Using Ramadan fasting as a natural experiment, we estimate the long-run impacts
of in-utero health and nutrition shocks on adult outcomes. We exploit administrative
tax return data comprising the universe of income tax returns filed in
Pakistan during 2007–2009. The data allow us to link in-utero Ramadan exposure
of individuals with their later life labor market outcomes. We find a robust negative
effect of Ramadan exposure on earnings (a lower-bound estimate of around 2–3 percent).
The exposed individuals are less likely to be in high-skilled occupations
and less likely to be in the top of the income distribution. Using nationally
representative survey data we show that our results are unlikely to be driven by
selective timing of conception.
Sin taxes are increasingly being used to discourage the consumption of internalityproducing
goods. A major concern with sin taxes is that they are regressive. But
the behavioral biases they seek to correct may also be concentrated among the
poor. The welfare consequences of sin taxes are therefore not clear, depending
inter alia on the distribution of the demand elasticity and behavioral bias in the
population. In this project, we intend to exploit the introduction of South Africa’s
Health Promotion Levy to estimate the extent and distribution of welfare gains
created by a representative sin tax. We will extend the framework of Allcott et al. (2019)
to allow imperfect pass-through of the levy, following O’Connell & Smith (2021).
Combining administrative data from CIT, VAT, and Customs records with
survey data and using event-study and difference-in-differences research designs
we will then estimate the parameters required to assess how the levy affected the
welfare of the poor and rich in South Africa.
It has been argued that reflexive predictions pose a methodological problem to the
social sciences by jeopardizing objective theory testing. This problem apparently arises
from a reflexive prediction’s ability to modify the evidential import conferred by an
observation upon a tested theory. As a classic example, reflexive predictions can lead to a
spurious confirmation of a theory (Merton, 1949). In a recent article, Kopec has argued
that spurious confirmations constitute just one of multiple undesirable consequences of
reflexive predictions, which he argues include both overstating and understating an
observation’s evidential support as well as mistaking counter-evidence for evidence
(Kopec, 2011). I agree that reflexive predictions may alter the evidential import of an
observation; however, I will argue that this is not a genuinely methodological problem for
social science but instead a technical one. Specifically, I will show that the reflexivity of a
prediction is merely a type of statistical bias and, as such, can be dealt with using standard
econometric methods. After developing a new definition of reflexive predictions, I will
show that econometric methods can in principle eliminate reflexivity from the predictions
used to test theories in social science.