1. Discussion of Baugh (2015) “What happens when payday borrowers are cut off from payday lending? A natural experiment”
Brian T. Melzer
Kellogg School of Management
Northwestern University
Household Economics & Decision-making Conference
FRB Cleveland, September 2015

2. Overview
Important and policy relevant question: how does credit access affect consumption and financial distress among low-income borrowers?
Many empirical studies use measures of economic and financial distress – missed bill payments (mortgage, rent and utilities), bankruptcy and foreclosure
Few examine consumption directly (Zaki 2015; Dobridge 2015)
This paper has:
Great data from online account aggregator that provides direct measures of household income, expenditures, borrowing and distress (overdraft/bounced check)
Clever methodology: regulatory intervention to shut down unlicensed and illegal online lenders creates variation in supply of credit

3. Operation Choke Point
DoJ shuts down operations of several online payday lenders - shock to credit supply
Dif-in-dif methodology
Treatment group: borrowers that rely on these lenders may lose access to credit
Comparison group: borrowers that use licensed online lenders and maintain access through this period
Identifying assumption: parallel trends in consumption and distress; some differences in treatment and control groups, so one has to trust that HH FEs soak up this variation

4. Results
Borrowing and loan repayments decline substantially among treatment group, by $101 and $185 per month, respectively
“Financial distress” also declines by 7% among treatment group, as proxied by days with negative balance in checking account and incidence of bounced checks
Effects larger among heavy borrowers and high MPC borrowers
Consumption shows modest and statistically insignificant rise of $15 in treatment group; sub-group of heavy borrowers shows significant increase in consumption

5. Striking evidence on borrower income
My understanding: online payday loans are among most expensive loans
Nevertheless, median income of borrowers in sample is over $40K per year
In borrower surveys, one can explain away high average income as upward bias due to non-response and false reports
But these are administrative data!
Raises possibility that high-cost borrowing is not due to large, unavoidable income shocks, but may be due to consumption choices – perhaps poor budgeting, self- control problems

6. Some diagnostics for dif-in-dif
Build case for parallel trends:
Examine income as dependent variable in dif-in-dif
Examine pre-trends in treatment and control groups – one way to do this is by specifying a fake Operation Choke Point date one year earlier

7. Note on natural experiment
OCP affects treated group in two ways:
They lose credit access
AND
For those with outstanding loan as of intervention, they have a loan forgiven
Paper aims to measure impact of (1)
Perhaps better to construct sample that excludes concurrent borrowers

8. Note on overdraft outcome
Because online payday loans are administered through direct deposit and debit, there will be a mechanical relationship between borrowing and overdraft
Less borrowing will result in less loan default and less overdraft, even if loan access has no effect on household’s ability to meet non-PDL obligations
As a result, hard to interpret finding that overdraft/bounced checks decline for treatment group
Detailed data provides a way forward: separate bounced check events by type of payment and measure incidence of bounced checks on non-PDL payments

9. Notes on consumption results
Adding up? If income is unchanged and loan interest payments decline by $84, why does consumption rise by only $15? Where does remainder go?
Problems with outliers in OLS estimation? I would run the model with log consumption
Show results for models including all expenditures, including check payments
Are users of high-interest loans really committing large amounts to savings?

10. Caution on interpretation
Challenging to draw welfare conclusions when simply observing change in consumption
Results reject view that average HH uses payday loans to their financial benefit – removing access does not result in decline in consumption
But observing increase in consumption in treatment group does not mean those HH are better off
Consumption smoothing: even many proponents of PDL would grant that household will consume less because of interest paid to borrow – more of pie goes toward paying interest, but remainder is distributed more efficiently over time