Public-to-Private Buyouts and Innovation

We study the effect of public-to-private buyout transactions on investments in innovation using an international sample over the 1997-2017 period. We use patent counts and citations to proxy for the quantity, quality, and economic importance of innovation. Our results are based on time analysis and matched sample regressions. The data indicate that buyouts are associated with a significant reduction in patents and patent citations, including a reduction in radical (i.e., more scientific) patents. When we split the sample into institutional and management buyouts, the negative effect of buyouts is confirmed only for institutional buyouts. This suggests that only institutional buyouts prevent target firms from adopting long-term investments. This finding is confirmed by reductions in innovator employment and innovation efficiency subsequent to going private. Moreover, the data indicate that the negative effect is most prevalent for transactions where the cost of the deal’s debt financing is higher than that of the debt post-buyout. We rule out some alternative explanations for these findings, including but not limited to outliers, truncation bias, and endogeneity.


Introduction
The global economy has undergone a profound shift in ownership structure over the past few decades. A significant share of firms is now owned by institutional investors from the private equity (PE) industry, and the effect on target firms continues to be debated. PE firms acquire publicly listed firms, delist them, and restructure them. Post-buyout transaction, existing theories suggest that, theoretically, target firms' operating performance, investment, and productivity should improve (Jensen, 1989). The intuition is straightforward: PE managers are value-adding active investors that put into place efficient incentive and monitoring mechanisms, together with debt discipline, to enhance firm productivity and performance (Ahlers et al., 2017;Amess, Stiebale and Wright, 2016;Cornelli and Karakaş, 2015;Jensen, 1989).
On the other hand, however, critics argue that PE firms are transitory organizations (Kaplan, 1991). They have an overly strong focus on projects with short-term payoffs, and tend to reduce investments in long-term projects in order to ensure they can meet their debt servicing obligations (Rappaport, 1990). One example of the "dark" side of PE deals is the buyout of Debenhams, a public-to-private deal that took place in 2003 in the U.K. This deal generated enormous profits for the PE owners, but it left the firm with massive debt, and its value plummeted after the IPO. 1 In subsequent years, it was not able to service its debt, and was taken over by its lenders in 2019. 2 Another example is the $24 billion buyout of Dell Technologies Inc. by PE firm Silver Lake in 2013, which currently stands as the largest technology firm buyout. In this case, the company was in tatters by 2018, with its financial position described as: during the most recent buyout wave of 2006-2007 and the post-2008-2009 global financial crisis. The 2006-2007 buyout wave allowed the PE firm to exploit cheap access to credit and this might have changed their motives for such transactions. We note that prior work on buyouts and innovation generally preceded the global financial crisis, and focused largely on U.S. and U.K. data.
In this paper, we use a comprehensive sample of public-to-private buyout transactions, and a dataset that covers the most recent buyout wave and financial crisis. We study two specific transaction types: institutional buyouts (IBO), and management buyouts (MBO). In an IBO, the PE fund acquires a controlling interest in the target firm, hires new management, and typically exits within five years; in an MBO, current management takes a large ownership stake in the company. The goals of the two groups may be very different. IBO investors are mainly focused on delivering returns on the transaction; MBO investors are focused on servicing sustainable debt, as well as on long-term planning.
To study how public-to-private buyouts impact long-term innovation, we use a unique international dataset over the 1997-2017 period. Our measures of innovation are based on patents registered in each country's office provided by EPO's Worldwide Patent Statistical Database (PATSTAT). The depth of the data allows us to create measures that have not been used before, such as radical innovation and innovation efficiency.
Our tests are based on before-after buyout analysis, with fixed effects and differencein-differences methodology for a buyout and control sample of public firms. We find that buyouts generally reduce investments in innovation as measured by the number of patents and citations. These effects are quite substantial in terms of quantity and quality. We observe a 17% post-buyout decline in the number of patents in year 2, and up to 22% by year 3. The drop in quality ranges from 24% to 45%, and is observed mostly in years 2 and 3. When we distinguish Electronic copy available at: https://ssrn.com/abstract=3295199 between institutional and management buyouts, we find that, in the case of public-to-private buyouts by institutional investors, the effect on innovation remains negative. The analysis of public firms taken private by management is inconclusive.
We find a consistently negative effect following public-to-private buyouts for a sample of firms that engaged in what we refer to as radical innovation, i.e., a higher level of scientific innovation that cites non-patent literature. Furthermore, we test whether target firms become more efficient in terms of innovative activities. We find that innovation efficiency decreases after a public-to-private buyout. We also find that the negative effect depends on PE investors' syndicate size.
We also try to explain the underlying reasons for the negative effect of buyouts on innovation, and find it is mostly driven by the relative cost of debt. In particular, we find that, if the acquirer cannot lock in financing for the buyout transaction at a lower rate than other market participants' cost of debt, it may negatively impact investment and therefore innovation.
This article contributes to the literature on the effects of ownership changes on innovation, and, in particular, the effects of buyout transactions on innovation. Lerner, Sorensen and Strömberg (2011) find that innovation increases after leveraged buyouts (LBOs).
Their U.S.-based sample extends through 2005, and the vast majority of their deals are privateto-private transactions. Similarly, Amess, Stiebale and Wright (2016) show an increase in innovative activity for their sample of U.K. deals, although they state that most of the effect comes from private-to-private transactions.
In contrast, our study differs in a number of ways. For example, we study public-toprivate buyouts, and our sample is international. The distinction between public-to-private and private-to-private deals is important as suggested in previous studies, yet none of them focused on a detailed analysis of public-to-private deals. We also capture a different time span, which covers the global financial crisis and the second LBO wave (2005)(2006)(2007). We find some evidence that this negative effect of buyouts on innovation is mostly existing in post-2006 period, versus pre-2006 period. We provide evidence based on "time-trend" analysis for the buyout sample, as well as "difference-in-differences" results for the matched sample, which mitigates potential endogeneity concerns.We also distinguish between management and institutional buyouts, which affect changes in innovation in very different ways.
The rest of this paper is organized as follows. The next section outlines a literature review and develops our hypotheses. The third section discusses our research design, while the fourth section presents the data. Our main results are in the fifth section, followed by a sixth section presenting additional analyses. The last section concludes.

