Panel 3_Benjamin J. Keys


Our discussant is Benjamin
Keys from Wharton. Great. Well, thanks so much
for having me today. Thanks to the organizers. And I want to commend
the organizers, as well, for putting a paper like
this one on the program, to get in front of both
policymakers and folks from the financial advice
industry, to think about some of these problems that I
think have been laid out as cleanly as I’ve ever seen
them laid out in this paper. I think this is a paper that
really takes a clever approach to identification, that brings
really unique data to bear, and answers a question that
where you’d think the deck would be stacked very
much against finding the types of results
that they find. So I was really pleased
to have the chance to discuss this paper. Actually, a very
similar motivation in terms of thinking about
this question and why it’s important, about
half of households– thanks to the work
of Annamaria Lusardi and her co-authors
and their surveys pushing these questions
to be more widespread– and so we’re getting a
better sense of this type of financial hardship. About half of all
households say they probably or definitely could not come
up with $2,000 in 30 days. And that’s a pretty
shocking number, especially given how unrepresentative it
is of the people in this room. So this is something that most
of the people in this room would have absolutely
no trouble doing in terms of, I don’t
know, maybe just what’s in your wallet right
now, but more likely what’s coming from
your next paycheck or that’s certainly on
your credit card in terms of available liquidity. And it’s a tough act to
face, but most Americans have very little liquid savings. But at the same time, they face
very frequent expense shocks. So a really useful
survey from Pew in 2015 found that 60% of
households experienced a significant financial
shock in the past year. And the median amount of
the most expensive shock was $2,000, which lines up
well with my first statistic– half of people can’t come
up with $2,000 very easily. And these are shocks,
not just medical bills. And we’ve heard a lot
about health expenses, but also things like auto
repair, home repair, or income volatility, so
shocks to their job in terms of either
hours or wages. And there’s been a lot of
well-documented research that’s shown that income
volatility has risen sharply over the last 25 years. What’s unique
about this paper is that it’s looking in
exactly the opposite place. It’s looking in a place where
you wouldn’t expect the dog to bark. These are social
security benefits that are highly predictable
and that are very certain. So you know you’re getting
a check next month. You probably can go and
consult the calendar and say, OK, well, the
next month, I have 28 days. And the month after that I have
35 days until my next check. So this is extremely easy to
forecast when this next check is coming with certainty. It’s among a population– at
least part of this population is elderly, and another
part is disability. But this is a group that
generally has a relatively low poverty rate, overall. So we would think that among
the retired group of people, who have relatively
low poverty, who may have other resources in
the form of other pensions or other savings, they should
be able to plan around this very predictable event successfully. Whether it’s 28 days or 35 days
just simply should not matter. And what the authors show,
and some really convincing evidence– and I’ve just
added one of the pictures from the paper at
the bottom here– is they show rising overdrafts
over the course of the month among these recipients. And it’s higher in
the months where it’s 35 days between checks. And so even among
this population, where you just simply
wouldn’t expect to find this type of
excess sensitivity to the timing of
paycheck receipt, the authors are
still finding it. And they’re finding
it very clearly. And it’s reflected in a range
of different financial hardships that the authors
measure in terms of overdrafts, in terms
of bounced checks, and some other measures. And so what this paper
is really suggesting is that these short-term budget
constraints are highly binding. And this has been shown in
a range of other studies, as well. So it’s not just
among this group of folks where it’s
this predictable, but also around the
timing of paychecks where your hours may vary. And so there may be more
variability paycheck to paycheck. It’s been shown that there’s
a strong monthly cycle and expenditure on food
among food stamp recipients. So this paper is showing– using these two really
neat new data sets– that the timing of
social security checks matters for a range
of financial outcomes. And I think the highlight
here is this really clean identification, that
there’s no difference in the baseline
characteristics of the people who receive their check
on the second Wednesday, versus the third Wednesday,
versus the fourth Wednesday. But we see these
important differences in their financial
outcomes ex post. So the key findings
here are first, that the five-week
month matters. So whether it’s 28 days
versus 35 days matters. And the other is that
