Access to Capital and the IPO Decision: An Analysis of US Private Firms

Finance and Economics Discussion Series Federal Reserve Board, Washington, D.C. ISSN 1936-2854 (Print) ISSN 2767-3898 (Online) Access to Capital and the IPO Decision: An Analysis of US Private Firms Andres Almazan, Nathan Swem, Sheridan Titman, Gregory Weitzner 2025-102 Please cite this paper as: Almazan, Andres, Nathan Swem, Sheridan Titman, and Gregory Weitzner (2025). “Access toCapitalandtheIPODecision: AnAnalysisofUSPrivateFirms,”FinanceandEconomics DiscussionSeries2025-102. Washington: BoardofGovernorsoftheFederalReserveSystem, https://doi.org/10.17016/FEDS.2025.102. NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.

Access to Capital and the IPO Decision: ∗ An Analysis of US Private Firms † ‡ § ¶ Andres Almazan Nathan Swem Sheridan Titman Gregory Weitzner November 2025 Abstract We analyze firms’ IPO decisions using detailed financial data on US private firms. We find that firms with higher external capital needs are more likely to go public. Following the IPO, firms increase their investment and debt issuance, resulting in leverageratiosclosetotheirpre-IPOlevels. Finally,newlypublicfirmsborrowfrom an expanded pool of lenders at improved terms, with a decrease in the within-firm dispersion in banks’ private risk assessments. Our evidence is consistent with firms going public to improve their access to capital, which is facilitated by a reduction in asymmetric information. ∗We thank Jason Abrevaya, Daniel Bias, Adolfo De Motta, Mike Ewens, Slava Fos, Paolo Fulghieri, Charlie Hadlock, Helyoth Hessou, Kose John, Steve Kaplan, Philipp Krueger, Borja Larrain, Zack Liu, Michelle Lowry, Simon Mayer, Amrita Nain, Aaron Pancost, Lubos Pastor, Gordon Phillips, Jay Ritter, Amir Sufi and conference and seminar participants at the Fed Board, FRA, GREFA, University of Houston, McGill, MFA, NBER Fall Corporate Finance Meeting, Oregon Summer Finance Conference, Private Equity Research Symposium, PUC Chile, Texas A&M Young Scholars Consortium, University of Texas Austin, and Virtual Corporate Finance Friday for the helpful comments and discussions. We thank Farah Bernstein, Siddhartha Lewis-Hayre, Pierre Alexandre-Larouche and Lin Yang for excellent research assistance. The views expressed in this paper are solely those of the authors. They do not necessarily reflect the views of the Federal Reserve Board or the Federal Reserve System. †University of Texas at Austin. Email: andres.almazan@mccombs.utexas.edu. ‡Federal Reserve Board. Email: nathan.f.swem@frb.gov. §University of Texas Austin. Email: sheridan.titman@mccombs.utexas.edu. ¶McGill University. Email: gregory.weitzner@mcgill.ca.

1 Introduction Improved access to capital is often cited as a primary motive for firms going public.1 Intuitively, IPOs increase firms’ transparency, thereby reducing information asymmetries, which allows firms to raise capital more easily and at a lower cost. However, empirical support for this rationale is mixed,2 and even if access to capital was an important motive for going public in the past, the rapid growth of private capital markets in recent years raisesthequestionofwhetherthispresumedadvantageofpublicmarketsisstillrelevant.3 This paper provides evidence that improved access to capital, driven by reductions in asymmetric information, is a key motive for firms going public. Our analysis uses the Federal Reserve Y-14Q data, which includes all corporate loans over one million dollars extended by large US bank holding companies from 2012 onward. This data is uniquely suited to examine the access to capital motive for two reasons. First, it contains extensive financial information on private firms in the US—by far the largest IPO market in the world—including balance sheet and income statement information and granular information on firms’ bank loans. Second, the data contains banks’ internal risk assessments of borrowers, which, as we describe in more detail below, allow us to examine how both firms’ cost of capital and the degree of asymmetric information change after the IPO. Our main hypothesis is based on the idea that public firms are more transparent and hence less subject to informational asymmetries than private firms and, consequently, face fewer adverse selection and hold-up problems when they raise capital. The increased transparency is both due to public firms being subject to stringent disclosure rules and because information is revealed during the security trading process.4 Motivated by these ideas, we investigate three related issues: i) Are private firms with greater needs for external capital more likely to go public?, ii) do firms gain improved access to capital and increase their investment expenditures after going public?, and if so, iii) is the improved access to capital after the IPO due to a reduction in information asymmetries? We first show that firms that are more reliant on external capital, as proxied by their financing deficit, i.e, (capex - EBITDA)/assets, are more likely to go public in the 1Oneoftheexpressgoalsofthe2012JOBSActwastospurIPOactivitytoimproveaccesstocapital markets (see The JOBS Act: A Landmark Reform to U.S. Securities Laws). 2See Lowry et al. (2017) and Bernstein (2022) for excellent discussions of this issue. 3See Ewens and Farre-Mensa (2020), who show that the deregulation of securities laws has led to an increase in the supply of private capital to late-stage private startups. 4There is a large theoretical literature providing reasons why reducing information asymmetries can improve a firm’s access to capital. These include reductions in adverse selection costs (e.g., Stiglitz and Weiss(1981)andMyersandMajluf(1984))andhold-upproblems(e.g.,Sharpe(1990)andRajan(1992)). In addition, the information reflected in the stock prices of public firms can improve their investment decisions (e.g., Subrahmanyam and Titman (1999)). Other papers that analyze how differences in the information environment of public and private firms influence the IPO choice include Holmstr¨om and Tirole (1993), Chemmanur (1993), Pagano and R¨oell (1998), and Chemmanur and Fulghieri (1999). 1

future. Specifically, we find that a one standard deviation increase in a firm’s financing deficit increases its likelihood of going public by 73%. In addition, when we separately analyze the two components of the financing deficit, we find that ex-ante investment, i.e, capex/assets, positively predicts future IPOs, while profitability, i.e., EBITDA/assets, negatively predicts future IPOs, where the relationship between ex-ante investment and going public is stronger for less profitable firms.5 We next analyze firms’ investment and financing choices before and after they go public by matching IPO firms with comparable firms that remain private. Compared to matched control firms, IPO firms’ capex and total assets increase by over 40% four years following the IPO, with growth observed in both tangible and intangible assets. This post-IPO asset growth is not just financed with the influx of equity capital from the IPO, but is largely funded by increases in debt. Moreover, although the leverage ratios of newly public firms initially drop after going public, they are not significantly different than those of control firms four years later. Whydofirmsrebalancetheircapitalstructureaftergoingpublicratherthanbecoming more reliant on equity? One possibility is that going public allows firms to raise debt capital on more favorable terms. Indeed, we find that interest rates drop on firms’ bank debt after they go public. However, the drop in interest rates may be purely driven by reductions in risk. To rule out this possibility, we exploit banks’ private assessments of each loan’s probability of default (PD) and loss given default (LGD), which are available in the Y-14Q data.6 We find that after controlling for these risk assessments, firms’ borrowing costs drop by 38bps after going public, suggesting that the drop in interest rates at least partially reflects an improvement in borrowing terms. Ourexplanationfortheimprovementinbankloantermsisthatthereductionininformation asymmetries following the IPO reduces the information rents informed financiers can extract from the firm.7 Given that firms often borrow from multiple banks at the same time, we can construct a proxy for the degree of asymmetric information based on the within-firm dispersion in banks’ PD assessments. Consistent with a reduction in asymmetric information, we find that PD dispersion drops after the IPO.8 5The link between capital needs and going public was articulated by John Collison, the Stripe Cofounder and President, who recently stated that more profitable firms have less of a need to go public because internally generated cash flows can fund their investments (Stripe in ‘no rush’ to go public as cash flow turns positive). 6Beyhaghi,Fracassi,andWeitzner(2025)showthat1)theseriskassessmentsstronglypredictfuture loan performance and 2) interest rates no longer predict firm performance after controlling for them. 7E.g., Rock (1986), Sharpe (1990) and Rajan (1992). In these models, the adverse selection problem that arises from asymmetric information across financiers causes the increase in firms’ cost of capital. 8Differences of opinion, such as in bond ratings and analyst forecasts, is a common proxy for asymmetricinformation(e.g.,Morgan(2002),FlanneryandKwan(2004),Iannotta(2006)andLivingstonand Zhou (2010)). PD dispersion may also arise from differences in subjective beliefs (e.g., Diether, Malloy, and Scherbina (2002)). However, as we argue in further detail below, if differences in subjective beliefs werethesoledriverofPDdispersion,therewouldbelittlereasonforthisdisagreementtosystematically decrease following an IPO unless asymmetric information is reduced. 2

