LETTER LETTER REPLY TO RIEGER AND WAGNER: Context matters when studying purportedly harmful cultural practices David W. Lawson a,1 , Susan James b , Esther Ngadaya c , Bernard Ngowi c , Sayoki G. M. Mfinanga c , and Monique Borgerhoff Mulder b,d Rieger and Wagner (1) present three lines of cri- tique. First, Rieger and Wagner (1) argue that our (2) demonstration that polygyny predicts higher child weight-for-height z-scores (WHZ) in two out of three ethnic groups practicing polygyny should be disre- garded because of inadequacies of this measure. We defend our use of WHZ because: (i) although no mea- sure is perfect, many studies unequivocally confirm that WHZ is a useful indicator of acute malnutrition, highly predictive of child mortality (3); and (ii) regard- less of construct validity, WHZ scores have guided the international development sector for decades, our primary audience in addressing claims that po- lygyny is universally harmful. We also emphasize that our interpretation that polygyny may, in certain contexts, serve female interests rests not only on differences in WHZ, but also on our demonstration that (male-headed) polygynous households are rel- atively food secure and wealthy compared with monogamous households. Second, Rieger and Wagner (1) write that we dem- onstrate that child height-for-age z-scores (HAZ) are “systematically and negatively correlated with polygyny both at the individual and the village levels” (1), and that although our individual-level comparisons are sta- tistically nonsignificant, this can be attributed to small sample size because their study of polygyny across 26 African nations revealed a statistically significant nega- tive correlation at a similar magnitude (4). This critique fundamentally misses our central conclusion: context matters when studying purportedly harmful cultural practices. At the village level, the association between polygyny prevalence and child HAZ is accounted for by underlying contextual differences in ecological vulner- ability (rainfall) and socioeconomic marginalization (ed- ucation). At the individual level, our estimates are based on a specific area of Tanzania, and so cannot be mean- ingfully contrasted with Wagner and Rieger’s (4) esti- mate based on Africa as a whole. Indeed, consistent with our analysis (2), Wagner and Rieger’s cross-national study identifies considerable heterogeneity, with confi- dence intervals crossing zero for 15 of 26 countries and a positive (statistically nonsignificant) association be- tween polygyny and HAZ in Tanzania (ref. 4, p 17). Africa is a diverse continent and polygyny a diverse in- stitution, encompassing variable norms of residence, resource sharing, and spousal recruitment (5). A true understanding of polygyny can only be gained by ac- knowledging this diversity and designing analyses that take context into account (2). Third, Rieger and Wagner (1) suggest our analyses (2) are flawed because we don’t include interactions with child age. Reanalyzing our data selecting only children over 30 mo, Rieger and Wagner (1) report a negative association between polygyny and HAZ. However, our data are not suitable to test for age de- pendencies, which can only be confidently assessed via longitudinal analysis. Furthermore to achieve ade- quate sample size, Rieger and Wagner resort to (i) pooling data across ethnic groups, and so cannot rule out confounding with ethnicity, and (ii) crude compar- isons neglecting wife rank (proxied by household head sex) that proved crucial in our original analysis. Notably, once adjusted for village differences, Rieger and Wagner’s (1) effect estimate also falls short of standard levels of statistical significance (i.e., P < 0.05). We also observe a double standard: Wagner and Rieger’s (4) own cross-national study of polygyny neglects age interactions. In studying “harmful cul- tural practices” it is vital that we apply equivalent stan- dards of evidence independent of whether results meet or contradict conventional expectation. a Department of Population Health, London School of Hygiene and Tropical Medicine, London WC1E 7HT, United Kingdom; b Savannas Forever Tanzania, Arusha, Tanzania; c National Institute for Medical Research, Muhimbili Medical Research Centre, Dar es Salaam, 11101, Tanzania; and Department of Anthropology, University of California, Davis, CA 95616 Author contributions: D.W.L., S.J., E.N., B.N., S.G.M.M., and M.B.M. designed research; D.W.L. analyzed data; D.W.L. and M.B.M. wrote the paper; and S.J., E.N., B.N., and S.G.M.M. collected data. The authors declare no conflict of interest. To whom correspondence should be addressed. Email: david.lawson@lshtm.ac.uk. d www.pnas.org/cgi/doi/10.1073/pnas.1601420113 PNAS Early Edition | 1 of 2