search

QuickPay-Operational-Performance / Data-and-code

Data and code for econometric analysis
0 stars 0 forks source link

Why Defense contracts only (agency code = 97) #47

Closed vob2 closed 4 years ago

vob2 commented 4 years ago

Hi Vibhuti,

Another question on the sample selection. Why defense contracts only (agency code = 97)? (https://github.com/QuickPay-Operational-Performance/Data-and-code/wiki/Results:-Difference-in-Differences-model)

Thanks, Vlad

vibhuti6 commented 4 years ago

Hi Vlad, our sample selection is largely driven by the Barrot & Nanda paper. They use MOCAS data (provided by Dept Of Defense) to show that payment terms sharply fell for fixed priced contracts to small businesses. See Figure 1 here.

vibhuti6 commented 4 years ago

On page 12 to be specific.

vob2 commented 4 years ago

Sure. But why limit to DoD only? Did they limit to DoD? I might be misreading, but I thought they are using MOCAS to validate that payments were received sooner. Because we do not have MOCAS (which works for DoD), why should we restrict to DoD only?

Just trying to figure out what justification we can give in the paper.

vibhuti6 commented 4 years ago

Thanks, Vlad. You are right — their paper is not restricted to DoD contracts. But their unit of analyis is firm or county-sector so it’s probably not a big issue for their paper. What I mean is that a given firm is likely to receive accelerated payments on atleast some of its contracts after quickpay. So on average, treatment and control groups would be comparable.

Our analysis, however, is at the contract-level. We cannot know for sure whether the payment timing for a given set of contracts was similar before quickpay, and different afterwards. This is especially important because the law was implemented on a rolling basis. E.g., following quote from Barrot & Nanda:

acceleration of payment through the Quickpay reform took place between April and September of 2011

So it could be that two contracts receive payments from different payment systems. Though MOCAS implemented this right away, other systems may have been slow to adapt. In the absence of payment data, the best we can do is restrict our sample based on characteristics verified in the Barrot & Nanda paper.

Does that make sense? Please let me know if you have any comments or suggestions. Thanks!

vob2 commented 4 years ago

Yes, this makes sense. Can we allocate companies in treatment and control groups by changing the treatment date so that we err on putting some of the treated firms into the control group? Probably taking the latest date when the law applies would achieve that. If we find the effect then, this argues that for robustness. But this is not a priority for us now.

I guess if DoD is the bulk of contracts and we know from Barrot and Nanda what happenned to DoD contracts. So we are OK with what we have.