discussion vastly oversimplifies both papers. I used the Bonferroni correction, but both papers consider a variety of multiple-testing adjustments, including those that, like Bonferroni, control the family-wise error rate (FWER: the probability of getting at least one false rejection), as well as those that control the false discovery rate (FDR: the fraction of rejections that are incorrect). The methods work, whichever approach you take.<br />
<br />
The Fafchamps and Labonne paper goes on to discuss how this approach might reorganize other aspects of the research process: data management might be divided between the portion of a research team that controls and anonymizes the whole dataset and a separate group that formulates and tests hypotheses in the split-sample while writing the pre-analysis plan; journals might accept papers based only on the pre-analysis plan and the analysis in the first half of the dataset, without knowing what remains significant in the second half.<br />
<br />
The Anderson and Magruder paper goes on to show how their approach could have changed the conclusions of <a href="https://www.povertyactionlab.org/sites/default/files/publications/45_reshaping%20institutions%20QJE.pdf" rel="nofollow">the Casey, Glennerster, and Miguel paper</a> that brought pre-analysis plans to prominence in the context of field experiments in development economics. Anderson and Magruder’s finding serves as a warning: a pre-analysis plan does bind researchers’ hands against data mining and p-hacking, but may also bind them against some important discoveries.<br />
<br />
A caveat.<br />
<br />
There is a looming problem, hinted at by both papers. Lunch (or, in this case, a pre-analysis plan with lots of hypotheses) still isn’t free. Anderson and Magruder report two statistics: among recently-published field experiments, the median T-statistic is 2.6; among recently-filed pre-analysis plans, the median number of tests is 128. The contradiction here is that if your expected T-statistic is 2.6, your unadjusted power is 74 percent. If you adjust the FWER for 128 tests, your power is down to 17 percent. How do we reconcile this? Perhaps field data collection will have to be on a larger scale than before, or only some coefficients require multiple test corrections. Fafchamps and Labonne’s proposed division of labor also appears to necessitate a larger research team than has previously been typical. This trend may place some types of research out of reach for graduate students, or for researchers who are “only” able to secure a few hundred thousand dollars in research funding. No matter how you slice the data, multiple test correction and pre-analysis plans combine to drive the required sample sizes up considerably. If these requirements are disproportionately applied to field experiments, they may be raising the bar in precisely the wrong places: “specification searching and publication biases are quite small in randomized controlled trials,” as <a href="http://evavivalt.com/wp-content/uploads/2014/12/Vivalt_JMP_latest.pdf" rel="nofollow">Vivalt (2016)</a> and the amazingly-titled <a href="https://www.aeaweb.org/articles?id=10.1257/app.20150044" rel="nofollow">Brodeur, et al. (2016)</a> (ungated <a href="https://sites.google.com/site/yanoszylberberg/home/Star_Wars.pdf" rel="nofollow">here</a>) conclude.<br />
<br />
All is not lost. With the rise of “big data” comes massive sample size, and thus the required statistical power. If they arrive sequentially, early waves of “big data” can act as the first split that helps write the pre-analysis plan for later waves. (This only helps, of course, if “big data” somehow obviates the need for the kind of bespoke data collection that is common in current field experiments.) Finally, if you are still having a hard time writing your pre-analysis plan, or you worry that your pre-analysis plans won’t pan out, just do as Anderson, Magruder, Fafchamps, and Labonne have done: <a href="https://twitter.com/willafriedman/status/830156132589195265" rel="nofollow">write papers <em>about</em> writing pre-analysis plans instead</a>.<br />
<br />
<br />
PS – <a href="http://economics.ozier.com/owen/other/ozier-blog--fafchamps-labonne-anderson-magruder-20170711.do" rel="nofollow">here is a short piece of Stata code</a> that produces all the calculations above.<br />
Wed, 12 Jul 2017 08:37:00 -0400Owen OzierDeworming improves child cognition. Eventually.