Our final set of tests examines whether the reduction in asymmetric information after the IPO allows firms to borrow from a broader pool of lenders. First, we document an increase in the number of banks that IPO firms borrow from after going public. Second, we find that IPO firms increase their use of syndicated loans and bonds dramatically after going public.9 These results suggest that going public not only provides firms with additional equity but also facilitates access to syndicated loan and public bond markets, the former of which is consistent with anecdotal evidence that IPO activity is an important determinant of aggregate bank lending.10 The analysis in this paper builds on the literature that uses data on private firms to examine the ex-ante determinants and the ex-post outcomes following firms’ IPO decisions (e.g., Pagano, Panetta, and Zingales (1996), Pagano, Panetta, and Zingales (1998), Fischer (2000), Chemmanur, He, and Nandy (2010), Aslan and Kumar (2011), Gopalan and Gormley (2013) and Maksimovic, Phillips, and Yang (2020)).11 The seminal paper in this literature, Pagano, Panetta, and Zingales (1998), analyzes a sample of private Italian firms from 1982 to 1992 and documents several pieces of evidence inconsistent with the access to capital channel. In particular, they find that more profitable firms are more likely to go public and show that investment, profitability, and leverage drop after the IPO. More recently, Chemmanur, He, and Nandy (2010) and Aslan and Kumar (2011) find a positive relationship between both ex-ante and ex-post investment among samples of private US manufacturing firms and UK firms, respectively.12 Our contribution to this literature is as follows: First, our analysis uses the most detailed data on US private firms in the literature.13 In comparison, many papers use the Census Longitudinal Business Database (LBD), which contains incomplete income 9Syndicated loans involve elements of both traditional, private relationship bank lending and public bondofferings, giventhattheloansareoftenwidelyheldbydispersedinvestors(DennisandMullineaux (2000) and Gadanecz (2004)). 10See US companies going public could lift related bank lending. This complementarity between equity and debt financing is also consistent with Hartman-Glaser, Mayer, and Milbradt (2024), who show that improved access to equity markets increases firms’ debt capacity. 11A related literature compares the behavior and outcomes of public and private firms (e.g., Brav (2009),SaundersandSteffen(2011),KovnerandWei(2014),Asker,Farre-Mensa,andLjungqvist(2015), Gilje and Taillard (2016), Acharya and Xu (2017), Phillips and Sertsios (2017), Maksimovic, Phillips, and Yang (2017), Sheen (2020), Dambra and Gustafson (2021) and Sanati and Spyridopoulos (2023)). Bernstein (2022) and Ewens and Farre-Mensa (2022) survey this literature. Some papers also analyze a small set of private firms in which pre-IPO data is more prevalent (e.g., Lerner (1994), Helwege and Packer (2003), and Aghamolla and Thakor (2022b)). 12In addition, Jain and Kini (1994) document an increase in capital expenditures following IPOs using other public firms as a control group. Kim and Weisbach (2008) show that a large portion of IPO proceeds are used for capex and R&D, and Mikkelson, Partch, and Shah (1997) show that 64% of firms include new investments as a use of proceeds in the IPO prospectus. Lowry (2003) shows that proxies for demand for capital are important determinants of IPO volume at the aggregate level. 13Our data is also more detailed than most foreign data used in this literature. For example, we are not aware of any paper using banks’ internal risk assessments in the IPO literature. 3

statementandbalancesheetinformation, andnoinformationonfirms’borrowingterms.14 Second, we provide evidence from many different angles that ex-ante investment needs predict IPOs and access to capital improves after the IPO, the latter of which we link to reductions in asymmetric information. Third, we introduce several new results to the literature. Specifically, our paper is the first to use banks’ private risk assessments to show that after going public, firms’ borrowing costs drop conditional on the risk of the borrower and this coincides with a decrease in the dispersion in banks’ private credit assessments. Our results that 1) ex-ante profitability negatively predicts going public and 2) that this effect is stronger when ex-ante investment is high are also new and differ from studies using European data (e.g., Pagano, Panetta, and Zingales (1998) and Aslan and Kumar (2011)).15 Finally, our paper is the first to show that, due to large increases in debt, firms’ leverage levels revert to their pre-IPO levels within four years of the IPO and that firms increase their use of market-based debt financing after going public.16 The ex-post part of our analysis relates to the more recent literature that focuses on the causal impact of IPOs on subsequent outcomes. This literature, starting with Bernstein (2015), uses data on firms that file to go public but may ultimately withdraw, instrumenting for the completion decision with market-wide returns (e.g., Babina, Ouimet, and Zarutskie (2020), Borisov, Ellul, and Sevilir (2021), Cornaggia et al. (2021), Cornaggia et al. (2022) and Larrain et al. (2025)). The closest paper to ours in this literature is Larrain et al. (2025), which shows that European firms’ sales and profitability increase after going public, which they argue is facilitated by reductions in financial constraints. Our evidence is complementary to theirs in that we 1) show that firms with higher ex-ante external capital needs are more likely to go public, 2) provide direct evidence of access to capital improving after the IPO, and 3) show that this improved access to capital can be attributed to a reduction in information asymmetries. As we discuss below in more detail, for several reasons, the Bernstein (2015) instrument has limited power during our sample period, precluding us from taking this 14Amongthesepapers,severalusetheCensusofManufacturersandtheAnnualSurveyofManufacturers data, which contains sales and capital expenditures at the plant-level for firms in the manufacturing industry (e.g., Chemmanur, He, and Nandy (2010), Chemmanur and He (2011), Chemmanur et al. (2018),andChemmanuretal.(2022)). However,thedata,whichiscollectedeveryfiveyearsforallfirms and annually for plants with more than 250 employees, does not contain any information about firms’ balance sheets or income statements beyond sales and capital expenditures/stock. In contrast, our data contains a quarterly panel of detailed firm financials for an extremely broad set of private firms. 15OurresultsmaydifferfromPagano,Panetta,andZingales(1998)fortworeasons. First,asPagano, Panetta, and Zingales (1998) note, firms that go public in Italy are much older and more profitable than in the US, suggesting that the capital markets are fundamentally different than those in the US. Second,becauseoursampleismorerecent,thereasonfirmsgopubliccouldhavefundamentallychanged. However, given the recent rise of private capital markets, we would think that, if anything, access to capital would be less important for public firms than it was 30 years ago. 16Whilesomepapersanalyzefirms’firstbondissuances(e.g.,Datta,Iskandar-Datta,andPatel(2000) and Hale and Santos (2008)), we are the first to analyze how bond issuance evolves after the IPO using a set of control firms that remain private. 4

approach.17 However, while it is clearly important to isolate the treatment effects of IPOs, selection effects (i.e., which types of firms choose to IPO) are also interesting and important. For example, if the IPO results in a reduction in the cost of external capital, whichwefindsupportforinouranalysis, firmswillinvestmoreasaresultofgoingpublic, but also will be more likely to go public when they expect to invest more in the future. As a result, our results capture both of these important effects. Nonetheless, several of our findings regarding the access to capital channel are difficult to explain through selection alone. For example, if the convergence in bank risk assessments induces firms to go public, we would expect this to occur prior to the IPO; however, the convergence only occurs after the IPO. Similarly, we find that firms receive improved terms on their bank debt after going public; however, selection effects would predict that firms should be less likely to go public if they anticipate an improvement in terms in the future, regardless of their listing status.18 Finally, there are other studies that find evidence of public firms obtaining lower interestratesonbankdebtthanprivatefirms(e.g., Pagano, Panetta, andZingales(1998), Schenone (2010) and Saunders and Steffen (2011)). Schenone (2010) finds that firms’ borrowing costs decrease after they go public19 and Saunders and Steffen (2011) show that public firms borrow at lower average interest rates than private firms in the UK. While some of our evidence is consistent with these studies, our analysis differs in several key respects. First, by controlling for firms’ underlying risk, as perceived by their lenders, we show that the decrease in borrowing costs is not simply due to a reduction in firm risk. Second, we document a drop in the dispersion in banks’ risk assessments after the IPO, providing evidence for a reduction in asymmetric information as a potential mechanism for the improvement in borrowing terms. Third, we show that firms increase their usage of bonds and syndicated loans after the IPO.20 2 Data Inthissection, wedescribeoursourceofdataonprivatefirms, IPOs, andtheconstruction of our firm-level and loan-level samples. 17For example, our sample period occurs after the 2012 JOBS Act, which allowed firms to file and withdraw the IPO confidentially, potentially reducing the number of withdrawn IPOs observable to researchers (Dambra, Field, and Gustafson (2015), Boeh and Dunbar (2021) and Bias (2021)). 18We also find that firms increase their use of market-based debt financing after the IPO, but not before. 19Duetoalackofdataonprivatefirms,thisstudydoesnotcompareborrowingcoststocontrolfirms, but rather analyzes how borrowing costs change among IPO firms only. 20Additionally, in contrast to Schenone (2010) and Saunders and Steffen (2011), our data allows us to analyze how firms’ financing structure changes after firms go public, relative to similar control firms that remain private. 5

2.1 Sample of Private Firms OurmainsourceofdataisScheduleH.1oftheFederalReserve’sY-14Qdata. TheFederal Reserve began collecting the Y-14 data in 2011 to support the Dodd-Frank-mandated stress tests and the Comprehensive Capital Analysis and Review (CCAR). The sample includes corporate loans from all bank holding companies (BHCs) with $50bn or more in total assets, accounting for 85.9% of all assets in the US banking sector as of 2018:Q4 (Frame, McLemore, and Mihov (2020)). Qualified BHCs are required to report detailed quarterly loan-level data on all corporate loans that exceed one million dollars in size. These loans constitute over 97% of these BHCs’ corporate exposure (Beyhaghi (2022)) and represent about 70% of all commercial and industrial loan volume in the US extended by BHCs that file a FR Y-9C report (Bidder, Krainer, and Shapiro (2020)). Our sample of private firms starts in 2012, when borrower financial data became fully populated. We apply several filters that are consistent with other papers that use the Y-14 data (e.g., Gustafson, Ivanov, and Meisenzahl (2020) and Beyhaghi, Fracassi, and Weitzner (2025)). Specifically, we drop firms with missing taxpayer identification numbers (TINs), firms headquartered outside the US, firms with loans denominated in foreign currencies, borrowers that appear to be high-net-worth individuals, financial firms (NAICS code 52), real estate firms (NAICS code 92), and public administration and government entities (NAICS code 53).21 We also drop firms with less than $10mm in assets because these firms are unlikely to go public. Additionally, because we are interested in comparing private firms that go public to those that remain private, we remove all public firms that are never private in our sample period. We identify public firms, as well as their subsidiaries, in the Y-14 data using a multi-step process similar to that of Beyhaghi et al. (2024), which we discuss in detail in Appendix A.2. Many firms in the Y-14 data borrow from multiple banks in a given quarter, and potentially have multiple loans from those banks. Hence, we take the median financial record across each loan within each firm-quarter.22 After we aggregate the loan-level data to the firm-level, we have a quarterly panel with over one million firm/quarter observations with over just over 100,000 unique private firms. 2.2 IPO Firms We obtain US IPO data from SDC Platinum from LSEG Data and Analytics. Following Bernstein (2015), we exclude financial firms (SIC codes between 6000 and 6999), unit trusts, closed-end funds, real estate investment trusts (REITs), American depositary 21See Appendix A for additional details. We also make several minor adjustments to the firm’s financial data, which we describe in Appendix A.1. 22Firm financials often differ across banks, given that there are no standard reporting requirements for private firms. 6