Literature Review and Hypothesis Development
Prior literature suggests that ownership structure plays an important role in corporate innovation, because it represents the financing choices, governance, and incentives of the owners. An early study by Aghion and Tirole (1994) explores the existence of innovation under different structures. Belenzon, Berkovitz and Bolton (2009) suggest that companies choose the corporate form that is the most conducive to undertaking research and development. Many studies have also examined how various ownership forms affect innovation, focusing explicitly on short-versus long-term value creation. Some research has found that firms may not invest in long-term projects due to shortterm performance pressures (Stein, 1988;Graham, Harvey and Rajgopal, 2005;Bushee, 1998Bushee, , 2001. Owners may expropriate firm resources and impede innovative activities. Manso (2011) suggests that, ideally, organizing and motivating systems should build in a certain amount of tolerance for failure, as well as reward for long-term success. Moreover, Ferreira, Manso and Silva (2012) show that going public is optimal when exploiting existing ideas, and going Electronic copy available at: https://ssrn.com/abstract=3295199 private is optimal when exploring new ideas. Empirical evidence shows that innovation generally declines after private firms go public (Bernstein, 2015).
Alternatively, some owner types may enhance innovative activities by acting as active monitors, encouraging management to invest in long-term projects (Shleifer and Vishny, 1986;Kahn and Winton, 1998;Burkart, Gromb and Panunzi, 1997;Gillan and Starks, 2000). For example, Aghion, Van Reenen and Zingales (2013) show that institutional ownership is associated with more innovation. Boot and Vladimirov (2018) show that ownership and innovation can even exhibit a U-shaped relationship when we take into account market collusion, where public ownership nurtures innovation when the probability of success is either very low or very high. Financing of innovation also matters. Atanassov, Nanda and Seru (2007) show that public firms that rely on equity or public debt tend to be more innovative.
The buyout transaction is a particular form of ownership change, generally undertaken by PE firms or firm management using a substantial external source of funding (usually debt).
Intuitively, the purpose is to restructure the target firm. Investors aim to install more efficient incentive mechanisms and monitoring, and to improve corporate governance and capital structure (Ahlers et al., 2017;Amess, Stiebale and Wright, 2016;Cornelli and Karakaş, 2015;Lerner, Sorensen and Strömberg, 2011;Jensen, 1989). There are several theories that motivate the value gains from public-to-private transaction through those restructuring activities that in theory rely on the reduction of shareholder-related agency costs. This agency conflict might impose significant costs on public firm shareholders due to fact that the manager acting as an agent has decision power and informational advantage over shareholders (Jensen and Meckling, 1976). The change in ownership should reduce those costs and improve the target firm value, its performance, and productivity. 5 However, although the intended goals of public-to-private buyouts are to improve target firm performance, the debt burden may ultimately have a negative effect on its long-term investment. Kaplan (1991) states that PE firms are transitory organizations that focus on projects with short-term payoffs while reducing investments in long-term projects. Rappaport (1990) claims that debt discipline and concentrated ownership can impose significant adjustment costs. Debt significantly increases the leverage of target firms, and default risk becomes a primary concern. Moreover, financing is often sourced from multiple debt providers, so refinancing becomes more difficult to achieve (Demiroglu and James, 2010;Graham and Leary, 2011;Colla, Ippolito and Wagner, 2012;Axelson et al., 2013).
We distinguish further between IBOs and MBOs, the two types of buyouts. In the case of IBOs, the PE fund, as the owner, is in fact an intermediary that must provide returns to its investors. PE firms represent limited partners that provide funding. The limited partners typically expect to be repaid within five to ten years. Therefore, although the investment search for the PE fund may take two to three years, the actual turnaround period can last for three to seven years. Subsequently, PE funds plan exit that might include the return of the target firm to public ownership or sale to another acquirer. Most exercise that option between the second and fifth year post-buyout (Kaplan, 1991). Therefore, PE funds' investment horizons are generally up to five years. Therefore, while in general the effect of IBOs on innovation might be positive due to efficient incentive mechanisms, concentrated monitoring, and improvements in corporate governance and capital structure, the short term turnaround periods and excessive debt pressure might dampen the long term investments in innovation.
