Limitations

[datarichard]

Limitations to be discussed include:

• narrow and potentially self-serving selection of income variables/wealth inequality + poorly specified model
• other models will of course provide better fits to this data (e.g., a log-normal models are typically used to model income effects), but such models will not provide a clear/distinct change point. An implication is the inflection point we identify may be much smoother than implied by our analysis. We can still defend a weak version of our argument which merely contends the location of the inflection point shifts, if it can be identified.

[datarichard]

Additional limitations include:

• the amount of variance in wellbeing explained by income seems very small (but effect size interpretation is relative - depends on context, theory and consequences). For instance, we have previously reported much larger effects of major life-events on changes in wellbeing for individuals (can we compare delta here?). And the importance will depend on whether wellbeing is more likely to vary between vs within individuals.
• the coefficient (beta) is also small, but this will be biased towards zero due to measurement error in household income and wellbeing so the true coefficient may be larger.

[datarichard]

Potential limitations include:

• If household income is pre-tax then over $100Kish you have to double your income to get a 50 percent increase in income. Our household income measure is after tax
• We expressed income in a linear scale, when many other studies use a log scale. However even if the effect of income on happiness is linear in the log of income, that's basically the same as saying it's not linear in income, and that above some threshold or ceiling you'd need to increase your income by a lot in order to see any substantial increase in wellbeing