Correct Treatment?

[Page007]

Find the fee change effect on suit probability and delinquency dynamics on 2 different samples:

1. A sample which contains all control counties and only those treated counties that experience a contemporaneous "other" treatment -- for example, if the set of all treated counties is the set of counties experiencing civil filing fee change, then the sample in pt. (2) is made from the subset of counties that experience small claims fee change the same year as the civil filing fee change.
2. A sample which contains all control counties and only those treated counties dropped in pt. 2.

What does this accomplish?

• It tells us which fee variation changes suit costs for plaintiffs.
• It tells us which fee variation changes people's delinquency patterns.

[Page007]

Trimmed Sample

From a monthly panel of borrowers (between 2009-17) drawn at random from the BTU data, select the set of borrowers that are present in: (a) Control counties -- counties that do not experience a ff change (b) treated counties for which the econometrician observes at least 2 pre- and post- periods.

(In the new DiD literature parlance, select treatment counties that are treated "in the middle" of the sample time period.) We construct such an estimation panel because we suspect that the downward trending coefficient pattern in the 5th, 6th and 7th lag in the figure below, is driven by only a subset of counties that are treated very early on in the sample time frame.

Particularly, for the computation of the 5th, 6th and 7th lag, the treatment group is mainly composed of counties treated early on in the sample. Thus, there is a high probability that the downward trending pattern is driven by treatment effect specific to these counties.

Estimation on the Trimmed Sample

We estimate the delinquency elasticity to ff changes on the trimmed sample. Some statistics about the trimmed sample:

When Treatment Is Civil FF Change

Number of control counties: 8 Num. of treated counties: 8 Treated counties and their cohorts:

When Treatment is Small Claims FF Change

Number of control counties: 35 (25 CA + 10 others) Number of treated counties: 6 Treated counties that their cohorts:

[Page007]

The link for the ES plots for MH regressions on trimmed sample is here For easy comparison, the link for ES plots for MH regressions on untrimmed sample is here