MBOs are subject to similar pressures as IBOs, yet the additional factor that plays a role in these types of deals is the shareholders' intolerance of failure. Kamoto (2017) shows that it deteriorates managerial innovation incentives in public firms. Therefore, going private in an MBO deal releases the management from being subject to the dismissal risk posed by the Electronic copy available at: https://ssrn.com/abstract=3295199 shareholders' intolerance of failure. Yet, we do not know to what extent the short-termism of investors on managerial innovation incentives affects the post-MBO long term investment.
Subsequently, the corporate governance issues exacerbated by dispersed ownership aggravate agency problems. Thompson and Wright (1995) suggest that through MBO the bureaucratic incentives are being replaced by market-based incentives. The reunification of ownership with control after an MBO should motive owner-managers to profit maximise. In the context of patenting, we would therefore expect that managers involved in an MBO to be financially motivated to pursue patenting activity if they believe that patenting activity is consistent with maximising profit.
Thus, theory remains somewhat unclear about the actual effect of public-to-private buyout transactions on long-term investment, but the empirical evidence does not offer a compelling answer either. The literature has mostly debated the effects of buyout transactions on operating performance, productivity, and employment, with mixed results. We summarize below.
Early evidence suggests that the impact of PE buyouts showed positive effects on productivity based on plant-level data (Lichtenberg and Siegel, 1990). There is also some evidence of improved operating performance during the first buyout wave (Kaplan, 1989;Baker and Wruck, 1989;Smith, 1990). More recently, Davis et al. (2014) show that, while buyouts can lead to job losses, they also tend to bring productivity improvements. Guo, Hotchkiss and Song (2011) find evidence of a positive effect on productivity after a buyout.
Similarly, Acharya et al. (2012) and Weir, Jones and Wright (2015) find small improvements in operating performance post-LBO for a U.K. sample, while Boucly, Sraer and Thesmar (2011) and Bergström, Grubb and Jonsson (2007) find larger operating improvements post-LBO for other countries. Harford and Kolasinski (2013) study wealth creation at the time of PE investor exit, and find no evidence for the short-termism view of buyouts.
Other studies, however, present a different view of the effect of buyouts on target firm efficiency. Bharath, Dittmar and Sivadasan (2014) use a U.S. sample, and find that going private does not seem to change firm productivity. In fact, they find some evidence of underinvestment. Leslie and Oyer (2008) and Cohn, Mills and Towery (2014), using a sample of U.S. LBOs, find little or no evidence of operating improvements following a buyout.
Similarly, Ayash and Schütt (2016) find no economically significant improvement in operating performance following buyouts, and Ayash and Rastad (2017) question productivity improvements claimed in prior literature. In a U.K. buyout context, Wood (2014a, 2014b) show that the performance and productivity of IBOs tend to decrease post-transaction.
Moreover, the effect of buyouts may depend on investor type, as Ughetto (2010) finds for private-to-private transactions in Europe. There is some evidence that PE IBOs tend to have a negative effect on employment and productivity Wood, 2014a, 2014b;Guery et al., 2017), but the opposite effect has been found for MBOs. Kaplan (1989), Smith (1990), and Smart and Waldfogel (1994)  To summarize the theoretical arguments and extant empirical evidence, it is not clear ex ante whether public-to-private transactions have a positive or negative effect on innovation.
On the one hand, the improvements in corporate governance, managerial incentives, and discipline should positively impact innovation. On the other hand, the debt burden and the short-term constraints imposed on PE investors can significantly hamper innovative activity.
The effect may also differ depending on investor type (institutional or management).
Ultimately, leave it as empirical question.