https://blogs.worldbank.org/africacan/deworming-improves-child-cognition-eventually
<p>
You could be forgiven if you found deworming to be something of an enigma. Some have hailed it as <a href="http://www.povertyactionlab.org/policy-lessons/education/student-participation" rel="nofollow">one of the most cost effective interventions for improving school participation in developing countries</a>. Yet two recent review papers, drawing together the lessons from many studies, find <a href="http://academics.wellesley.edu/Economics/mcewan/PDF/meta.pdf" rel="nofollow">insignificant effects of deworming on learning specifically</a> and only <a href="http://archive.lstmed.ac.uk/3219/" rel="nofollow">uncertain evidence on cognition</a> more generally. How could this be?<br />
<br />
The short answer is that, until a few months ago, both views could be right. I explain why in this 7-minute talk highlighting my <a href="http://documents.worldbank.org/curated/en/2014/10/20258469/exploiting-externalities-estimate-long-term-effects-early-childhood-deworming" rel="nofollow">recent research</a>.</p>
<div class="asset-wrapper asset aid-45 asset-video"> <strong >
Africa Big Ideas: Health </strong>
<div class="content">
<div class="field field-name-field-asset-video-file field-type-emvideo field-label-hidden"><div class="field-items"><div class="field-item even"><iframe src="//cdnapisec.kaltura.com/p/619672/sp/61967200/embedIframeJs/uiconf_id/24449191/partner_id/619672?iframeembed=true&playerId=kaltura_player_1415206006&entry_id=1_gx1yzouy&flashvars[akamaiHD.loadingPolicy]=preInitialize&flashvars[akamaiHD.asyncInit]=true&flashvars[twoPhaseManifest]=true&flashvars[streamerType]=hdnetworkmanifest" width="480" height="300" allowfullscreen webkitallowfullscreen mozAllowFullScreen frameborder="0" style="width: 480px; height: 300px;" itemprop="video" itemscope itemtype="//schema.org/VideoObject">
<span itemprop="name" content="Africa Big Ideas: Health"></span>
<span itemprop="description" content=""></span>
<span itemprop="duration" content="431"></span>
<span itemprop="thumbnail" content="//cdnsecakmi.kaltura.com/p/619672/sp/61967200/thumbnail/entry_id/1_gx1yzouy/version/100011/acv/211"></span>
<span itemprop="width" content="480"></span>
<span itemprop="height" content="300"></span>
</iframe>
</div></div></div><div class="field field-name-field-asset-video-desc field-type-text-long field-label-hidden"><div class="field-items"><div class="field-item even"></div></div></div></div>
</div>
<p>
But if you prefer to read rather than watch the video, allow me to explain.<br />
<!--break--></p>
<div>
Deworming treats a group of neglected tropical diseases caused by parasitic worms. They are generally not lethal, but they live for the most part in the human gut, and absorb key nutrients (including iron) that would otherwise be nourishing their human hosts. Not long ago, widespread worm infections were problematic not only in the tropics and developing world, but in rich countries as well. Where sanitation was poor, parasitic worms flourished. When, early in the twentieth century, a treatment campaign was rolled out in the American South, <a href="http://qje.oxfordjournals.org/content/122/1/73.short" rel="nofollow">it increased school attendance and literacy</a>.<br />
<br />
This study of deworming in the American South is a rarity. We simply don’t have many studies of the impact of deworming on cognition and learning. These outcomes don’t change overnight, so only long-run studies are well-positioned to uncover any effects. Without many studies, reviews of the literature are bound to come up short.<br />
<br />
To make matters worse, the epidemiology of worm infections may have frustrated many early attempts at answering the question at hand. Treating one person’s infections helps everyone around that person by reducing the chance that they are newly infected. This means that studies comparing neighbors who did and didn’t receive deworming drugs could find small effects, in part because untreated study participants actually benefit from medications that their neighbors take (i.e., they don’t get worms from their neighbors). This means that – despite appearances – in many older studies, there simply was no real control group that was not impacted by the intervention.<br />
<br />
That is where things stood, a few months ago.<br />
<br />
But a handful of researchers (myself included) had undertaken long-term follow-ups of cluster-randomized deworming studies. Cluster-randomized means that whole communities were (or weren’t) given deworming medications, overcoming the methodological problem that frustrated earlier work. Now, the results are available for all to see.<br />
<br />
Kevin Croke follows children who received deworming medication through Child Health Days in Uganda from 2000-2003 (<a href="http://www.jstor.org/stable/40699371" rel="nofollow">initially studied by Harold Alderman and co-authors</a>), and <a href="http://scholar.harvard.edu/files/kcroke/files/ug_lr_deworming_071714.pdf" rel="nofollow">finds that ten years later, children who received deworming medication do better on tests of mathematics</a>.<br />
<br />
Sarah Baird, Joan Hamory Hicks, Michael Kremer, and Edward Miguel follow children who began receiving <a href="http://www.poverty-action.org/sites/default/files/miguel_worms.pdf" rel="nofollow">deworming medication in Kenyan schools in the late 1990s</a>, and ten years later, <a href="http://emiguel.econ.berkeley.edu/research/worms-at-work-long-run-impacts-of-child-health-gains" rel="nofollow">find improvements in educational outcomes for women, as well as a variety of beneficial labor market impacts </a>for both genders.<br />
