Kevin Croke

Collecting high frequency panel data in Africa using mobile phone interviews

Abstract As mobile phone ownership rates have risen in Africa, there is increased interest in using mobile telephony as a data collection platform. This paper draws on two pilot projects that use mobile phone interviews for data collection in Tanzania and South Sudan. In both cases, high frequency panel data have been collected on a wide range of topics in a manner that is cost effective, flexible and rapid. Attrition has been problematic in both surveys, but can be explained by the resource and organisational constraints that both surveys faced. We analyse the drivers of attrition to generate ideas for how to improve performance in future mobile phone surveys.

Deliberate disengagement: How education decreases political participation in electoral authoritarian regimes

A large literature examining advanced and consolidating democracies suggests that education increases political participation. However, in electoral authoritarian regimes, educated voters may instead deliberately disengage. If education increases critical capacities, political awareness, and support for democracy, educated citizens may believe that participation is futile or legitimizes autocrats. We test this argument in Zimbabwe—a paradigmatic electoral authoritarian regime—by exploiting cross-cohort variation in access to education following a major educational reform. We find that education decreases political participation, substantially reducing the likelihood that better-educated citizens vote, contact politicians, or attend community meetings. Consistent with deliberate disengagement, education’s negative effect on participation dissipated following 2008’s more competitive election, which (temporarily) initiated unprecedented power sharing. Supporting the mechanisms underpinning our hypothesis, educated citizens experience better economic outcomes, are more interested in politics, and are more supportive of democracy, but are also more likely to criticize the government and support opposition parties.

Does Mass Deworming Affect Child Nutrition? Meta-analysis, Cost-Effectiveness, and Statistical Power

The WHO has recently debated whether to reaffirm its long-standing recommendation of mass drug administration (MDA) in areas with more than 20 percent prevalence of soil-transmitted helminths (hookworm, whipworm, and roundworm). There is consensus that the relevant deworming drugs are safe and effective, so the key question facing policymakers is whether the expected benefits of MDA exceed the roughly

100 Years of Mass Deworming Programmes: A Policy Perspective From the World Bank's Disease Control Priorities Analyses

0.30 per treatment cost. The literature on long run educational and economic impacts of deworming suggests that this is the case. However, a recent meta-analysis by Taylor-Robinson et al. (2015), (hereafter TMSDG), disputes these findings. The authors conclude that while treatment of children known to be infected increases weight by 0.75 kg (95 percent CI: 0.24, 1.26; p=0.0038), there is substantial evidence that MDA has no impact on weight or other child outcomes. This paper updates the TMSDG analysis by including studies omitted from that analysis and extracting additional data from included studies, and finds that the TMSDG analysis is underpowered: Power is inadequate to rule out weight gain effects that would make MDA cost effective relative to comparable interventions in similar populations, and underpowered to reject the hypothesis that the effect of MDA is different from the effect that might expected, given dewormings effects on those known to be infected. The hypothesis of a common zero effect of multiple-dose MDA deworming on child weight at longest follow-up is rejected at the 10 percent level using the TMSDG dataset, and with a p value < 0.001 using the updated sample. In the full sample, including studies in settings where prevalence is low enough that the WHO does not recommend deworming, the average effect on child weight is 0.134 kg (95 percent CI: 0.031, 0.236, random effects). In environments with greater than 20 percent prevalence, where the WHO recommends mass treatment, the average effect on child weight is 0.148 kg (95 percent CI: 0.039, 0.258). The implied average effect of MDA on infected children in the full sample is 0.301 kg. At 0.22 kg per U.S. dollar, the estimated average weight gain per dollar is more than 35 times that from school feeding programs as estimated in RCTs. Under-powered meta-analyses are common in health research, and this methodological issue will be increasingly important as growing numbers of economists and other social scientists conduct meta-analysis.