Research Design
Our research design focuses on two sets of results. First, we analyze the "before-after" time trends for the sample of firms that went private. We compare innovation levels after going private to those exhibited when the firms were public. Second, we implement "difference-indifferences" (DiD) tests to analyze the changes in innovation of going-private firms compared to a control group of matched firms that remained public.

The "before-after" methodology
In order to examine the changes within the going-private group, we run the following Ordinary Least Squares (OLS) regression: where , is the outcome variable (innovation measures), are dummy variables that equal 1 for the year k after the buyout transaction (negative values correspond to years before the buyout), Controls is a vector of country characteristics, and FE are firm-and year-country fixed effects. The term , stands for residual error. The estimated coefficients on betas are the average effect of the buyout transaction for a particular year.
The "difference-in-differences" methodology The "before-after" analysis of innovation for firms that went private is ultimately driven by country-, industry-, or firm-related characteristics such as age and size. In order to eliminate this potential source of endogeneity, we form a matched control group for each going-private Electronic copy available at: https://ssrn.com/abstract=3295199 firm in our sample of buyouts. Similar to the procedure for the going-private firms, we first ensure that the control group firms have patent activity. Then, we select up to five matched control firms that remained public based on country, industry, event year (announcement), age, and size. We thus have "cells" of one going-private firm, and up to five matched controls. We delete "cells" where the number of control firms is lower than three. We estimate the following regression model: where , is the outcome variable (innovation measures), and are dummy variables that equal 1 for the year k after the buyout transaction (negative values correspond to years before the buyout). We omit year 0. FEC are "cell" fixed effects. The term , stands for residual error. The estimated coefficients on deltas are the average treatment effect of the buyout transaction for a particular year compared to the control sample.

Sample construction
To establish our sample, we first obtain buyout transactions from the Zephyr database. 6 We only analyze deals where the acquirer bought 100% of the listed target firm. We choose the Zephyr database because it shares common identities with the Orbis database. We then merge Zephyr and Orbis with the detailed patent data derived from EPO's Worldwide Patent Statistical Database (PATSTAT), for which Bureau van Dijk has assigned unique applicant firm identifiers. 7 The PATSTAT provides data on patent applications filed in over 90 offices around the world. It contains basic bibliographic information, including date of application, date of patent grant, track record of patent citations, and inventor identification for each patent application. PATSTAT is published biannually; we use the 2017 Autumn edition.
The PATSTAT database covers patents filed in ninety-three countries. It therefore provides even greater coverage than the National Bureau of Economic Research (NBER) Patent and Citation database, which is compiled from information in the United States Patent and Trademark Office (USPTO) (Moshirian et al., 2015). The USPTO only aggregates patents filed in the U.S.
In summary, using databases that share common identifiers allows us to avoid many pitfalls. Both Zephyr and Orbis are provided by the same supplier, Bureau Van Dijk, so we can match deal information to firm-level data more accurately. We further match these data with PATSTAT. Using PATSTAT data, we can directly measure firms' innovation levels, regardless of where the patent application was filed.
We include all completed buyout transactions from 1997 to 2011 for an international sample of countries that includes Austria, Belgium, Canada, Switzerland, Czech Republic, Germany, Denmark, Spain, Finland, France, the United Kingdom (U.K.), Israel, Italy, Japan, Korean Republic, Netherlands, Norway, Poland, Sweden, Singapore, and the United States (U.S.). Our sample is mostly dominated by U.S. deals, followed by those in Canada, Japan, France, the U.K., and Germany. Our sample of buyout deals terminates in 2011, because we require six years' of post-buyout patent data in order to construct the patent citation measures.
We only include buyout deals where the target firm had at least one successful patent applied for and granted from the three years prior to the three years after the transaction (similarly to Lerner, Sorensen and Strömberg, 2011). Our final sample is comprised of 307 going-private deals involving 26,360 patents. 33% of those transactions are done in cross-border deals where the target firm and the acquirer are form different countries. Generally, 38% of acquirers are classified as a very large company (i.e. total assets greater than 260 million USD), 10% as a large company (i.e. total assets greater than 26 million USD), 10% as a medium size company (i.e. total assets greater than 2.6 million USD), and 42% as a small company.

Measuring innovation
Our primary goal is to measure innovation quantity and quality. We use a simple patent count to proxy for innovation quantity. In order to evaluate the quality and importance of innovation, we use two other measures. The first is absolute citation count, which captures citations made within the three-year period from the year of patent grant date to the three years afterward. We use this measure to mitigate the issue of truncation at the end of the sample. The second is relative citation count. This measure calculates citations received for patents filed and subsequently granted during the year of the patent grant through the three years afterward, less the average number of citations during the period received by the matching patents. We follow Lerner, Sorensen and Strömberg (2011), and define matching patents as those granted in the same year and assigned to the same technology class. 8 Because absolute and relative citation measures require three years of forward patent data, and because our study requires citation measures for three years from the date of the buyout, we require a total of six years of patent data from the date of the buyout. This limits us to considering buyout transactions up to 2011.

