Read on Twitter
This is a useful thread which happens very kind toward my own research. Thanks @Simon_Mongey and @adam_tooze. Let me take the opportunity to articulate one thought and I apologize in advance because I will oversimplify it. https://twitter.com/Simon_Mongey/status/1294664169850130433
Empirically, a number of studies found large expenditure responses to anticipated transfers at the bottom of the distribution. Theoretically, financial frictions (binding constraints or wedges btw borrowing and lending) do exactly that. We've also known this for a decade at least
Quantitatively, though, it's another story. In the data (e.g. the SCF for the US), only about 10% of households, at most, are poor hand-to-mouth, i.e. have zero *net worth*, and the median household has around $100K.
If you solve a one-asset model with constraints, calibrate it to match these facts and try to replicate Chetty's figure you would NOT be able to do it. The response would be strong only for those at the very bottom, but not for the remaining 15-20 pct in bottom quartile/tercile
Why? Simple. In the model, they are not constrained. What you need to reconcile model and data is to separate liquid wealth from the rest of net worth in the model, i.e. include in the canonical income fluctuation model two assets with different degree of liquidity.
In the data, median *liquid wealth* is around $5,000K. The rest of US household wealth is in not easily accessible form (like housing or 401k). Around 20-25% of households are 'wealthy' hand-to-mouth, i.e. have sizable wealth, but it's all illiquid.
A model with these features can replicate Chetty's figure. In the new paper with Kaplan and Moll on the pandemic, which soon will see the light of day, our model replicates Chetty's findings and similar findings from the brilliant team working with #JPMCInstitute data.
Why do households who have wealth not always use it to smooth consumption? As said by @Simon_Mongey and
@mattkahn1966, the logic is similar to @JohnHCochrane
AER 1989 beautifully exposed argument.
@mattkahn1966, the logic is similar to @JohnHCochrane
AER 1989 beautifully exposed argument.
The losses from imperfect consumption smoothing are often second order relative to the gains from investing in a higher-return asset. See Kaplan, Violante, Weidner (BPEA)
https://brookings.edu/wp-content/uploads/2016/07/2014a_Kaplan.pdf
https://brookings.edu/wp-content/uploads/2016/07/2014a_Kaplan.pdf
Themes for another thread: (1) heterogeneity in preferences (like discount factors or IES) and rates of return on saving across households --essentially heterogeneous wedges in Euler equations-- can also deliver some of these same features, but they have their own shortcomings...
(2) when income shocks are large, people want to dip into their illiquid assets. For example the provision of CARES Act that allows individuals to withdraw for free is great. With
@GregWKaplan we are preparing an article for the Annual Review where we explain all this in detail
@GregWKaplan we are preparing an article for the Annual Review where we explain all this in detail
Finally, I don't think it is very useful to frame this debate in terms of what is neoclassical and what is not. It polarizes the discussion in an ideological manner away from content. Let's stick to facts and competing models without attaching labels. Thanks for listening!