receipts (ADRs), limited partnerships, special purpose acquisition vehicles (SPACs), and spin-offs. Thesefiltersresultin1,390uniquefirmsintheSDCdatathatgopublicbetween 2012 and 2023. To merge the IPO data into the Y-14 dataset, we use a multi-step matching process, which is described in further detail in Appendix A.3. Using this process, we identify 423 unique IPOs (a 30% match rate) in the Y-14 data within a three-year window prior to the IPO.23 Our data does not include the entire universe of IPOs. Some IPO firms are not in our data because 1) they do not have bank debt, 2) if they have bank debt, it is not from one of the BHCs in the Y-14, or 3) the size of the loan is below $1mm.24 Although our match rate is less than half on an absolute basis, we match the majority of larger IPOs. Specifically, Appendix Table C1, shows that the matched IPOs account for 61% of the aggregate IPO proceeds, suggesting that our sample of IPOs tends to be larger private firms.25 Appendix Table C2 compares IPO firm-quarters for IPOs in our sample to those outside our sample, using pre-IPO financial data from Compustat. IPOs in our sample are indeed much larger in terms of both sales and assets. Table C1 also shows that our sample of IPOs is not fully representative of all IPOs over this period in terms of industries. For example, there are over 500 pharma and biotech IPOs over this period and only 37 in our sample. However, this industry effect appears to be entirely explained by firm size. Appendix Table C3 includes regressions where the dependent variable is a dummy that equals one if the IPO firm is in our sample. Column (1) includes dummy variables for the ten largest industries in the overall IPO sample. Consistent with Table C1, the coefficients for pharma and biotech are negative andstatisticallysignificant, andtheF-statsuggeststhattheindustrydummiesarejointly significant. However, in column (2), when we include Log(1+Sales) as a control variable, pharma and biotech are much smaller in magnitude and no longer statistically significant. Moreover, the F-stat is below one, suggesting that the difference in industry composition in our sample can be fully explained by firm size. That our sample consists of relatively larger private firms that borrow from large banks suggests that these firms already have relatively good access to capital and, thus, lower needs for going public. In this respect, this sample selection likely works against the access to capital mechanism we explore in our analysis. 23For comparison, Maksimovic, Phillips, and Yang (2020) identify 48% of the firms from Jay Ritter’s IPOdataintheUSCensusdata. ThisislikelyduetotheCensusdatacontainingmoresmallfirmsthan theY-14data. Wechooseathree-yearwindowgiventhatourex-antetestsexaminewhichcharacteristics predict firms going public over a three-year window. 24Many of these firms enter our dataset after they go public. Indeed, we show below that firms expand the number of lenders they borrow from after going public, increasing the likelihood that firms are present in the Y-14 data. Consistent with this, Beyhaghi, Howes, and Weitzner (2022) show that over half of publicly traded firms are in the Y-14 data. 25Our sample includes 63% of IPOs larger than $250mm and 74% of IPOs larger than $500mm in proceeds. 7

2.3 Firm-Level Panel We use the Y-14 data to construct a quarterly panel of private firms from 2012 to 2023, which we supplement with data on VC financing from the Preqin VC funding database.26 We merge the Preqin data with the Y-14 data using industry, location, and firm name via the FedMatch text string matching algorithm (Cohen et al. (2021)).27 We use similar matching methods to identify private firms that are acquired based on the SDC Platinum mergers and acquisitions dataset and to identify the amount of bonds outstanding each firm has using the Mergent FISD dataset.28 Appendix Tables C4 and C5 display the industry and location composition of IPO firms in our sample, and Appendix Tables C6 and C7 display the industry and location composition of the other private firms in our sample. While the firms that go public represent a variety of industries and are located in a variety of cities, they are clustered in technology-related industries, and a large number are located in the San Francisco- Oakland-Hayward and Palo Alto-San Jose core-based statistical areas (CBSAs), which includes Silicon Valley. By contrast, the top industries for the broader sample of private firms tend to be consumer retail-related, such as auto dealers and restaurants, with locations more aligned with the overall US population. We define a firm’s IPO quarter as the latest quarter in which we observe Y-14 data within the one-year window preceding the IPO. We also create the dummy variable IPO, which equals one if the firm goes public in the next three years. We construct several standard corporate finance variables (e.g., investment and profitability).29 To minimize theimpactofoutliers,wewinsorizevariablesthatareratiosatthe1%and99%. Appendix B contains detailed definitions of the variables used throughout the paper. Table 1 includes summary statistics that compare financial information for IPO firmquarters to non-IPO firm-quarters. Unsurprisingly, IPO firms are larger in terms of assets andsalesthanthebroadersampleofprivatefirms. Moreover, IPOfirmsinvestmore(0.08 versus 0.05 Capex/Assets) and are less profitable (0.08 versus 0.16 EBITDA/Assets). As we discuss later, an important variable in our analysis is the financing deficit, which is defined as (Capex-EBITDA)/Assets, and measures the amount of external financing a firm needs beyond what it generates internally. This variable is standard in the literature (e.g., Frank and Goyal (2003)); however, we exclude some smaller components which are notavailableoraswellpopulatedintheY-14data, suchaschangesinnetworkingcapital. 26For some empirical tests we include post-IPO observations of recently private firms that have gone public during our sample period. 27The Preqin VC funding database includes many types of private equity investments (e.g., angel investments, seed financing, series A, etc). To be defined as a VC in Preqin, the investment firm must take a minority stake in the target firm. We refer to all of these deals as VC investments. 28We describe the details of the Preqin, SDC, and FISD merges in Appendix A.3. 29The latest instructions detailing what qualifying banks are required to report can be found here: FR Y-14 Instructions. 8

On average, IPO firms have a financing deficit close to zero, while the average financing deficit of non-IPO firms is significantly negative (-0.12), i.e., the average firm that stays private internally generates more cash flows than it needs to fund its investment. IPO firms are also more likely to be VC-backed than non-IPO firms (29% versus 2%) and are much more likely to be located in Silicon Valley (14% versus 2%) and operate in technology and life science industries (23% versus 6%).30 2.4 Loan-Level Sample We also construct a loan-level sample for our tests that examine changes in bank lending terms. To do this, we merge the firm-level balance sheet, income statement, cash flow, location, public status, IPO status, and private financing characteristics from our panel of private firms with the respective firm’s specific loan-level records from the Y-14 data, which contain the terms of each loan at origination. In addition to basic information about the loan, such as its size, interest rate, maturity and syndication status, the data contains each bank’s private assessments of the borrower’s probability of default (PD) and the loan’s expected loss given default (LGD). Specifically, PD is a long-run, borrower-level annual default rate, and LGD is a long-run expected loss per dollar of exposure in default. According to the Basel Committee on Banking Supervision, internal estimates of PD and LGD “must incorporate all relevant, material and available data, information and methods. A bank may utilize internal data and data from external sources (including pooled data).”31 Moreover, banks must update these estimates regularly and immediately after any material changes: “Borrowers and facilities must have their ratings refreshed at least on an annual basis... In addition, banks must initiate a new rating if material information on the borrower or facility comes to light.” Recent work has also shown that these risk assessments strongly predict future realized default (Beyhaghi, Fracassi, and Weitzner (2025) and Weitzner and Howes (2025)) and public equity and bond returns (Beyhaghi, Howes, and Weitzner (2022)). Following Beyhaghi, Fracassi, and Weitzner (2025), we apply several additional filters to the loan-level data. Specifically, we exclude observations where the interest rate is zero or negative. We also drop observations in which the PD or LGD is missing, zero, or greater than one. Loan records can appear in the data for multiple quarters so long as the loan remains on the lending bank’s balance sheet, but because we are interested in the terms at origination, we only keep newly originated loans. Table 2 contains summary statistics for the loan-level sample. 30See Appendix B for the list of NAICS codes we consider technology and life science firms, which closely follows Chemmanur et al. (2022). 31The most recent instructions for calculating these risk assessments are available at Calculation of RWA for credit risk. 9