Control variables
Many factors drive innovative activity at the country and firm level. Following previous literature, we control for these characteristics. In particular, we include the intellectual property protection index created by Park (2008), and the level of a country's innovativeness as measured by patent applications scaled by GDP. Nanda and Rhodes-Kropf (2013) and Hsu, Tian and Xu (2014) show that financial market development affects innovative activity. Thus, we include equity development measures as proxied for by the value of shares traded and scaled by GDP, and two credit market development measures. CMD1 is domestic credit to the private sector. This is an important indicator of the ability to finance production, consumption, and capital formation, which in turn affects economic activity. CMD2 is domestic credit provided by the financial sector scaled by GDP, which measures banking sector depth and financial sector development in terms of size. We also include GDP growth of a country to proxy for general economic conditions. We provide definitions for all variables and data sources in Appendix AI. [ Table 1 here] [ Table 2 here]

Summary statistics
Subsequently, we compare both samples of treated and control observations in terms of their innovation. The firms targeted in the buyout transactions have an average of 11.39 patents.
The relative and absolute citations are 10.60 and 5.35, respectively. In Figure 1  [ Figure 1 here]

Baseline regressions
In the multivariate analysis, we use patent count and citations as dependent variables. Given that the patent count variable is highly skewed, we transform it into ln(1+patent count) in the regression analysis. 9 In column (1) of Table 3, we present the results of a "before-after" analysis, where we include industry, firm, and country-year fixed effects, and cluster standard errors by firm. We find a significant decline in the number of patent applications post-buyout transaction, ranging from 18% in year 2, to 22% in year 3. 10 This is also economically significant, translating into up to 3 patents less each year.
The innovation drop may be due to the fact that PE firms tend to buy certain firms. In this analysis, we match the buyout firms to public firms by age, profitability, year, and country in order to mitigate those concerns. Our empirical tests are based on DiD methods, where we compare change in innovation among firms that went through a public-to-private buyout (the treatment group) with change in innovation among a matched group of public firms that remained public (the control group).
In column (2) of [ Table 3 here] In columns (1) and (2) of Table 4, we present the results of "before-after" analyses where the dependent variables are Absolute Citations and Relative Citations, respectively. We include industry, firm, and country-year fixed effects, and we cluster the standard errors by firm. We find a significant decline in the number of citations post-buyout transaction, ranging from 32% to 41% in year 2, and from 36% to 46% in year 3. 11 In column (2) of Table 4, we present the results from the difference-in-differences methodology. The results show a decline in the number of citations of 24% to 33% in year 2, and 22% to 30% in year 3, compared to public firms matched by year, size, three-digit industry, and age. The 46% reduction in patent citations measured as absolute citations translates into up to 5 citations less each year compared to the mean number of absolute citations; and the 36% reduction in patent citations measured as relative citations translates into up to 2 citations less each year compared to the mean number of relative citations.

Addressing endogeneity of the going-private decision
A decision to delist a public firm is not random, and therefore our analysis is subject to endogeneity. In the previous section, we attempt to mitigate this concern by performing a DiD analysis that matches on industry, size, age, and year. In this subsection, we extend this analysis by employing two alternative matching methods. First, we construct a sample of matched firms that are similar in terms of going-private characteristics. In particular, following Bharath and Dittmar (2010), we identify several characteristics as future predictors of going-private. We create a sample that is similar to the buyout sample in terms of total assets, sales, R&D, Capex, dividends, free cash flow, debt, cash, and net fixed assets 12 measured at the time of IPO, which is on average thirteen years prior to going private. Similarly, as discussed in subsection 2.2., we include the "cell" effects.
Second, in order to mitigate concerns that pre-buyout innovation may affect the results, we construct a sample of matched firms that are similar in terms of pre-innovation characteristics. This should solve any concerns that we are analyzing firms at different timelines in the innovation cycle. We present the results from these two alternative DiD analyses in Table 5. In columns (1) and (2), we present the DiD results when firms are matched on the going-private characteristics; in columns (3) and (4) to 34% in year 2, and from 15% to 25% in year 3.
[ Table 5 here] Third, we verify whether the results are robust to a different matching technique. We match treated and control observations based on the Mahalanobis distance measure on firm characteristics such as industry, size, age, and year. The results are in Table 6. We continue to find a negative and statistically significant effect of public-to-private buyouts on innovation.
The coefficients are also of similar magnitude, i.e., we observe a drop in the number of citations of 19% to 26% in year 2, and up to 27% in year 3.

Institutional and management buyouts
In this subsection, we distinguish between IBOs and MBOs, because we expect institutional investors to have different incentives and long-term objectives than insiders such as firm management. Theoretically, going private in a highly leveraged IBO transaction does not relieve a target firm from short-term pressures. In fact, servicing a huge debt may preclude a firm from realizing long-term investment strategies. In contrast, in an MBO, the insiders may be focused on servicing debt as well as on long-term planning. They may have reputational and career concerns, and, as a result, they may desire to keep the firm in solid shape after returning it to the public sector. We present the summary statistics when we divide the sample of buyout into IBOs and MBOs in Table AII. MBOs have a slightly lower level of innovation measured in the number of patents or citations, yet higher level of radical innovation (measured as number of patents granted to firm i in year t that have at least one citation to non-patent literature). The also happen to take place in countries with higher investor protection and less developed equity markets. statistically significant effect on absolute or relative citations post-buyout. These results indicate that the negative effect from buyouts is predominantly observed for IBOs, but not for MBOs.

