7,627
edits
Changes
Jump to navigation
Jump to search
The Effect of Leverage Buyouts on Innovation (view source)
Revision as of 13:40, 10 July 2020
, 13:40, 10 July 2020no edit summary
Brian asked for a range on the Tobin's Q given in table 2. Table 2 should have been produced in DataSetup.do, but the descriptive statistics section appeared missing. Possibly it was cut and paste out during the creation of Version 2. The LaTeX says that the Table was generated from the sheet Descriptives, which is in Tables-Most Recent. It appears to come from some tabstats but on data that had been adjusted for inflation with the CPI, which wasn't loaded until Analysis.do
==Addressing Concerns==
===Beyond JF===
'''Steve Kaplan''':
*Seems consistent with Davis et al. for public-to-privates.
*Would spend more time explaining how it differs from the recent Cumming et al paper. Results seem effectively the
same.
So, Steve didn't read the paper at all. But I don't see the Davis cite in our reference, though I know Brian has written a comment on at least one of his LBO papers. The relevant cite is probably:
Davis, S. J., Haltiwanger, J., Handley, K., Jarmin, R., Lerner, J., & Miranda, J. (2014). Private equity, jobs, and productivity. American Economic Review, 104(12), 3956-90.
Of course, our paper is not consistent with Davis, at least concerning their findings on TFP gains. And we do differentiate our paper from Cummings. We could perhaps differentiate it further though.
'''Josh Lerner''':
#Certainly, our work on PE and jobs, productivity, etc. suggests that public-to-private deals are real outliers relative to other transactions along many dimensions. Would it make sense to look at a broader spectrum of deals? Another advantage is that it would substantially increase your sample size.
#Seagate is a real outlier. As you saw in our paper, we excluded it. How do you handle it? Or do you not include it in your sample?
#One issue we really struggled with was spin-off of entities around the time of IPOs, and differences in policies regarding the assignment of patents. How do you think about this issue?
#The industry categories you use seem quite broad. Thus, I would like to see more exploration of the matching procedure and the robustness of the results to different approaches.
#It would be helpful to talk more about the sources for the sample of transactions, to compare against other databases and the data-sets used in other papers (e.g., the Kaplan work, Kaplan-Stromberg JEP paper, and so forth).
#I could not quite square the numbers in Table 2 with the reported number of buyouts in the overall sample in the text. In general, being more careful on sample sizes, etc., and explaining the details would be helpful.
#Can you look at R&D expenditures after the buyouts for at least some firms (e.g., if they still make security filings)?
Responses:
*Josh's first and fourth comments appear somewhat disingenuous. We made a great play of how it was important to isolate public-to-private transactions and not mix them as he has done. Mixing provides the opposite results, which is good press for the industry... So let's ignore those.
*We address an outlier firm without naming it on page 23 ("One firm held 426 patents...") and page 28 ("Our mean sell-off statistic includes, but is not driven by, a single leveraged buyout selling 798 patents – this is the same firm that went on to have 426 patents by the end of the year
of its buyout, shown in Table 4."). Going through the data, this firm is OWEN-ILLINOIS INC, and the sale was of 767 patents, not 798. Seagate is in the original dataset as SEAGATE TECHNOLOGY PLC (gvkey=150937) and SEAGATE TECHNOLOGY-OLD (gvkey=9545). Neither of these records are marked as having had an LBO. We should look at the source data to see if there is some reason why Seagate isn't in our dataset.
*If spinoffs were being assigned patents, and these assignments were recorded by the USPTO, we would see them as sales in our data. We may want to add a comment to this effect.
*Yes, the industry categories are quite broad. We could spend eternity exploring the matching procedure, so let's ignore that one.
*The comment on sample sizes and explaining details is fair. We need to be clearer about what the unit of observation is when we report results.
*Yes, we can look at R&D filings and we did, we just didn't include it in the write-up. The results are in Graphs-Edited.xlsx
=Version 2=