3 Empirical Analysis Our empirical analysis is divided into five parts. In Section 3.1, we test which characteristics predict the decision to IPO in the cross-section of private firms. In Section 3.2, we examine the dynamics of firm outcomes before and after the IPO based on a matched sample of firms that remain private. In Section 3.3, we examine how firms’ borrowing costs drop after the IPO. In Section 3.4, we analyze how asymmetric information changes after the IPO. In Section 3.5, we explore how the composition of firms’ debt changes after they go public. Finally, in Section 3.6, we discuss the extent to which our findings capture both treatment and selection effects. 3.1 Which Characteristics Predict IPOs? We first examine which ex-ante characteristics predict whether firms go public. To do this, we estimate the following linear probability model: IPO = ΓX +βMTB +δ +u , (1) i,t+1:t+12 i,t j,t t i,t+1:t+12 where i, j, and t index firm, industry, and quarterly date, respectively. The dependent variable IPO is a dummy variable that equals one if the firm goes pubi,t+1:t+12 lic within the next twelve quarters, which we multiply by 100, and X is a vector of i,t firm-level predictors. The primary variable of interest is the financing deficit ((Capex- EBITDA)/Assets), which proxies for the firms’ expected external financing needs. Under the access to capital channel, we expect that firms with higher financing deficits are more likely to go public. Our other predictors, which closely follow those in Pagano, Panetta, and Zingales (1998), include: 1) sales (log(1+Sales)), as there are likely large fixed costs associated withgoingpublic, makingtheIPOmoreattractiveforlargerfirms(e.g., Ritter(1987)); 2) sales growth, because faster growing firms have higher market valuations; 3) debt/assets, as firms with higher leverage may have less financial flexibility and find it more difficult to raise external capital; and 4) the industry-level market-to-book ratio, which, as Pagano, Panetta, andZingales(1998)discuss, canproxyforeitherbetterinvestmentopportunities or mispricing in the industry. To control for time-specific shocks, we also include date fixed effects (δ ); however, in some specifications, we also include date by industry, date t by CBSA, and date by industry by CBSA fixed effects. We cluster our standard errors by firm.32 Column (1) of Table 3 displays the estimated coefficients of (1) with date fixed effects. First, we find a statistically significant relationship between firms’ propensity to IPO and 32Thestandarderrorsareverysimilarthroughouttheentireanalysisifwedoubleclusterbyfirmand date. 10

their size. Specifically, a 10% increase in sales increases the likelihood of a firm going public by 12% from its base rate of 0.20%. Second, firms with higher trailing oneyear sales growth are also more likely to IPO. A one standard deviation increase in sales growth (0.42) is associated with a just over 100% increase in the likelihood of a firm going public. These effects are consistent with larger, faster-growing firms being more likely to go public. Third, firms with higher leverage are more likely to go public. However, this effect is a bit smaller on a relative basis: a one standard deviation increase in leverage (0.26) is associated with a 10% increase in IPO likelihood, and the coefficient is only statistically significant in the specifications with more saturated fixed effects. Fourth, firms with higher financing deficits are more likely to go public. Specifically, a one-standard deviation increase in Financing Deficit (0.23), is associated with a 73% increaseinlikelihoodofafirmgoingpublic. Thisresultsuggeststhatfirmswithhigherexternalcapital needsare morelikelyto gopublic. Moreover, aftercontrolling forFinancing Deficit, the coefficient on investment (Capex/Assets) is not significant. This implies that in an equivalent regression in which investment and profitability are treated as separate predictors, 1) investment positively predicts IPOs and profitability negatively predicts IPOs, and 2) we cannot reject the null hypothesis that the coefficients on profitability and investment are equal in magnitude.33 The fact that profitability negatively predicts firms going public is the opposite of Pagano, Panetta, and Zingales (1998), who estimate similar regressions on a sample of privateItalianfirms,34 andisconsistentwithrecentevidencethatfirmsdelaygoingpublic when their profits increase, allowing them to generate cash flows internally. For example, the Stripe Co-founder and President, John Collison, recently stated that more profitable firms do not need to go public because internally generated cash flows can fund their investments.35 This also mirrors evidence from LBOs, where firms that go private tend to generate higher cash flows internally and have lower investment needs (e.g., Opler and Titman (1993)). Finally, we find that the industry-level market-to-book ratio has a positive relationship with the propensity to go public, consistent with firms with greater investment opportunities being more likely to go public. In columns (2), (3), and (4) of Table 3, we estimate the same regressions but include industry by date fixed effects, CBSA by date fixed effects, and industry by CBSA by date fixed effects, respectively.36 Across these alternative specifications with more restrictive fixed effects, the coefficients remain fairly similar except for the coefficient of leverage, 33SinceFinancingDeficitisequaltoCapex/Assets-EBITDA/Assets,thecoefficientforCapex/Assets tests whether Capex/Assets and EBITDA/Assets are different in magnitude. 34Aslan and Kumar (2011) also find the same result in a sample of private firms in the UK. 35See Stripe in ‘no rush’ to go public as cash flow turns positive. 36Column (2) has more observations than column (1) because it does not include the industry-level market-to-book ratio, which is not available for a few industries in the Y-14 data. 11

which becomes more than three times as large and statistically significant once we control for industry effects.37 Taken together, our initial results suggest that firms with higher external capital needs and better investment opportunities are more likely to go public. If less profitable firms go public due to a lack of internal funds to finance investment, we expect this effect to be stronger for firms with more investment needs. Intuitively, under the access to capital motive, internally generated cash flows should only matter to the extent that firms have substantive investment needs to begin with. To test this hypothesis, we re-estimate a similar regression as Table 3, but include EBITDA/Assets and Capex/Assets separately, as well as an additional interaction term between them. As shown in Table 4, the interaction coefficient is negative and statistically significant across all specifications. These results suggest that the relationship between ex-ante investment and going public is even stronger for firms that generate fewer internal cash flows, and are thus more reliant on external capital. Hence, these results provide further evidence that firms are more likely to go public when they have higher external capital needs. 3.1.1 Technology and Life Science Subsample Analysis One might expect that this isn’t important for tech and biotech firms. First, a large portion of our IPOs are these types of firms. Second our effects are larger. Third, these firms actually engage in Capex. In Appendix Table C8, we reestimate the same regressions, but instead of restricting the sample to VC-backed firms, we limit the sample to technology and life science firms. 3.1.2 VC-Backed Subsample Analysis If firms with higher external capital needs are more likely to go public, one might expect this effect to be weaker for VC-backed firms with access to private capital markets. Moreover, VC-backed firms tend to have proprietary technologies, which can more easily be expropriated when these firms are public.38 On the other hand, firms that seek VC financing may require more capital and be more subject to asymmetric information to begin with; hence, when their external capital needs are high, they may find public markets particularly attractive. In Table 5, we re-estimate the same regressions from Table 3 but only include firms that we identify as VC-backed. Across all specifications, the coefficient estimates for Financing Deficit are larger. These results suggest that even firms with access to private 37In unreported results, we also find qualitatively similar results if we estimate a probit regression with date fixed effects. 38E.g., Campbell (1979), Bhattacharya and Ritter (1983), Yosha (1995), Dambra, Field, and Gustafson (2015), Farre-Mensa (2017), Aghamolla and Thakor (2022a), Davydova et al. (2022) and Bennett, Stulz, and Wang (2023). 12

equity capital go public when their external capital needs are high.39 Moreover, the fact that the magnitudes are even larger than those in our baseline tests is consistent with firms that have VC backing being more subject to asymmetric information, and hence, the benefit of being public increases more with their external capital needs.40 3.1.3 Robustness Tests In this section, we conduct various robustness tests for our main results in Table 3. Oneconcernisthatsomeofthesmallerprivatefirmsinoursamplethatremainprivate are acquired or otherwise drop out of the sample much earlier than the firms that IPO. In Appendix Table C11, we show that our results are robust to excluding firms that exit via acquisition. In addition, in Appendix Table C12 we show that our results are robust to excluding all firms that exit the sample for any reason within three years. Another concern is that linear independent variables and fixed effects are not sufficient to control for the potentially large differences between IPO and non-IPO firms. In particular, Table 1 shows that IPO firms are, on average, almost ten times as large (in terms of assets) as firms that do not go public. In Appendix Table C13, we estimate the same regressions in Table 3, but interact the existing fixed effects with an additional asset decile fixed effect to compare firms of similar sizes and find very similar qualitative results. In addition, we show that our results are robust to including all firms smaller than $10 million in assets (Appendix Table C14) and excluding all firms smaller than $50 million in assets (Appendix Table C15). Although we include industry/date fixed effects in our regressions, firms within industries may still not be completely comparable, particularly for the high-tech firms, which comprise a high share of the IPO firms in our sample. One concern could be that these “high-tech” firms that IPO tend also to have higher financing deficits. For example, many biotech firms whose drugs have not been approved by the FDA have zero revenue before going public. In Appendix Table C16, we show that the main results in Table 3 are robust to excluding all technology and life science firms and firms located in Silicon 39ThepointestimatesforFinancingDeficitarelargerthanourbaselineresultsinTable3,andwefind evenlargereffectsinAppendixTableC9, wherewerestrictthesampletoVC-backedtechnologyandlife science firms. This latter result suggests that VC-backing is not simply picking up “tech” effects, but rather VC-backing has an independent relationship with external capital needs and the IPO decision. 40The main results also hold if we reestimate Table 3, but control for whether the firm is VC-backed (Appendix Table C10). Although our analysis below suggests that the increased transparency benefits firms in terms of their ability to raise capital, we do not claim that the costs of increased transparency are irrelevant. To the extent that transparency is costly for certain firms, we would expect those firms to be less likely to go public. Second, not all information disclosure affects asymmetric information and expropriation risk similarly. For example, public firms are required to report audited financial information regularly. While audited financial information clearly reduces asymmetric information, it is not obvious that it increases the potential of expropriation. Similarly, a technology company may not reveal the exact details of its proprietary algorithm or production methods in an IPO. However, this lack of disclosure need not result in substantial asymmetric information. Relatedly, Boone, Floros, and Johnson (2016) show that firms often redact proprietary information in their IPO prospectuses. 13