Radical innovation
Thus far, we have analyzed various general measures of innovation. However, the nature of innovation can differ. Certain patents, for example, may refer directly to scientific literature, and therefore may be considered more radical than incremental in nature. Following Griffith and Macartney (2014), we thus define radical innovation as the total number of patents granted to firm i in year t that have at least one citation to non-patent literature (NPL). NPL generally refers to scientific journals, and, therefore, patents making citations to NPL are likely to be new and represent radical innovations. In order to identify the effect of buyouts on radical innovation, we limit our sample to target firms that had at least one radical patent applied for and granted within the period of three years before to three years after the buyout.
The results are in Table 8. We find that the number of radical innovations tends to drop after the buyout transaction. We observe a statistically and economically significant decrease in radical innovation one, two, and three years post-buyout.

Innovation efficiency or short term-payoffs?
In previous subsections, we demonstrated that innovation generally drops after going private.
This may be due to PE firms restructuring R&D departments. We therefore look next at innovator employment changes. If the PE firms are focused on long-term investment projects, we expect them to expand and keep the R&D units operational. Alternatively, if their focus is solely short term, we expect to observe employment reductions in innovator employment.
We create a novel measure of innovation efficiency, computed as the number of patent applications filed and subsequently granted during the year, divided by the number of unique innovators. We consider unique innovators as those listed on the patent application. If the same person is included in multiple applications, we count that person only once. This measure also considers how efficiently a firm uses its R&D team following a buyout. Innovation efficiency can be improved by either increasing the number of patent applications while keeping the size of the R&D team constant, or by producing the same number of patent applications using a smaller R&D team.
The results for innovation efficiency are in Table 9. Similarly to the findings for patent counts, radical patent counts, absolute citations, and relative citations, we find that buyouts have a significantly negative effect on innovation efficiency. The drop in innovation efficiency results from the fact that rate of decrease in innovation is higher than the rate of decrease in the number of unique innovators.
[ Table 9 here] Syndicate size PE investors may form syndicates, which are a consortium of multiple investors that are financing the same portfolio firm. There are several reasons why PE firms may partner to form syndicates. Some concerns exist that such PE partnerships may be colluding to depress prices.
For example, Officer, Ozbas and Sensoy (2010) find an overall 40% "club deal discount." However, Boone and Mulherin (2011) suggest other reasons for consortium formations, such as scale, risk, and bidder expertise. Cumming (2006) suggests that syndication, through better screening and selection of investments, might reduce agency conflicts. PE consortiums may be formed in order to certify deal quality of a highly levered transaction to debtholders (Officer, Ozbas, and Sensoy, 2010). Therefore, as Officer et al. (2010) suggest it may be relatively easier to obtain debt financing and on favorable terms if there are multiple PE firms syndicating the deal. We therefore expect for the deals involving a larger number of PE investors can obtain better terms of financing that will put less pressure on their investment decisions. That would result in less pressure on cutting investments in innovation.
In order to test this hypothesis, we divide our sample into deals that had one or more PE investor. The results are in Table 10. We observe negative and statistically significant effects of public-to-private buyouts on innovation only for deals with only one PE investor.
The coefficients suggest a decline in the number of citations of 36% to 47% in year 2, and 39% to 52% in year 3. We find no statistically significant evidence of public-to-private buyout transactions on innovation for deals backed by a syndicate. This result supports the intuition that syndicates, who have more favorable financing terms, would try to avoid cuts in long-term investments.

Buyouts and the cost of debt
Now that we have shown that innovation drops after going private, the question is: Why do PE firms pay for positive net present value (NPV) projects, and then abandon them? It would be illuminating to examine the underlying reasons for the post-buyout drop in innovation. Deep pocket investors may seem more likely to nurture innovation, because we expect they could better tolerate short-term failure. However, they frequently rely heavily on debt financing. The debt overhang theory of Myers (1977) posits that management of an excessively leveraged firm will forgo positive-NPV projects if the new projects benefit debtholders rather than equityholders.
In general, the buyout transaction is not only related to the change in ownership, it also changes the target firm's capital structure and shifts it toward higher leverage. The buyouts are financed mostly with debt, so as much as 80% of the transaction cost may be debt financing.
At the time of the buyout announcement, the acquirer and the lender have agreed upon the terms and payout structure. However, the debt portion may be a significant burden for the planned restructuring of the target firm during the buyout period. Moreover, financing may be received from multiple debt providers, which makes it more difficult to effect a refinancing (Kaplan andStein, 1993, Demiroglu andJames, 2010;Graham and Leary, 2011;Colla, Ippolito and Wagner, 2012;Axelson et al., 2013). The financing in these highly leveraged transactions is often defined as fixed debt plans. The valuation is made based on the assumption that debt is expected to be a function of time alone agreed at the time of the investment (Cooper and Nyborg, 2018;and Baldwin, 2001a,b). Most prior studies have analyzed the effects of buyouts on investment and productivity in isolation. However, investment and financing decisions are not typically separate. In this subsection, we analyze the effects together. It is critical for an acquirer to negotiate the best debt terms for the buyout transaction. If an acquirer can ex-post lock in deal financing for the subsequent post-buyout years at a lower rate than the one currently experienced by other market participants, it will have a critical investment advantage over the competition. Thus, the effect of the buyout on innovation should be positive. But the reverse also holds and an acquirer locks ex-post in deal financing at a higher rate than the ones for current market participants, this may have a negative effect on investment and innovation.
In order to control for the cost of debt, we include the relative ratio of the initial cost of debt at the time of announcement, and the cost of debt in the first, second, and third years postbuyout, respectively, in the regression analysis. Data on the cost of debt come from FRED Economic Data, and we use the corporate debt yield at the time of announcement at the subgroup country level.
Our results are presented in Table 11. They show that the effect of the relative cost of debt after a buyout transaction negatively affects innovation. In particular, the CD (cost of debt) in year 1 after the buyout (that is, the ratio Interestingly, the effect of the post-buyout years becomes positive in years 1 and 3, suggesting that the effect of the buyout on innovation is dependent on the relative cost of debt at the year of announcement relative to the current cost of debt. We posit that, if an acquirer is able to lock in a lower cost of debt over the duration of the restructuring compared to the postbuyout cost of debt, then the incentives to innovate will be stronger. However, if the current cost of debt is lower than the cost of debt at announcement, the investment in innovation will no longer be lucrative, and the incentives to innovate will decrease.