<br />
And in <a href="http://documents.worldbank.org/curated/en/2014/10/20258469/exploiting-externalities-estimate-long-term-effects-early-childhood-deworming" rel="nofollow">my work, I go one step further</a>: Ten years after the deworming, I found the children who were in infancy when the deworming was conducted in <a href="http://www.poverty-action.org/sites/default/files/miguel_worms.pdf" rel="nofollow">those same Kenyan communities</a>, and who didn’t receive treatment at the time. I looked for them because we know that health in early childhood can have lasting implications. And because the schoolchildren in their communities were dewormed in the late 1990s, infants in those communities faced lower rates of infection than they otherwise would have. I conducted surveys and cognitive tests of these children 10 years later to find out whether their improved health in early childhood – thanks to the deworming of their older siblings and neighbors – had a long-term effect on their cognition. It did. <a href="http://documents.worldbank.org/curated/en/2014/10/20258469/exploiting-externalities-estimate-long-term-effects-early-childhood-deworming" rel="nofollow">Children whose communities were dewormed before they were one year old perform markedly better on tests of reasoning than their counterparts in communities where deworming began later.</a> Remember, in early childhood, these children were not treated directly; they just reaped the spillover benefits of fewer worms in their communities. At this critical period in child development, the effect of deworming is long-lasting and easy to see.<br />
<br />
The new evidence has already led to <a href="http://blog.givewell.org/2014/10/03/a-promising-study-on-the-long-term-effects-of-deworming/" rel="nofollow">some re-evaluation of the findings about deworming: it looks like it has cognitive benefits, after all</a>.</div>
Thu, 16 Oct 2014 11:02:00 -0400Owen OzierMicrofranchising in Nairobi hits the BBC
https://blogs.worldbank.org/allaboutfinance/microfranchising-in-nairobi-hits-the-bbc
<p><img src="/allaboutfinance/files/allaboutfinance/s7-final-for-blog.jpg" alt="Photo credit: Sophia Jones-Mwangi/IRC " width="270" height="172" hspace="3" align="right">This week, the BBC and the International Rescue Committee blog both featured a project that I am evaluating together with coauthors Maddalena Honorati and Pamela Jakiela. IRC approached us because they were interested in conducting a rigorous impact evaluation of their project.</p><p>Here are a few of the things IRC has to say about its project:</p><p><i>"NAIROBI, Kenya —</i></p><p><i> In many ways, 19-year-old Susan Kayongo is a typical Kenyan teenager. Brought up by her grandmother in Eastleigh, one of Nairobi’s poorest neighborhoods, she did well in primary school but could not afford to continue her education. Her future looked bleak, like so many young women in her country with little education and work...</i></p><p><i> </i></p><p><i>Susan partnered with nine other teenagers like herself to open the Downtown Salon. Located in a repurposed freight container left behind in the inner city, the parlor is surprisingly inviting, its white walls decorated with bright posters of trendy cuts. The women sell beauty products and hair extensions as well as style hair."</i></p><p><i> </i><!--break--></p><p>The evaluation is still in its early stages, so for now, we won't delve into the program impacts. Because IRC's program provides not only training and startup capital, but also a template business model for young women to follow, we are curious how the effects of this program differ from those found in existing evaluations of other microcredit and microenterprise interventions. If the lack of a viable business plan constrains the effectiveness of capital interventions, then providing one might have a large impact.</p> <p>In the mean time, here is a BBC audio interview with one participant:</p> <p><a href="http://www.bbc.co.uk/news/business-20300644">http://www.bbc.co.uk/news/business-20300644</a></p> <p>And here is a longer description at IRC's blog:</p> <p><a href="http://www.rescue.org/blog/girl-power-kenya-style">http://www.rescue.org/blog/girl-power-kenya-style</a></p> <p>The impact evaluation is ongoing, but we hope to have preliminary results in 2013. Stay tuned!</p>Fri, 16 Nov 2012 10:19:30 -0500Owen OzierEarly Childhood Interventions Conference
https://blogs.worldbank.org/impactevaluations/early-childhood-interventions-conference
<p>Why aren't all early childhood interventions most effective at the same age? Should we be checking that our randomizations are balanced according to genes that influence behavior? Should we be gathering biological outcomes, in addition to economic ones, even when the intervention does not involve biology?<BR><BR>Early childhood interventions - usually working through either health or education – can have very long-lasting effects, some of which are even transmitted to the next generation. Two weekends ago, the Chicago Initiative for Economic Development and Early Childhood (CEDEC) held a conference to survey what is known in this area and provide a forum for sharing findings from recent projects.</p>
<p>In today's post, I highlight a few bits of the presentations that taught me something I didn't know, gave me a reference I wanted to hold on to, or put old findings in a new perspective.</p>
<p><!--break-->The conference program is in two parts, linked here: <A href="http://home.uchicago.edu/~fernandoperez/CEDEC/CEDEC/CEDEC.html">Part 1</A>, and <A href="http://mfi.uchicago.edu/humcap/groups/eci/20120421_conference.shtml">Part 2</A>. Video from the presentations should be available online in the coming week or two.</p>