Valley. Similarly, our results are robust to controlling for firm fixed effects that absorb firm-specific differences not captured by industry (Appendix Table C17). 3.2 How Do Firm Outcomes Evolve after the IPO? Afteranalyzingwhichfirmcharacteristicspredictfirms’decisionstoIPOinthefuture, we nowexaminehowfirmoutcomesevolveaftertheIPO.Specifically, weperformananalysis in which we match IPO firms to control firms in the last quarter available in the year prior to the IPO. We form a matched sample by estimating propensity scores based on our ex-ante regression (1) with date by industry by CBSA fixed effects, but only including IPO firms’ last quarter in the year prior to the IPO. The matched sample includes the three closest firms in terms of propensity scores while requiring exact matching based on their two-digit NAICS industry and VC-backing status. After identifying cohorts of treated and matched control firms, we employ a cohort generalized difference-in-differences strategy using a window of 3 years prior to the IPO up to 4 years after the IPO. Specifically, we analyze the difference in outcome Y for i,c,t each treated firm i after the IPO relative to before and compare it with the difference in outcome of its matched control firms within the same cohort c using the following regression: 4 (cid:88) Y = β (d ×λ )+α +δ +u , i,c,t k i,c y,k,c i,c t,c i,c,t k=−3 where d is a dummy that equals one if the firm is treated (i.e., is an IPO firm), λ is i,c y,k,c a dummy equal to one if year y is equal to k and zero otherwise, where k = 0 is the year of the IPO. We include firm cohort fixed effects α to compare the change in outcome i,c within the same firm. We include time-cohort fixed effect δ to ensure that the IPO firm t,c is compared only with the matched control firms at each point in time. Standard errors are again clustered by firm. For each regression, we plot the time series of coefficients, i.e., β , with 90% confik dence intervals, while omitting the year prior to the IPO, i.e., β as the reference point. −1 We estimate annual rather than quarterly coefficients to obtain more precise estimates; however, because the year of the IPO may also contain quarters prior to the IPO, the effect is often smaller in year zero than in years in which all quarters occur after the IPO. First, we examine the dynamics of firms’ capex (in logs) around the IPO. Figure 1 showsthatIPOfirms’capexincreasesdramaticallyaftertheIPOascomparedtomatched firms that do not go public. Specifically, capex jumps after the IPO and increases by a statistically significant 46% relative to matched non-IPO firms four years after the IPO. This increase in investment translates into higher total assets. In Figure 2, we plot total assets (in logs) and find that IPO firms’ assets increase 40% more than control firms four years after IPO. 14

While capital expenditures are clearly an important form of investment, certain firms, particularly technology-related ones, also invest in intangible assets such as R&D. Although we do not have data on R&D and intangible investment specifically, we can back out total intangible assets based on the firms’ total assets and tangible assets, which are both available in the Y-14 data.41 In Figure 3 we plot the time series of coefficients for intangible assets (in logs) and find that IPO firms’ intangible assets increase 23% more than matched non-IPO firms four years after IPO. Given that firms dramatically increase their assets and investment after going public, an obvious question is how firms finance this investment. Is it purely financed through new equity, or do firms use the IPO to facilitate non-equity capital raises? To answer this question, we first analyze how firms’ debt levels evolve after the IPO. Figure 4 shows that IPOfirmsincreasetheirdebtaftertheIPO,wherebyyear4, theamountofdebttheyhave increases by about 65% relative to control firms. Moreover, Figure 5 shows that in the first year, there is a statistically significant drop in leverage of about 4 percentage points. However, after year one, leverage reverts back such that in years 2, 3 and 4, there is no statistically significant difference between IPO firms and their matched counterparts, with the coefficient even positive by year 4. This result differs from Pagano, Panetta, and Zingales (1998) and Aslan and Kumar (2011), who find permanent reductions in leverage after IPOs in Italy and the UK, respectively. 3.3 Going Public and Bank Borrowing Costs In Section 3.2, we find that firms do not simply issue equity after they go public. Rather, they finance their asset growth and investments with debt, such that their leverage is unchanged four years after the IPO. In this section, we analyze how the terms of the debt that firms issue change after they go public. In particular, going public may allow firms to raise debt capital on more favorable terms. This can occur if the decrease in information asymmetry following the IPO reduces the amount of information rents informed investors can extract (e.g., Rock (1986), Sharpe (1990), Rajan (1992), Kurlat (2016), and Beyhaghi, Fracassi, and Weitzner (2025)). An empirical challenge to testing for such an improvement in borrowing terms is that firms’ risk levels also change after the IPO. For example, firms business models may fundamentally change around IPOs as they move from the innovation to exploitation stage. This could result in banks being less risky. Hence, it is not sufficient to simply show that interest rates decline after firms go public. Fortunately, the availability of banks’ internal risk assessments (PD and LGD) in the Y-14 data allows us to make this distinction among firms’ bank loans. As mentioned 41Tangible assets in the Y-14 data encompass any assets that have a physical existence, including cash. 15

earlier, these risk assessments strongly predict future realized default and public equity and bond returns.42 Moreover, Beyhaghi, Fracassi, and Weitzner (2025) show that after controlling for these risk assessments, interest rates no longer predict default at all, consistent with the risk assessments being sufficient statistics for the underlying risk of the borrower. Hence, we follow their approach and test how interest rates change after controlling for the banks’ assessed risk of their underlying loans. For these tests, we use loan-level data and restrict the sample to newly issued loans. Because we are analyzing new loans, there are not enough observations to do the same matched sample analysis as before; however, as we argue above, the main unobserved confounding variable is the credit risk of the borrower, which our data allows us to observe. To test for an improvement in bank loan terms after the IPO, we estimate the following regression: IR = β (IPO ×Post )+Γ X +Γ Z +β PD +β LGD j 0 i t 0 i,t 1 j 1 j 2 j +β (PD ×LGD )+α +δ +u , 3 j j i,b ,t j where IR is the interest rate on a new loan j from bank b to firm i in quarter t. As j independent variables, we include the same vector of firm-level controls as in Section 3.1 (X ), a vector of loan-level controls (Z ), which include log(maturity), log(amount) and i,t j facility type fixed effects,43 as well as banks’ internal risk assessments: probability of default (PD) and loss given default (LGD) and their interaction (expected loss). The variable of interest is IPO ×Post , which represents the change in firm i’s borrowing cost i t after going public, controlling for bank b’s change in the perceived risk of the firm. We also include bank by firm fixed effects α to control for any time-invariant relationshipi,b specific effect on borrowing costs. The results are displayed in Table 6. In column (1), we estimate the regression without loan-level controls, bank risk assessments, or bank by year-quarter fixed effects. The estimated coefficient −0.511, is statistically significant, showing that credit spreads decrease by 51bps after firms go public. We find lower magnitudes when we include loanlevel controls in column (2) and add bank by date fixed effects in column (3). Finally, in column (4), we also include bank risk assessments. The expected loss is positively related to the loan’s interest rate; however, the coefficient for IPO×Post also remains negative and large in magnitude (-0.384). This 38bp drop in borrowing costs compares to an all-in average interest rate of around 401bps and a credit spread of 183bps (compared to the average 10-year treasury rate) for IPO firms prior to going public. Hence, credit spreads drop by 21%, even after controlling for the underlying risk of the firm, as perceived by 42E.g., see Weitzner and Howes (2025), Beyhaghi, Howes, and Weitzner (2022), and Beyhaghi, Fracassi, and Weitzner (2025). 43See Instructions for the Capital Assessments and Stress Testing Information Collection for the list of facility types in the data. 16

the bank.44 These results suggest that borrowing from banks becomes more attractive after firms go public. The most plausible mechanism behind this channel is that by increasing their transparency after going public, information asymmetry drops. This reduction in asymmetric information allows firms to borrow from more banks and at a lower cost, as banks can extract fewer information rents from public firms.45 In the next section, we provide direct evidence for a reduction in asymmetric information after the IPO. 3.4 Going Public and Asymmetric Information Our next set of tests examines whether going public reduces information asymmetries. For these tests, we create a proxy for the degree of asymmetric information based on the within-firm dispersion in banks’ PD assessments. Intuitively, if there is less asymmetric information, banks’ beliefs should more closely coincide with each other. This approach is in line with the literature that uses split bond ratings (e.g., Morgan (2002), Iannotta (2006) and Livingston and Zhou (2010)) and analyst dispersion (e.g., Flannery and Kwan (2004)) as proxies for asymmetric information.46 Our measure has the advantage of incorporating the private information of multiple sophisticated financial institutions that have direct financial incentives to accurately assess borrower risk.47 Our measure of dispersion is the cross-sectional standard deviation of PD estimates across banks within each firm-quarter. For this analysis, we employ a different matching approach than in our previous tests. Because the level of PD and the number of banks are likely correlated with the cross-sectional standard deviation of PD estimates, we match based on the Mahalanobis distance measure using PD and the number of banks as non-exact matching variables, while, as before, requiring exact matching based on their two-digit NAICS industry and VC-backing. This approach allows us to create a wellmatched control group specifically for analyzing the differences in PD dispersion between IPO and non-IPO firms. Figure 6 shows that, compared to matched private firms, the cross-sectional standard deviationinbanks’PDestimatesdecreasessignificantlyafterfirmsgopublic. Thisdecline begins immediately after the IPO and persists through the four-year post-IPO period we analyze. By year four, the dispersion in PD estimates for IPO firms is 6 percentage points 44A potential concern with analyzing banks’ risk assessments is that banks may misrepresent them (e.g., Plosser and Santos (2018) and Behn, Haselmann, and Vig (2022)). However, the inclusion of bank by time fixed effects absorbs any effects of banks’ incentives to misreport at the aggregate level. 45SeealsoBird, Karolyi,andRuchti(2019), whoshowthatinformationsharingreducesinterestrates on bank debt. 46Usingtheseotherproxiesisnotpossiblegiventhatprivatefirmsrarelyhavecreditratingsoranalyst coverage prior to going public. 47Forexample, thefactthatcreditandequityresearchanalystsarenotpaiddirectlyfortheaccuracy of their ratings, but rather reputational concerns, can lead to dishonest reporting (e.g., Ottaviani and Sørensen (2006)). 17