Robustness Analysis and Limitations
In subsequent untabulated tests we analysed whether the results hold for different subsamples.
We split our sample into public-to-private US and non-US buyouts. The negative results were stronger for the US subsample, yet they also hold for the non-US subsample. We also divided the data into pre-2006 and post-2006 subsamples. Most of the previous studies report the results for the pre-2006 data. We checked whether the specific sub period might be driving the negative results. We find that the results are mostly insignificant for pre-2006 subsample, and they are highly significant for post-2006 subsample. We also collected the data for private-toprivate buyouts. The sample for private-to-private buyouts is much larger than the sample of public-to-private buyouts. We find no significant evidence for the full sample of private-toprivate buyouts that they have any effect on innovation, yet when we split this sample into preand post-2006 we find negative association in year 3 post private-to-private buyout.
The PE funds also might engage in secondary buyouts (SBOs). It has been shown that under pressure PE engage more in SBOs that later on underperform (Arcot et al., 2015). Also, Secondary Management Buyouts (SMBOs) have been found to perform worse than regular MBOs (Zhou et al., 2014;Jelic et al., 2019). We therefore, checked the extent to which our results might be affected by these deals. We only have around 10% of deals that are secondary buyouts. We excluded those transactions and the estimates remain robust.
We are unable to disentangle the effects of board advisory and PE human capital on innovation. Yet, in order to mitigate any concerns that PE characteristics play in our analysis we run a model where we additionally include PE firm fixed effects. The results remain robust.
We also tried if the results are robust to alternative estimation methods, where we correct for serial correlation. We therefore applied the linear dynamic panel-data model and included lagged value of the dependent variable. The results are similar to our main results and show negative and statistically significant effect of buyouts on innovation in years two and three post-buyout.
We are also cautious about the fact that our results might overestimate the negative effects of buyout transactions as we only observe what happens in the post-buyout firm. It is possible that divisions or subsidiaries are sold to another firm and the patenting activity continues there.
While our data are not rich enough to analyse remaining research questions we leave them for future research to explore. For example, future studies might explore the effect of PE firm reputation on innovation, use different measures of innovation such as textual analysis that would analyse the patent documents and their disclosure. The future studies might also analyse in details to what happens with the divisions and subsidiaries that are sold post-buyout transaction. Furthermore, in untabulated analysis we find that there is not much evidence of a drop in innovation for a sample of private-to-private deals. We suggest that future research can focus on the determinants of those differences and compare the debt levels between public-toprivate versus private-to-private deals.

