Questions about Contract financing model

[vibhuti6]

Hi everyone,

I have been thinking about our current model for contract financing and I have a few concerns:

• We haven't considered that QuickPay itself could have changed the nature of contract financing. Because small projects are being paid quicker, this may come at the expense of lower rates of contract financing. To check this, I plotted the proportion of projects receiving financing over time. From the plot, it appears that fewer small projects receive financing since the inception of JOBS act (which increased availability of credit to small business), and the rate fell even further after QuickPay.

• This change in contract financing over time is not necessarily reflective of improved financial strength of the contractors. For example, the trend for "receives_grants" variable is nearly constant over time. The "receives_grants" variable measures if the contractor receives financial assistance from the government.

• We set our indicator CF to be one if the project receives contract financing. This means two projects of the same contractor in the same time period can have different values of CF. This makes it difficult to argue whether or not the contractor is financially constrained. Ideally, the contractor should be considered financially constrained on both projects.

• I think a more accurate measure of financial constraints would be to check if the contractor was receiving financing on any project in the quarters before Sept 2010. If true, then all projects of the contractor should have CF = 1. This is also a more standard approach. For example, Barrot (2016) defines a firm as financially constrainted based on its financial performance three years prior to the trade credit reform.

• Lastly, I think we should include Treat x CF in our regression equation for this model. We had omitted this term following the specification in Barrot (2016), but they have a robustness check in Table VII where this term is included. I am not sure of any other reason for excluding this term.

I reran the regressions with the above approach, and the results are positive and statistically significant across specifications. The effect is also robust across specifications with Logit model (which was previously insignificant in some columns). For linear regression, this approach also gives robust results if we use the variable "receives_grants" as a proxy for financial constraints.

Please let me know if you think we should use this approach in the paper, and if you have any comments or suggestions. Thank you!

[vob2]

Good question.

I forgot, is the number of small and large projects changing after QuickPay? Is the change in proportions due to numerator or denominator?

When you change the definition of what counts as financially constrained, did results change? I am not sure what you mean by results being positive. Or to be more accurate, I do not remember what we had before.

[vibhuti6]

Hi Vlad, the number of projects hasn't changed much -- please see the plot below. So it appears the change is due to numerator.

The results are still consistent with our theory -- delays are greater for contractors that are financially constrained.

[JNing0]

Thanks, Vibhuti. Good point about the effect of QP on whether a project receives contract financing. It means that, as you illustrated, the criteria for a project to receive contract financing tightens after QP. So we could under-estimate the effect of CF because we assume projects that receive CF before and after QP react to QP the same way.

I also agree with adding the Treat x CF term in the regression.

I am not sure about the other two fixes though. Projects are the subject of our analysis, not contractors. The government policies also clearly state that contract financing is given to aid the completion of a project, not the development of a firm. I find it conceivable to have a contractor with projects that receive and not receive CF. It really depends on the resources demanded by the project. Barrot paper uses the firm metric because its analysis is done on a firm level. Our data and analysis have finer granularity.

As for "receives_grants" variable, I am not sure how it connects with contract financing. As you can see here and from the data dictionary, "receives_grants" is not a means of contract financing...

Let's discuss more in the afternoon.

[vibhuti6]

Hi everyone, I ran the regressions using "receives_grants" as a proxy for financial constraints, and that one seems pretty robust across different models. There's actually not a lot of overlap between projects that "receive contract financing" and "receive grants" so we need to think more carefully about the two proxies. Please see the latest results here -- Section 5 shows the results using "receives_grants". Thanks!