lower than for matched control firms that remain private, which represents close to one standard deviation (6.4 percentage points). To ensure that our results are not driven by changes in the composition of lending banks after the IPO, we conduct the same analysis while fixing the set of banks for each firm throughout the sample period (Appendix Figure C1). This modified test shows a similar decrease in PD dispersion after the IPO.48 These results, which indicate that information asymmetry declines after firms go public, provide a potential mechanism for why newly public firms can borrow at better terms. Finally, while we only observe the dispersion in bank investors’ beliefs, the post-IPO convergence in banks’ risk assessments likely reflects improvements in the broader information environment that benefit the firm’s ability to raise capital from all types of investors. A relevant concern is that firms’ business models may fundamentally change when they go public. For example, firms may be at the stage of commercialization rather than innovation (e.g., Rajan (2012), Bernstein (2015), Larrain et al. (2025), Bernstein (2022)). This could lead to a reduction in dispersion in risk assessments, not because of new information being revealed from the IPO, but rather because the fundamentals are more easily assessed. However, if this were the case, we would not expect it to happen immediately after the IPO. To the extent the business is changing, we should observe a slower movement in PD dispersion beginning prior to the IPO.49 3.5 Going Public and Firms’ Debt Composition In Section 3.4, we show that after firms go public, the dispersion in banks’ private risk assessmentsdrops,whichweattributetoareductioninasymmetricinformation. Reduced information asymmetries not only can affect the terms of firms’ debt, but also the types of debt firms issue. In this section, we explore how firms’ debt composition changes after they go public. First, reduced information asymmetries can enable firms to borrow from a broader pool of banks as the adverse selection problem is lessened. To test this hypothesis, we use the same matched time-series regressions as in Section 3.2 and plot the estimated coefficients using the number of banks the firm borrows from as the dependent variable. As shown in Figure 7, four years after going public, IPO firms borrow from just under one more bank relative to control firms, starting from a baseline average of 2.3 banks. In the Appendix, we also estimate the regression using a fixed-effect Poisson model (e.g., Cohn, Liu, and Wardlaw (2022)) and find very similar results. 48It is possible that the dispersion in banks’ PDs reflects differences of opinion, i.e., disagreement, ratherthandifferencesininformation. However,intheory,lowerdisagreementleadstolowerprices(e.g., Miller (1977)), whereas in contrast, we observe higher prices, i.e., improved terms on bank debt, after a reduction in PD dispersion. 49If this process occurs after the IPO, then we would expect to see a slow drop in dispersion after the IPO, not the large drop we see right away. 18

Reduced information asymmetries may also facilitate the use of market-based debt financing, such as public bonds and syndicated loans. For example, Dunn (2025) states:50 “In addition to the capital raised in an IPO, going public provides the company with access to the public capital markets for future financings, including for debt and hybrid securities.” Similarly,DennisandMullineaux(2000)arguethatfirmsaremorelikelytousesyndicated loans when information about the borrower is more transparent.51 Consistent with this idea, firms often receive new syndicated credit facilities from banks after going public.52 Based on this motivation, we test whether firms increase their usage of syndicated debt and public bonds after going public. Figure 8 analyzes the change in the ratio of syndicated debt to total debt, i.e, Syndicated Debt/Debt.53 Prior to the IPO, there is no evident pre-trend in the use of syndicated loans; however, after going public, the proportion of IPO firms’ debt increases by about 15 percentage points relative to peers that remain private. This represents a 167% increase relative to the baseline ratio of syndicated debt to total debt of 9% among IPO firms.54 We next examine changes in the use of public bonds after the IPO by creating a variable, Bonds/Debt, which is the ratio of total bonds outstanding, obtained from FISD, over total debt. Prior to going public, IPO firms’ bonds make up 3% of their total debt. Figure 9 shows that after going public, IPO firms experience a large increase in the proportion of bonds in their debt, with an almost 8 percentage points increase by year four. Moreover, this increase only occurs after the IPO and not before. Taken together, these results suggest that firms’ mix of debt becomes more concentrated in market-based 50Moreover, Florou and Kosi (2015) find that after an increase in the transparency of financial statements, firms are more likely to issue public bonds than borrow privately from banks, and Kovner and Wei (2014) finds that public bond spreads are higher when they are issued by private firms. 51As discussed in Dennis and Mullineaux (2000) and Gadanecz (2004), syndicated loans involve elements of both traditional, private relationship bank lending and public bond offerings, given that the loans are often widely held by dispersed investors. In recent years, syndicated loans are more often securitized and sold to CLOs and hedge funds (Bhardwaj and Mukherjee (2022)). 52See US companies going public could lift related bank lending. 53To calculate the total amount of syndicated loans on borrower firms’ balance sheets, we divide the loan size of each syndicated loan by the lender’s share of the loan, which is recorded in Y-14. We then identify loans across different banks in the same syndicate if the loans are classified as syndicated, have the same origination date, the same maturity date, and the same loan type. We then calculate the median of the adjusted size across the loan observations identified as being within the same syndicate. We then sum up the total amount of syndicated debt across all syndicated loans. The benefit of this approach is that it accounts for syndicated debt not held by Y-14 banks, as long as at least one Y-14 bank holds it. 54These results may lead one to wonder whether our results regarding the drop in interest rates are entirely due to the increase in loan syndication (e.g., Table 6). However, in Appendix Table C18, we reestimate the regressions, but control for whether the loan is syndicated and find almost identical results. Similarly, Appendix Figure C3 shows that our results on PD dispersion are similar, albeit with less power, when we only include non-syndicated loans in the measure of PD dispersion. 19

debt after going public, which we argue is facilitated by the increased transparency and resulting reduction in asymmetric information following the IPO. 3.6 Treatment Versus Selection Effects The analysis in this paper explores 1) the relationship between ex-ante characteristics and the IPO decision, and 2) changes in firm outcomes following their IPOs. While the first tests explicitly examine selection effects (i.e., which firms choose to go public), the second set of tests captures both treatment (i.e., the causal impact of the IPO on firm outcomes) and selection effects. For example, the reason that we observe an increase in investment after the IPO is almost certainly due to both treatment and selection effects. If the IPO leads to a reduction in the cost of external capital for firms, as we find evidence for, firms should invest more after going public. However, firms that anticipate investing more in the future should also find it more advantageous to go public, which we argue is a critical reason why firms go public. Asmentionedintheintroduction, thereisagrowingliteraturewithempiricaltestsdesigned to isolate the treatment effect of IPOs on firm outcomes. Specifically, these papers consider samples of firms that announce IPOs, some of which are completed and others of which are withdrawn. Because the decision to complete an IPO is endogenous, these papers instrument for IPO completion using plausibly exogenous market-wide stock returns (Bernstein (2015)) after the IPO filing date. Unfortunately, this instrument has limited power in our sample, which precludes us from using it in our time-series analysis for three reasons. First, our sample begins in 2012, giving us a relatively short sample period. Second, the 2012 JOBS Act allowed firms to file and withdraw their IPOs confidentially, potentially reducing the number of withdrawn IPOs observable to researchers (Dambra, Field, and Gustafson (2015), Boeh and Dunbar (2021), and Bias (2021)). Third, as mentioned above, we only have a subsample of completed (or withdrawn) IPOs over this period. We do believe, however, that several of our findings regarding the change in access to capital and asymmetric information that we document are difficult to explain through selection alone. For example, it is not obvious why a firm would go public in anticipation of a convergence in bank risk assessments, and if it did, we would expect this convergence to occur before the IPO; however, we observe no such convergence prior to the IPO. For the same reason, this makes it unlikely that firms are going public in response to becoming more transparent.55 Similarly, if firms anticipate an improvement in terms of their bank debt in the future, we would expect these firms to be less likely to IPO. In contrast, we find that these firms are more likely to IPO. Hence, we argue it is difficult 55This also makes it hard to argue that firms are using more market-based debt financing for reasons other than being public, given that we only observe this behavior after the IPO. 20

to explain the improvement in bank debt terms purely through selection alone. 3.7 Alternative Mechanisms Our paper does not claim other motives are not important. Rather it shows that access to capital is important. Brau and Fawcett (2006) provide several other motives. Lubos pastor paragraph. Firms file S-1. Stock price trading. Standardization story does not mean firms don’t go public for need of capital Barry and Mihov (2015) find that 11 percent of IPO firms have no debt. This excludes convertible bonds and preferred stock. 4 Conclusion An often-cited reason firms go public is to improve their access to capital. However, in recent years, private capital markets have expanded substantially, potentially reducing this presumed benefit of public markets. In this paper, we provide evidence that, despite thistrend, improvedaccesstocapitalisanimportantmotiveforwhyfirmsgopublic. Our evidence also suggests that at least part of this improvement in access to capital is due to the reduction in asymmetric information that arises from the increase in transparency following IPOs. We show that private firms with higher financing deficits (i.e., lower profitability coupled with elevated investment needs) are more likely to go public. After going public, these firms increase their investments in both tangible and intangible assets relative to a matched sample of firms that remain private. In addition to the equity raised in the IPO, newly public firms expand their use of debt, resulting in leverage ratios that are similar to those of control firms that remain private four years after the IPO. Consistent with a reduction in adverse selection, we show that firms’ borrowing costs, conditional on their risk, decline after the IPO. We also find evidence of reduced information asymmetries, i.e., the dispersion in banks’ private risk assessments drops after going public. Finally, we find evidence that going public facilitates access to syndicated loan and public bond markets. While our analysis focuses on the cross-section of firms’ IPO decisions, our results may also provide insights into aggregate IPO activity. In particular, the access to capital channel we study may also help explain the well-documented decline in both the number ofIPOsandpublicfirmsintheUSoverthepast25years.56 Whiletherehavebeenseveral 56E.g., Gao, Ritter, and Zhu (2013), Doidge, Karolyi, and Stulz (2013), Doidge, Karolyi, and Stulz (2017), Ewens and Farre-Mensa (2022), Chemmanur et al. (2022), and Doidge et al. (2025). 21