Should the WHO withdraw support for mass deworming

For more than 100 years, countries have used mass drug administration as a public health response to soil-transmitted helminth infection. The series of analyses published as Disease Control Priorities is the World Banks vehicle for exploring the cost-effectiveness and value for money of public health interventions. The first edition was published in 1993 as a technical supplement to the World Banks World Development Report Investing in Health where deworming was used as an illustrative example of value for money in treating diseases with relatively low morbidity but high prevalence. Over the second (2006) and now third (2017) editions deworming has been an increasingly persuasive example to use for this argument. The latest analyses recognize the negative impact of intestinal worm infection on human capital in poor communities and document a continuing decline in worm infection as a result of the combination of high levels of mass treatment and ongoing economic development trends in poor communities.

Tools of single party hegemony in Tanzania: evidence from surveys and survey experiments

1 World Bank, Washington DC, United States of America, 2 Harvard T. H Chan School of Public Health, Boston, Massachusetts, United States of America, 3 Center for Effective Global Action, University of California, Berkeley, Berkeley, California, United States of America, 4 Department of Economics, University of California, Berkeley, Berkeley, California, United States of America, 5 Department of Economics, Harvard University, Cambridge, Massachusetts, United States of America

Commentary: Exploiting randomized exposure to early childhood deworming programmes to study long-run effects: A research programme in progress

ABSTRACT Political systems dominated by a single party are common in the developing world, including in countries that hold regular elections. Yet we lack knowledge about the strategies by which these regimes maintain political dominance. This article presents evidence from Tanzania, a paradigmatic dominant party regime, to demonstrate how party institutions are used instrumentally to ensure the regimes sustained control. First, I show that the ruling party maintains a large infrastructure of neighbourhood representatives, and that in the presence of these agents, citizens self-censor about their political views. Second, I provide estimates of the frequency with which politicians give goods to voters around elections, demonstrating that such gifts are more common in Tanzania than previous surveys suggest. Third, I use a survey experiment to test respondents’ reaction to information about corruption. Few voters change their preferences upon receipt of this information. Taken together, this article provides a detailed picture of ruling party activities at the micro-level in Tanzania. Citizens conceal opposition sympathies from ten cell leaders, either because they fear punishment or seek benefits. These party agents can monitor citizens’ political views, facilitating clientelist exchange. Finally, citizens’ relative insensitivity to clientelism helps explain why politicians are not punished for these strategies.

Influence of Organizational Structure and Administrative Processes on the Performance of State-Level Malaria Programs in Nigeria

A recent article by Jullien et al. offers a critical appraisal of three studies which address the long-run impact of deworming, including my 2014 working paper, ‘The Long-Run Effects of Early Childhood Deworming on Literacy and Numeracy: Evidence from Uganda’. This working paper is part of an ongoing research project which continues to incorporate new data from the study setting as it is publicly released. First I offer responses to the authors’ critique of the 2014 working paper, and then provide information about the ongoing research project. My 2014 working paper examined the long-run impact of deworming, delivered through an earlier cluster randomized deworming trial by Alderman et al., on numeracy and literacy outcomes. In the original trial, all children (treatment and control) received a set of basic health services such as Vitamin A supplementation and vaccination, and those in the treatment group were given these interventions plus deworming treatment. Interventions were delivered every 6 months over a period of 3 years. Examination of the long-run impact of these interventions required a measurement of educational outcomes, for which I used a large-scale survey known as Uwezo, which since 2010 has measured basic numeracy and literacy annually in Uganda, Kenya and Tanzania. The geographical comprehensiveness of the Uwezo survey meant that literacy and numeracy data from the 2010 and 2011 survey rounds were available for a random sample of 710 children from the programme’s target age group during the programme period (from 2000 to 2003) in 22 out of the original 50 trial communities. The authors’ primary critique is that my study uses this subset of the clusters from the original deworming trial; the authors state that, as a result, ‘the analysis is at high risk of attrition bias, due to loss of clusters’. I disagree that there is a priori reason to expect bias in the estimated treatment effect. The communities for the original trial were partitioned into treatment and control randomly (via coin toss), and Uwezo documents state that their survey’s selection of communities was also random within every district of Uganda, and that the selection of households was by systematic random sampling within villages. Thus the clusters sampled by Uwezo should be, by design, a random subset of the full set of trial communities. A subsample of the original trial clusters that is randomly generated in this way should not be a biased sample. Selective or biased sampling should also manifest as imbalance between treatment and control. Although imbalance is possible given the relatively modest number of clusters, in my paper I found no significant differences between treatment and control communities across a range of demographic and socioeconomic variables. While the authors state in Table 3 that ‘some confounders [access to water and private education] appear unbalanced’, both the balance table provided in the original working paper and an augmented version, which I shared with the authors upon their request, show no significant differences between treatment and control communities on these variables. The access to water variable, which the authors highlight, is not significantly different between treatment and control despite relatively large differences because of a very high intra-cluster correlation for this variable (0.72) within communities. This point relates closely to another criticism, which is that the analysis does not address migration. Since the survey captures a random selection of treatment and control communities, there is no a priori reason to expect that selection into treatment 10–11 years earlier would affect migration patterns in late childhood or early adolescence. Nonetheless, migration could be an issue in two ways, both of which work against the possibility of detecting a significant effect. First, it adds measurement error to the definition of treatment, attenuating estimated treatment effects. Second, whereas preschool and primary school age children are unlikely to migrate for school or work, International Journal of Epidemiology, 2016, Vol. 45, No. 6 2159

