Fruit and Vegetable Consumption and Happiness in the United Kingdom, Australia, and South Africa: A Fixed Effect Instrumental Variable Analysis

Abstract

BACKGROUND: Fruit and vegetable (FV) consumption has been linked to better physical health. Recent studies also link FV consumption to better mental health. The World Health Organization (WHO) has recommended consuming at least five servings of FV per day, but recent statistics find most populations (in both developed and developing countries) achieve far below the target, despite knowing “FV is good for health”. Two recent studies using large survey data from Australia and from the U.K. using fixed effect (FE) regression found significant and large positive associations between FV consumption and life satisfaction; the authors caution that fixed effect regression cannot show causality. METHOD AND RESULTS: Using data from the U.K., Australia, and South Africa, we apply FE and instrumental variables (IV) regressions to show that the effect of FV on life satisfaction is plausibly causal. We use the consumer price index (CPI) of FV as the instrument (F statistic > 16). We test whether controlling for individual time invariant effects was necessary, by using correlated random effects as control functions; we find that controlling for time invariant effects is not necessary for the relationship between FV and life satisfaction. We further test whether the FV variable is endogenous after controlling for FE; we find FV remains endogenous even with FE. We calculate the elasticities of FV’s effect on life satisfaction and find similar elasticities across the three countries; the increase in life satisfaction from increasing average daily FV intake by one serving would more than offset the negative effect on mental health from becoming unemployed or developing a chronic illness. To explore whether the large effect size from IV estimates is due to heterogeneous treatment effect, we apply marginal treatment effect (MTE) estimation on WHO’s five-a-day recommendation. We find that the MTE curve is downward sloped: Although some may derive large gains in marginal utility by switching from not-meeting to meeting five-a-day, others may derive little gain. Examining the potential outcomes of meeting versus not meeting five-a-day (conditioning on the individual differences in observed and unobserved resistance to meeting WHO’s guideline), we find that anyone who meets the five-a-day derives a potential high life satisfaction -- regardless of the level of resistance to eating FV. The key difference lies in the individual differences in the potential outcomes when people consume low amounts of FV. Individuals with high resistance in achieving five-a-day still have relatively high life satisfaction despite not consuming five-a-day. However, individuals with low resistance in meeting five-a-day derive very low life satisfaction had they not consumed enough FV. DISCUSSION: A public information campaign to “educate” that people will be “happy” with more FV will likely be ineffective. This is because those who derive greater happiness with more FV consumption already have low resistance to treatment and thus already consume their five-a-day. Those who do not consume five-a-day also know they are just as happy eating versus not eating five FV servings a day, so the campaign to them is irrelevant.