explanations for this phenomenon,57 our results suggest that when external capital needs decrease, fewerfirmswillgopublic. Indeed, Covarrubias, Guti´errez, andPhilippon(2020) show that over this period, investment decreased and profitability increased, implying a decrease in external capital needs, which may have contributed to the decline in IPOs. Finally, while we focus on IPO decisions, the richness of the Y-14 data that we use allows for many potentially interesting analyses related to private firms’ financing decisions. For example, future work could analyze in more detail the types of private firms that are acquired. This issue is closely related to the analysis in this paper, given that being acquired is an alternate way in which firms gain access to public capital markets. 57Theseinclude1)increasedregulatoryandcomplianceburdensonpublicfirms(e.g., WeildandKim (2010) and Ewens, Xiao, and Xu (2024)), 2) M&A activity (Eckbo and Lithell (2025)), 3) increases in economiesofscope(e.g.,Gao,Ritter,andZhu(2013)),and4)thederegulationofprivatecapitalmarkets (e.g., Ewens and Farre-Mensa (2020)). 22

References Acharya, Viral and Zhaoxia Xu, 2017, Financial dependence and innovation: The case of public versus private firms, Journal of Financial Economics 124, 223–243. Aghamolla, Cyrus and Richard T Thakor, 2022a, Do mandatory disclosure requirements for private firms increase the propensity of going public?, Journal of Accounting Research 60, 755–804. Aghamolla, Cyrus and Richard T Thakor, 2022b, Ipo peer effects, Journal of Financial Economics 144, 206–226. Asker, John, Joan Farre-Mensa, and Alexander Ljungqvist, 2015, Corporate investment and stock market listing: A puzzle?, The Review of Financial Studies 28, 342–390. Aslan, Hadiye and Praveen Kumar, 2011, Lemons or cherries? growth opportunities and market temptations in going public and private, Journal of financial and Quantitative Analysis 46, 489–526. Babina, Tania, Paige Ouimet, and Rebecca Zarutskie, 2020, Ipos, human capital, and labor reallocation, Available at SSRN 2692845 . Badertscher, Brad A, Dan Givoly, Sharon P Katz, and Hanna Lee, 2019, Private ownership and the cost of public debt: Evidence from the bond market, Management Science 65, 301–326. Barry, Christopher B and Vassil T Mihov, 2015, Debt financing, venture capital, and the performance of initial public offerings, Journal of Banking & Finance 58, 144–165. Behn, Markus, Rainer Haselmann, and Vikrant Vig, 2022, The limits of model-based regulation, The Journal of Finance 77, 1635–1684. Bennett, Benjamin, Ren´e M Stulz, and Zexi Wang, 2023, Does greater public scrutiny hurt a firm’s performance?, Working paper, National Bureau of Economic Research. Bernstein, Shai, 2015, Does going public affect innovation?, The Journal of finance 70, 1365–1403. Bernstein, Shai, 2022, The effects of public and private equity markets on firm behavior, Annual Review of Financial Economics 14, 295–318. Beyhaghi, Mehdi, 2022, Third-party credit guarantees and the cost of debt: Evidence from corporate loans, Review of Finance 26, 287–317. 23

Beyhaghi, Mehdi, Cesare Fracassi, and Gregory Weitzner, 2025, Adverse selection in corporate loan markets, Journal of Finance forthcoming. Beyhaghi, Mehdi, Murray Z Frank, Ping McLemore, and Ali Sanati, 2024, Monetary policy and corporate investment: The equity financing channel, Available at SSRN 4793393 . Beyhaghi, Mehdi, Cooper Howes, and Gregory Weitzner, 2022, The information advantage of banks: Evidence from their private credit assessments, Working Paper . Bhardwaj, Abhishek and Saptarshi Mukherjee, 2022, Do lending relationships matter when loans are securitized?, Available at SSRN 4360620 . Bhattacharya,SudiptoandJayRRitter,1983,Innovationandcommunication: Signalling with partial disclosure, The Review of Economic Studies 50, 331–346. Bias, Daniel, 2021, Illiquid equity, labor mobility, and talent allocation, Swedish House of Finance Research Paper . Bidder, Rhys M, John R Krainer, and Adam Hale Shapiro, 2020, De-leveraging or derisking? how banks cope with loss, Review of Economic Dynamics . Bird, Andrew, Stephen A Karolyi, and Thomas G Ruchti, 2019, Information sharing, holdup, and external finance: Evidence from private firms, The Review of Financial Studies 32, 3075–3104. Boeh, Kevin K and Craig G Dunbar, 2021, Raising capital after ipo withdrawal, Journal of Corporate Finance 69, 102020. Boone, Audra L, Ioannis V Floros, and Shane A Johnson, 2016, Redacting proprietary informationattheinitialpublicoffering, Journal of Financial Economics 120, 102–123. Borisov, Alexander, Andrew Ellul, and Merih Sevilir, 2021, Access to public capital markets and employment growth, Journal of Financial Economics 141, 896–918. Brau, JamesCandStanleyEFawcett, 2006, Initialpublicofferings: Ananalysisoftheory and practice, The journal of Finance 61, 399–436. Brav, Omer, 2009, Access to capital, capital structure, and the funding of the firm, The journal of finance 64, 263–308. Campbell, Tim S, 1979, Optimal investment financing decisions and the value of confidentiality, Journal of Financial and Quantitative analysis 14, 913–924. Chemmanur, Thomas J, 1993, The pricing of initial public offerings: A dynamic model with information production, The Journal of Finance 48, 285–304. 24

Chemmanur, Thomas J and Paolo Fulghieri, 1999, A theory of the going-public decision, The review of financial studies 12, 249–279. Chemmanur, Thomas J and Jie He, 2011, Ipo waves, product market competition, and the going public decision: Theory and evidence, Journal of Financial Economics 101, 382–412. Chemmanur, Thomas J, Jie He, Shan He, and Debarshi Nandy, 2018, Product market characteristics and the choice between ipos and acquisitions, Journal of Financial and Quantitative Analysis 53, 681–721. Chemmanur, Thomas J, Jie He, Xiao Ren, and Tao Shu, 2022, The disappearing ipo puzzle: New insights from proprietary us census data on private firms, Available at SSRN 3556993 . Chemmanur,ThomasJ,ShanHe,andDebarshiKNandy,2010,Thegoing-publicdecision and the product market, The Review of Financial Studies 23, 1855–1908. Cohen, Gregory J, Jacob Dice, Melanie Friedrichs, Kamran Gupta, William Hayes, Isabel Kitschelt, Seung Jung Lee, W Blake Marsh, Nathan Mislang, Maya Shaton et al., 2021, Theussyndicatedloanmarket: Matchingdata, Journal of Financial Research 44, 695– 723. Cohn, Jonathan B, Zack Liu, and Malcolm I Wardlaw, 2022, Count (and count-like) data in finance, Journal of Financial Economics 146, 529–551. Cornaggia, Jess, Matthew Gustafson, Jason D Kotter, and Kevin Pisciotta, 2021, Does being private constrain geographic expansion?, Available at SSRN 3756590 . Cornaggia, Jess, Matthew Gustafson, Jason D Kotter, and Kevin Pisciotta, 2022, Initial public offerings and the local economy, Available at SSRN 3036176 . Covarrubias, Matias, Germa´n Guti´errez, and Thomas Philippon, 2020, From good to bad concentration? us industries over the past 30 years, NBER Macroeconomics Annual 34, 1–46. Dambra, Michael, Laura Casares Field, and Matthew T Gustafson, 2015, The jobs act and ipo volume: Evidence that disclosure costs affect the ipo decision, Journal of Financial Economics 116, 121–143. Dambra, MichaelandMatthewGustafson, 2021, Dotheburdenstobeingpublicaffectthe investment and innovation of newly public firms?, Management Science 67, 594–616. Datta, Sudip, Mai Iskandar-Datta, and Ajay Patel, 2000, Some evidence on the uniqueness of initial public debt offerings, The Journal of Finance 55, 715–743. 25

