variable: receives_financial_assistance

[vob2]

Hi Vibhuti,

What is the distribution of values for receives_financial_assistance variable?

If it is sufficiently populated, we might use it as a proxy for financially constrained firms.

I looked at 2012 numbers and about 18% of contracts have that as TRUE.

Thanks,

Vlad

[vob2]

Similar for receives_contracts_and_financial. Although I am not sure what they mean by receives contracts.

[vob2]

Another thought: c8a_program_participant https://www.mbda.gov/news/blog/2012/03/what-8a-business-development-program

might be another indicator of being financial disadvantaged

[vibhuti6]

Hi Vlad, thanks for sharing this information.

The variables on financial assistance are not present in our sample. I think the data on financial assistance (eg. loans) is provided on a different file, and is not available in the contracts data. The closest fields I could find in our sample is "receives contracts" or "receives grants".

For future reference, I have posted the list of variables we have in our sample here.

We do have the variable on c8a_program, and I will add its description to the summary statistics page. If you have any follow up comments or suggestions, please let me know. Thanks!

[vibhuti6]

One thing we could do is check if the firms in our sample received any financial assistance from the government. This information is currently not in our sample, but we may be able to extract this data from USASpending. For example, see this screenshot displaying the available filters. Our current sample is on contracts, but we could choose loans instead.

[vob2]

That's strange. I downloaded 2012 dataset from USASpending (to get a better feel for the data without having to wrangle with the full dataset) and it has receives_financial_assistance. Could it be that this variable is available in some years only? Could it be that my data is not contracts data (unlikely because variables seem to match mostly). As long as we can get this before and after Quickpay cutoff we can run an analysis on a subsample. Better than nothing.

Approximately 14% of 2012 contracts have receives_financial_assistance = t

Approximately 18% of 2012 contracts have receives_contracts_and_financial = t

Approximately 6% of 2012 contracts have c8a_program_participant = t

Approximately 23% of contracts have at least one of these = t. We can use this as an indicator of financial constraints.

Would you mind checking if we can get these variables for 2010 data?

[vob2]

Not sure if these filters apply to award_type variable or any others. But there are no instances of Loans in 2012 contracts data. They are probably split into other variables.

For example contracts vs IDVs are indicated in one variable but types and descriptions are in others. Not sure how they parse out queries from their web interface.

[vibhuti6]

Hi Vlad, in that case, I think this might be a new variable that is added to the data. In our sample, the column names are same for all years - 2009 through 2018.

It appears to me that any changes to the available fields in the data are applied retroactively. For instance, if you go to https://www.usaspending.gov/#/download_center/award_data_archive , you will see that there is a column stating "Data As Of". Currently, it is as of July 13, 2020. So files for all fiscal years (2008 onwards) will have the variables as of July 13, 2020.

The sample that we are using has variables as of Feb 5, 2020 -- and that raw data is here

[vob2]

Thanks! Can we use this variable?

[vibhuti6]

Yes, I think we should be able to do that. I will download the latest raw data from the website, and get back to you. Re-running the analysis with new dataset may take some time. But before doing that, I will check the distribution of these variables in the full sample. Thanks!

[vob2]

Not sure if we need to rerun the analysis. Can we just merge this variable (and a few others) into the already processed dataset we have?

[vibhuti6]

Yes, that would be better. Let me go through the new files and verify that we have common identifiers -- I say this because there was no unique contract ID in some previous versions of the data and we had to create one. I will keep you posted.

[vibhuti6]

Hi Vlad, I have updated the summary page with description for c8a participants, and competitive contracts (discussed in our last meeting). See items 10 and 11 of the table of content on this page.

I am working on the results for financial assistance variables. Thanks.

[JNing0]

It turns out that good predictors for a firm's financial constraints is a well-researched topic in finance. The last paragraph on p.1988 of Barrot (2016) uses this literature to identify financially constrained firms. I added a cited paper, Hadlock and Pierce (2010), in our Literature repository. Essentially, the firm size, which would be the sales amount in our data, and firm age, were found to be strong predictors of financial constraints. This will add to the contract-level financial assistance data we already have and allow us to separate small and large businesses.

[vob2]

That is correct.

But I think that indicator variable in the data (the one I asked her to check in Issue #40 ) will be a better source of financial constraints.

Size is positively correlated with age (do we have age of our firms?) If we base determination of financial constraints on size and age and sales, then basically all large firms will come out unconstrained and all small firms constrained. We will not have any variability in the financial constrained amount small firms (treatment) and amount large firms (control) to tell us the effect of QuickPay due to financial constrains.

[JNing0]

I agree. I am posting it in case we don't have enough data on these more specific metrics.

Correlating with small and large businesses is good news, because it would allow us to eliminate the assumption that small businesses are financially constrained and use well-accepted standards in literature. This also allows us to expand our sample to include firms who are labeled as both small and large businesses.

We did not collect "age" information last time but I think the dataset has it.

We may need to be careful with the "receives_financial_assistance" entry. I don't know whether it is the same as "contract financing," which indicates whether the supplier needs financing for the associated project. If so, it is a project-level indicator rather than a firm-level indicator. In the summary stats here, for example, a higher fraction of large businesses receive contract financing than small businesses.

Entry c8a probably is a firm-level indicator.

Do we want a firm-level or a project-level indicator in the analysis?

[vob2]

@vibhuti6 was correct. The previous name of variable receives_financial_assistance is receives_grants (https://github.com/fedspendingtransparency/usaspending-website/wiki)

And it looks like this refers to Federal Assistance Awards specifically. My guess is that grants was too specific (grants mean no expectations of the federal officer's involvement in the activities of the project)

[vob2]

@JNing0 My guess is that we need firm-level characteristics. At least theoretically, we want to capture that a firm has financial constraints. But money is money. If project receives financing = firm receives financing. Not sure.

[vibhuti6]

Hi Vlad, thanks for sharing this link! I hadn't seen the GitHub repository for the USASpending website. It looks like their team also has a bunch of other interesting repositories.

So for our analysis, do you think we can use the variable receives_grants or receives_contracts_and_grants to measure a firm's financial strength? Because it appears that only the name of the column has changed. If true, we don't need to merge the information between the two versions of the data?

[vob2]

If we have receives_grants in the existing data, we are probably fine. Not sure if there is a way to test selectively (e.g., for one year) if the data is the same as well.

[vibhuti6]

Yes, we do have "receives_grants" in the existing data. I will think of how we can quickly compare a subset of the two versions -- I will keep you posted. Thanks.

[vob2]

We are now using these variables in our analysis, so I am closing this issue