Can Job Training Decrease Women's Self-Defeating Biases? Experimental Evidence from Nigeria

Abstract Abstract—Studies have found links between organizational structure and performance of public organizations. Considering the wide variation in uptake of malaria interventions and outcomes across Nigeria, this exploratory study examined how differences in administrative location (a dimension of organizational structure), the effectiveness of administrative processes (earmarking and financial control, and communication), leadership (use of data in decision making, state ownership, political will, and resourcefulness), and external influences (donor influence) might explain variations in performance of state malaria programs in Nigeria. We hypothesized that states with malaria program administrative structures closer to state governors will have greater access to resources, greater political support, and greater administrative flexibility and will therefore perform better. To assess these relationships, we conducted semistructured interviews across three states with different program administrative locations: Akwa-Ibom, Cross River, and Niger. Sixty-five participants were identified through a snowballing approach. Data were analyzed using a thematic framework. State program performance was assessed across three malaria service delivery domains (prevention, diagnosis, and treatment) using indicators from Nigeria Demographic and Health Surveys conducted in 2008 and 2013. Cross River State was best performing based on 2013 prevention data (usage of insecticide-treated bednets), and Niger State ranked highest in diagnosis and treatment and showed the greatest improvement between 2008 and 2013. We found that organizational structure (administrative location) did not appear to be determinative of performance but rather that the effectiveness of administrative processes (earmarking and financial control), strong leadership (assertion of state ownership and resourcefulness of leaders in overcoming bottlenecks), and donor influences differed across the three assessed states and may explain the observed varying outcomes.

The impact of mass bed net distribution programs on politics : evidence from Tanzania

Occupational segregation is a central contributor to the gap between male and female earnings worldwide. As new sectors of employment emerge, a key question is whether this pattern is replicated. This paper examines this question by focusing on the emerging information and communications technology sector in Nigeria. Using a randomized control trial, the paper examines the impacts of an information and communications technology training intervention that targeted university graduates in five major cities. The analysis finds that after two years the treatment group was 26 percent more likely to work in the information and communications technology sector. The program appears to have succeeded only in shifting employment to the new sector, as it had no average impact on the overall likelihood of being employed. However, viewed through the lens of occupational segregation, the program had a surprising effect. For women who at baseline were implicitly biased against associating women with professional attributes, the likelihood that the program induced switching into the information and communications technology sector was more than three times as large than that of unbiased women. These results suggest that training programs can help individuals overcome self-defeating biases that could hamper mobility and reduce efficiency in the labor market.