<p>The conference was interdisciplinary, featuring economists, psychologists, child development specialists, and studies from a variety of countries (even a study of non-human primate behavior). Presentations included methodological innovations, analyses of causal effects of interventions, and novel statistical associations that survive inclusion of enough controls that one starts to wonder whether the links are indeed causal.<BR><BR>A basic insight driving the conference, verbalized by Jim Heckman and Martha Nussbaum (among others), is that if we are concerned with poverty or inequality, we may equally be concerned with poverty or inequality of opportunities and capabilities (in the direction of Amartya Sen's work). To the extent that those are constrained by the environment a child experiences early in life, it essential that we consider interventions that are useful at that age. These interventions may have great complementarities with subsequent investments.</p>
<p>Some highlights that stood out to me:<BR><BR><BR><B>Tools you can use</B></p>
<p>Lia Fernald presented a toolkit for cognitive assessments in early childhood - a wealth of information if you are interested in collecting data - linked <A href="http://siteresources.worldbank.org/INTCY/Resources/395766-1187899515414/Examining_ECD_Toolkit_FULL.pdf">here</A>.<BR><BR><BR><B>The role of biology: mechanisms for economic effects...</B></p>
<p>Biologically measurable processes play an important role in linking early childhood shocks to adult outcomes, even when outcomes are economic. Biomarkers for inflammation are one such group of measures, including C-reactive protein. Elevated levels of this and related biomarkers are known risk factors for disease, and in recent work, Andrea Danese and colleagues find elevated inflammation indicators for children who experienced maltreatment in early childhood.</p>
<p>(See the 2007 paper, linked <A href="http://www.pnas.org/cgi/doi/10.1073/pnas.0610362104">here</A>.)</p>
<p>Greg Duncan and colleagues have been working for some time now on US data from the Panel Study of Income Dynamics, showing that parents' income in the first several years of a child's life are much more predictive of that child's adult outcomes than are parents' incomes in later years. (See, for example, Duncan, Ziol-Guest, and Kalil 2010.) The remarkable connection to Danese's work is that this same identification strategy shows that a child whose parents had higher incomes in the first five years of the child's life also have lower rates of inflammation-linked conditions, such as arthritis and hypertension, in adulthood.</p>
<p><B>...and variation in effects based on biological markers.</B></p>
<p>The <A href="http://www.fragilefamilies.princeton.edu/">Fragile Families study at Princeton </A>is now looking into whether behavioral responses to shocks differ across individuals, according to genetic markers. (An example of this kind of research is <A href="http://www.pnas.org/content/108/20/8189">this 2011 PNAS paper </A>on genes that appear to influence psychological reactions to shocks.) Much of this work is still ongoing, but we can look forward to publications from this group in the coming year or two. Lots of related work is already linked on their website.<BR><BR><BR><B>Lasting impacts and heterogeneity</B></p>
<p>Paul Gertler discussed his work with Christel Vermeersch and an array of co-authors carrying out the latest follow-up of the very well-known Jamaica study by Sally Grantham-McGregor, Susan Walker, and others. Early childhood nutrition and stimulation among stunted children in Jamaica had very positive effects. The nutrition intervention yielded improved physiological growth, though those effects seemed to fade with time. The stimulation intervention, on the other hand, (basically, encouraging sophisticated play involving both children and their mothers or caregivers) had lasting effects on IQ in the teen years, and it seems, wages in adulthood.</p>
<p>Costas Meghir presented some very preliminary results from work that he, Sally Grantham-McGregor, Orazio Attanasio, Emla Fitzsimmons, and Marta Rubio Codina have undertaken in Colombia, building on the lessons from the project in Jamaica, but this time on a larger scale: randomizing treatment arms among 1,400 participants (compared to a few hundred in Jamaica).</p>
<p>Preliminary results in Colombia suggest that the stimulation intervention is having a large impact; perhaps the largest impact is appearing for children who were 18-24 months old at baseline, rather than 12-17 months old. (This might because these are children for whom more complex types of play are actually feasible.)</p>
<p>Jere Behrman put things in perspective: in the well-known INCAP study in four villages in Guatemala, where nutritional supplementation seems to have lasting effects, stunting rates in the population were close to 50%. In Guatemala, nutrition is a really serious issue. In Jamaica, the stunting rate is closer to 4%, so while Walker, Grantham-McGregor, and co-authors identified undernourished children in Jamaica for their study, nutrition simply isn't as prominent a concern in Jamaica in general as it is in Guatemala. Colombia, according to the numbers he showed, is somewhere in between these two cases. Bottom line: similar interventions have different effects, depending on the conditions in the area at the time.</p>
<p> </p>
<p><EM>Owen Ozier is an economist in the Development Research Group, currently studying health, education, and economic decisions in Kenya.</EM> </p>
<P> </P>Thu, 03 May 2012 08:00:38 -0400Owen Ozier