The height premium: a literature review and meta-analysis

Abstract

The association between adult height and labor-market wages, or the “height premium” (HP), is an important input for quantifying potential economic benefits of nutritional interventions promoting growth. A large economics literature has evaluated this association; however, HP estimates differ greatly depending on the study populations and statistical methodologies used. We conducted a meta-analysis of HP estimates to describe the differences in estimates with different statistical methodologies and to examine potential effect modification of the HP by sex and country income category. We performed meta-analyses for studies using instrumental variables (IV) and ordinary least squares (OLS) methods, separately. OLS estimates were separated into those that were “low-adjusted” for confounding variables and “high-adjusted” for at least one common mediator variable, specifically cognition or schooling. Overall, in a total of 12 studies, the pooled estimates for IV studies indicated that each centimeter increase in height was associated with 3.58% greater wages (95% CI: 1.62-5.54%; I2=97.5%, p<0.001)). In the 24 total OLS studies, low-adjusted estimates indicated an HP of 1.06% (95% CI: 0.85-1.28%, I2=95.5%, p<0.001), while for high-adjusted estimates the HP was only 0.57% (95% CI: 0.41-0.73%, I2=95.8%, p<0.001). Further, the meta-analysis found evidence of effect modification by sex in OLS estimates but not IV, and for both IV and OLS for country income category. Overall, the literature suggests a robust association between adult height and wages; however, the magnitude of the estimate appears to be dependent on statistical methods and covariates selected for multivariable models. Our findings also suggest there may be differences by sex and country income category. Additional analyses are needed taking into account a causal inference framework and, if adult height is being used to capture the cumulative effect on wages of nutritional exposures from conception through adulthood, studies should not adjust for potential mediators including cognition and schooling.