Conclusion
This paper explores the impact of public-to-private buyout transactions on the innovation of target firms. We analyze both quantity (patent count) and quality (citations) of patent activity.
We find that, following public-to-private buyouts, firms tend to have fewer patents overall, and to receive fewer citations on those patents. We also show that firms have fewer radical (e.g., scientific) innovations. We observe that the negative effect of public-to-private buyouts is only significant for institutional buyouts. We identify a significant decrease in innovation efficiency post-going private. We also show that the negative effect is most prevalent for transactions where the cost of the debt financing is higher than the post-buyout cost of debt.
Our results add to the previous literature, but also contrast with Lerner, Sorensen and Strömberg (2011) and Amess, Stiebale and Wright (2016), who show innovation increases after buyout transactions. However, their results are mostly driven by private-to-private transactions.
Our study contributes by showing contrasting results for public-to-private transactions. The evidence is based on a buyout sample and matched sample analysis.  This table presents the sample construction, and the distribution of sample by announcement year (panel A) and target industry (panel B), for deals announced from 1997 to 2011 with at least one patent granted to the target firm for the three years before to the three years after the transaction.   (1) and (2) present OLS panel regressions for before-after analysis, with the dependent variable ln(1+number of patents) in models (1) and (2). In model (1), we include industry, firm, and country-year fixed effects. Standard errors are clustered by firm. Column (2) presents difference-in-differences regression results. For each firm in the going-private sample, we include up to five public firms (based on data availability) that are matched to the going-private firms by year, size, three-digit industry, and age. In all models, we show the regression where the independent variables are the relative years pre-and post-buyout (event year 0 is the omitted base category, with a coefficient normalized to 1). All variables are defined in the Appendix. ***, **, and * denote 1%, 5%, and 10% significance levels, respectively.
Before-After DiD (1) (2) Coeff.   (1) and (2) present OLS panel regressions for before-after analysis, with the dependent variables ln(1+absolute citations) in model (1) and ln(1+relative citations) in model (2). In models (1) and (2), we include industry, firm, and country-year fixed effects. Standard errors are clustered by firm. Columns (3) and (4) present difference-in-differences regression results, with the dependent variables ln(1+absolute citations) in model (3) and ln(1+relative citations) in model (4). For each firm in the going-private sample, we include up to five public firms (based on data availability) that are matched to the going-private firms by year, size, three-digit industry, and age. In all models, we show the regression where the independent variables are the relative years pre-and post-buyout (event year 0 is the omitted base category, with a coefficient normalized to 1). All variables are defined in the Appendix. ***, **, and * denote 1%, 5%, and 10% significance levels, respectively.   (1) and (3), and ln(1+relative citations) in models (2) and (4). In columns (1) and (2) for each firm in the going-private sample, we include one public firm (based on data availability) that is matched to the going-private firms on variables that determine the propensity of going private measured for the prior thirteen years: total assets, sales, R&D, Capex, dividends, free cash flow, debt, cash, net fixed assets (we replace any missing values with industry-year averages). In columns (3) and (4) for each firm in the going-private sample, we include one public firm (based on data availability) that is matched to the going-private firms on pre-innovation measures: year, size, three-digit industry, and age. Standard errors are clustered by firm. In all models, we show the regression where the independent variables are the relative years pre-and post-buyout (event year 0 is the omitted base category, with a coefficient normalized to 1). All variables are defined in the Appendix. ***, **, and * denote 1%, 5%, and 10% significance levels, respectively.   (1) and ln(1+relative citations) in model (2). In columns (1) and (2) for each firm in the going-private sample, we include one public firm (based on data availability) that are matched to the going-private firms based on the Mahalanobis distance measure on firm characteristics such as industry, size, age, and year. Standard errors are clustered by firm. In all models, we show the regression where the independent variables are the relative years pre-and post-buyout (event year 0 is the omitted base category, with a coefficient normalized to 1). All variables are defined in the Appendix. ***, **, and * denote 1%, 5%, and 10% significance levels, respectively.   (1) and (2) present OLS panel regressions for before-after analysis, with the dependent variables ln(1+absolute citations) in model (1) and ln(1+relative citations) in model (2). In models (1) and (2), we include industry, firm, and country-year fixed effects. Standard errors are clustered by firm. Columns (3) and (4) present difference-in-differences regression results, with the dependent variables ln(1+absolute citations) in model (3), and ln(1+relative citations) in model (4). For each firm in the going-private sample, we include up to five public firms (based on data availability) that are matched to the going-private firms by year, size, three-digit industry, and age. In all models, we show the regression where the independent variables are the relative years pre-and post-buyout (event year 0 is the omitted base category, with a coefficient normalized to 1). All variables are defined in the Appendix. ***, **, and * denote 1%, 5%, and 10% significance levels, respectively. In all models, we show the regression where the independent variables are the relative years pre-and post-buyout (event year 0 is the omitted base category, with a coefficient normalized to 1). Standard errors are clustered by firm. All variables are defined in the Appendix. ***, **, and * denote 1%, 5%, and 10% significance levels, respectively.
(1)  In all models, we show the regression where the independent variables are the relative years pre-and post-buyout (event year 0 is the omitted base category, with a coefficient normalized to 1). Standard errors are clustered by firm. All variables are defined in the Appendix. ***, **, and * denote 1%, 5%, and 10% significance levels, respectively.  (1) and (3) and by relative citations in columns (2) and (4). In all models, we show the regression where the independent variables are the relative years pre-and post-buyout (event year 0 is the omitted base category, with a coefficient normalized to 1). Standard errors are clustered by firm. All variables are defined in the Appendix. ***, **, and * denote 1%, 5%, and 10% significance levels, respectively.   (1) and (2), and by relative citations in columns (3) and (4). In all models, we show the regression where the independent variables are the relative years pre-and post-buyout (event year 0 is the omitted base category, with a coefficient normalized to 1). Standard errors are clustered by firm. All variables are defined in the Appendix. ***, **, and * denote 1%, 5%, and 10% significance levels, respectively.