there’s this mismatch between the timing of bills
and the timing of benefits. So the people who receive their
benefit on the fourth Wednesday of the month appear
to do the best in terms of avoiding these
negative financial outcomes. And the authors speculate–
and I agree with them– that the reason is likely
because the expenditures are more closely aligned with the
timing of the benefit receipt. So there’s fewer
days in between where you have to think
about budgeting, or fewer days in between
where you experience some sort of household
shock, that would lead you away from being able to make
your predictable payments. And again, we’re really in an
extremely predictable setting. These are people who know they
have a mortgage payment coming or a car payment coming, know
they’re receiving a benefit check in the coming month,
and are still running into problems bouncing checks. So in terms of the project
and thoughts and directions that the authors could
push just a bit further, I wanted them to
really highlight which subpopulations
in their data are most affected by
these constraints. So we know that the people who
optionally choose to sign up for account aggregator
websites, like mint.com, may be slightly unique. Our concern would be that those
people are differentially more likely to be
disciplining themselves through these websites. I really need help with
my financial planning, so I’m going to these websites. But we wouldn’t think that
would matter across Wednesdays. So we should think that
the authors could still zoom in on the
subpopulations where they may be most sensitive. And the group that seems
obvious to start with is the group where the
social security or disability benefits are the majority source
of income in a given month. So this is going to be much
more binding for that group. If you saw in the
summary statistics, there are quite a few people
who are generating other income inside the household. There’s also attention in the
paper between the disability and the retirement
groups in thinking about how they may
be differentially sensitive to some
of these things. So it’d be nice to try– even if there isn’t a clear
flag in their data for which is which– to try to
separate those out. In terms of the
subpopulations that I think would be
interesting to study, there’s a group here,
clearly, whose spending is outstripping their income. So if you just look at
the summary stats again, you see that the average
expenditures in a given month are far above the
average incomes. And maybe that’s just a
very skewed distribution. But I’d like to see the authors
try to pin down some groups where that might
be more balanced. And then finally, the results
are a little bit less monotone than I would have expected. You would expect that the
day before you receive the check is always
going to be the most difficult day for people. So that day before the check
comes should be, far and away, the day where I’m most likely
to go and get a payday loan or bounce a check. And it doesn’t seem like that’s
as consistent a story as I would have expected. So the authors could,
at the very least, speculate on that path. In terms of policy,
I think, again, this is really the right audience
for this type of paper. And really thinking about what
seemed like small surprises, the difference between
28 days and 35 days on a social security check,
be enough to steer households off course. So is this about
income volatility? Is this about expense shocks? Is this about a
failure to budget? But we should think hard about
which of those dimensions is the real source
of the problem here. We don’t think for this group
that it’s income volatility. They’re receiving the same
size check each month. Then there’s a
question of, what’s the right policy response? Is it just as simple as
making direct deposits weekly or daily? So spread the check out across
every single day of the month. That’s not necessarily
better for some of the reasons that have
come up earlier, in thinking about behavioral biases and
lack of financial planning, that receiving the
lump sum may actually help some people
budget, especially that fourth Wednesday group,
where there’s just not a lot of time between
when that check comes in and when they need to
make their payments on the mortgage or the car. At the very least,
this five-week problem should be directly addressed
through improved communication. I think making the benefit
calendar more salient and seeing some experiments
along these lines going forward would be a really useful next
direction to say, how can we make it salient to
people that they’re going to need to stretch
that check for five weeks in this coming
month, rather than four? And then Gilan mentioned
some of the other tools that we can think
about, whether we just need to design the right
app to better align income and expenditures going forward. But I think this is exactly
the type of paper that really provokes a lot of
interesting policy-relevant questions. And hopefully the
conversation can continue. Thanks very much.

Leave a Reply

Your email address will not be published. Required fields are marked *