Davydova, Daria, Ru¨diger Fahlenbrach, Leandro Sanz, and Ren´e M Stulz, 2022, Why do startups become unicorns instead of going public?, Working paper, National Bureau of Economic Research. Dennis, Steven A and Donald J Mullineaux, 2000, Syndicated loans, Journal of financial intermediation 9, 404–426. Diether, Karl B, Christopher J Malloy, and Anna Scherbina, 2002, Differences of opinion and the cross section of stock returns, The journal of finance 57, 2113–2141. Doidge, Craig, G Andrew Karolyi, and Ren´e M Stulz, 2013, The us left behind? financial globalization and the rise of ipos outside the us, Journal of Financial Economics 110, 546–573. Doidge, Craig, G Andrew Karolyi, and Ren´e M Stulz, 2017, The us listing gap, Journal of Financial Economics 123, 464–487. Doidge, Craig, George Andrew Karolyi, Kris Shen, and Ren´e M Stulz, 2025, Are there too few publicly listed firms in the us?, Financial Review . Dunn, Gibson, 2025, Ipo handbook . Eckbo, B Espen and Markus Lithell, 2025, Merger-driven listing dynamics, Journal of Financial and Quantitative Analysis 60, 209–257. Ewens, Michael and Joan Farre-Mensa, 2020, The deregulation of the private equity markets and the decline in ipos, The Review of Financial Studies 33, 5463–5509. Ewens, Michael and Joan Farre-Mensa, 2022, Private or public equity? the evolving entrepreneurial finance landscape, Annual Review of Financial Economics 14, 271– 293. Ewens, Michael, Kairong Xiao, and Ting Xu, 2024, Regulatory costs of being public: Evidence from bunching estimation, Journal of Financial Economics 153, 103775. Farre-Mensa, Joan, 2017, The benefits of selective disclosure: Evidence from private firms . Fischer, Christoph, 2000, Why do companies go public? empirical evidence from germany’s neuer markt, Empirical Evidence from Germany’s Neuer Markt (May 2000) . Flannery, Mark J and Simon H Kwan, 2004, Market evidence on the opaqueness of banking firms’ assets, Journal of Financial Economics 71, 419–460. 26

Florou, Annita and Urska Kosi, 2015, Does mandatory ifrs adoption facilitate debt financing?, Review of Accounting Studies 20, 1407–1456. Frame, W Scott, Ping McLemore, and Atanas Mihov, 2020, Haste makes waste: Banking organization growth and operational risk . Frank, Murray Z and Vidhan K Goyal, 2003, Testing the pecking order theory of capital structure, Journal of financial economics 67, 217–248. Gadanecz, Blaise, 2004, The syndicated loan market: structure, development and implications, BIS Quarterly Review, December . Gao, Xiaohui, Jay R Ritter, and Zhongyan Zhu, 2013, Where have all the ipos gone?, Journal of Financial and Quantitative Analysis 48, 1663–1692. Gilje, Erik P and Jerome P Taillard, 2016, Do private firms invest differently than public firms? taking cues from the natural gas industry, The Journal of Finance 71, 1733– 1778. Gopalan, Radhakrishnan and Todd A Gormley, 2013, Do public equity markets matter in emerging economies? evidence from india, Review of Finance 17, 1571–1615. Gustafson, Matthew, Ivan Ivanov, and Ralf R Meisenzahl, 2020, Bank monitoring: Evidence from syndicated loans, Available at SSRN 2831455 . Hale, Galina and Joa˜o AC Santos, 2008, The decision to first enter the public bond market: The role of firm reputation, funding choices, and bank relationships, Journal of Banking & Finance 32, 1928–1940. Hartman-Glaser, Barney, Simon Mayer, and Konstantin Milbradt, 2024, A theory of asset-and cash flow-based financing, Working paper. Helwege, Jean and Frank Packer, The decision to go public: evidence from mandatory SEC filings of private firms (Citeseer 2003). Holmstro¨m, Bengt and Jean Tirole, 1993, Market liquidity and performance monitoring, Journal of Political economy 101, 678–709. Iannotta, Giuliano, 2006, Testing for opaqueness in the european banking industry: evidence from bond credit ratings, Journal of Financial Services Research 30, 287–309. Jain, Bharat A and Omesh Kini, 1994, The post-issue operating performance of ipo firms, The journal of finance 49, 1699–1726. Kim, Woojin and Michael S Weisbach, 2008, Motivations for public equity offers: An international perspective, Journal of Financial Economics 87, 281–307. 27

Kovner, Anna and Chenyang Jason Wei, 2014, The private premium in public bonds, FRB of New York Staff Report . Kurlat, Pablo, 2016, Asset markets with heterogeneous information, Econometrica 84, 33–85. Larrain, Borja, Gordon M Phillips, Giorgo Sertsios, and Francisco Urzu´a I, 2025, The effects of going public on firm profitability and strategy, The Review of financial studies hhaf026. Larrain, Borja, Giorgo Sertsios, and Francisco Urzu´a, 2021, The going public decision of business group firms, Journal of Corporate Finance 66, 101819. Lerner, Joshua, 1994, Venture capitalists and the decision to go public, Journal of financial Economics 35, 293–316. Livingston, Miles and Lei Zhou, 2010, Split bond ratings and information opacity premiums, Financial Management 39, 515–532. Lowry, Michelle, 2003, Why does ipo volume fluctuate so much?, Journal of Financial economics 67, 3–40. Lowry, Michelle, Roni Michaely, Ekaterina Volkova et al., 2017, Initial public offerings: A synthesis of the literature and directions for future research, Foundations and Trends® in Finance 11, 154–320. Maksimovic, Vojislav, Gordon M Phillips, and Liu Yang, 2017, Do public firms respond to investment opportunities more than private firms? the impact of initial firm quality, Working paper, National Bureau of Economic Research. Maksimovic, Vojislav, Gordon M Phillips, and Liu Yang, 2020, Are public firms really myopic? evidence from matching ipo firms at birth, Evidence from Matching IPO Firms at Birth (March 1, 2020). Tuck School of Business Working Paper . Mikkelson, Wayne H, M Megan Partch, and Kshitij Shah, 1997, Ownership and operating performance of companies that go public, Journal of financial economics 44, 281–307. Miller, Edward M, 1977, Risk, uncertainty, and divergence of opinion, The Journal of finance 32, 1151–1168. Morgan, Donald P, 2002, Rating banks: Risk and uncertainty in an opaque industry, American Economic Review 92, 874–888. Myers, Stewart C and Nicholas S Majluf, 1984, Corporate financing and investment decisions when firms have information that investors do not have, Journal of financial economics 13, 187–221. 28

Opler, Tim and Sheridan Titman, 1993, The determinants of leveraged buyout activity: Free cash flow vs. financial distress costs, The journal of finance 48, 1985–1999. Ottaviani, Marco and Peter Norman Sørensen, 2006, The strategy of professional forecasting, Journal of Financial Economics 81, 441–466. Pagano, Marco, Fabio Panetta, and Luigi Zingales, 1996, The stock market as a source of capital: Some lessons from initial public offerings in italy, European Economic Review 40, 1057–1069. Pagano, Marco, Fabio Panetta, and Luigi Zingales, 1998, Why do companies go public? an empirical analysis, The Journal of Finance 53, 27–64. Pagano, Marco and Ailsa Ro¨ell, 1998, The choice of stock ownership structure: Agency costs, monitoring, and the decision to go public, The Quarterly Journal of Economics 113, 187–225. Phillips, Gordon M and Giorgo Sertsios, 2017, Financing and new product decisions of private and publicly traded firms, The Review of Financial Studies 30, 1744–1789. Plosser, Matthew C and Joao AC Santos, 2018, Banks’ incentives and inconsistent risk models, The Review of Financial Studies 31, 2080–2112. Rajan, Raghuram G, 1992, Insiders and outsiders: The choice between informed and arm’s-length debt, The Journal of finance 47, 1367–1400. Rajan, Raghuram G, 2012, Presidential address: The corporation in finance, The Journal of finance 67, 1173–1217. Ritter, JayR,1987, Thecostsofgoingpublic, Journal of financial economics 19, 269–281. Rock, Kevin, 1986, Why new issues are underpriced, Journal of financial economics 15, 187–212. Sanati, Ali and Ioannis Spyridopoulos, 2023, Comparing capital allocation efficiency in public and private equity markets, Available at SSRN 4557395 . Saunders, Anthony and Sascha Steffen, 2011, The costs of being private: Evidence from the loan market, The Review of Financial Studies 24, 4091–4122. Schenone, Carola, 2010, Lending relationships and information rents: Do banks exploit their information advantages?, The Review of Financial Studies 23, 1149–1199. Sharpe, Steven A, 1990, Asymmetric information, bank lending, and implicit contracts: A stylized model of customer relationships, The journal of finance 45, 1069–1087. 29

Sheen, Albert, 2020, Do public and private firms behave differently? an examination of investment in the chemical industry, Journal of Financial and Quantitative Analysis 55, 2530–2554. Stiglitz, Joseph E and Andrew Weiss, 1981, Credit rationing in markets with imperfect information, The American economic review 71, 393–410. Subrahmanyam, Avanidhar and Sheridan Titman, 1999, The going-public decision and the development of financial markets, The Journal of Finance 54, 1045–1082. Weild, David and Edward Kim, 2010, Market structure is causing the ipo crisis—and more, Capital Market Series, Grant Thornton . Weitzner, Gregory and Cooper Howes, 2025, Bank information production over the business cycle, Review of Economic Studies forthcoming. Yosha, Oved, 1995, Information disclosure costs and the choice of financing source, Journal of Financial intermediation 4, 3–20. 30

Figure 1: Investment Dynamics Note: Inthisfigure,weanalyzethedynamicsoffirminvestment,i.e.,log(1+Capex),beforeand after the IPO using a matched sample. We form a matched sample by estimating propensity scores based on our ex-ante regression (1) with date by industry by CBSA fixed effects, but only including IPO firms’ last quarter in the year prior to the IPO. The matched sample includes the three closest firms in terms of propensity scores that are in the same two-digit NAICS industry and have the same VC-backing status. The dots are point estimates of the interaction coefficients between treated (IPO firms) and year dummies, and the bars are the 90% confidence intervals from the following regression: 4 (cid:88) Log(1+Capex) = β (d ×λ )+α +δ +u , i,c,t k i,c y,k,c i,c t,c i,c,t k=−3 where i, c, and t index firm, cohort (matched group), and time, respectively, d is a dummy i,c that equals one if the firm is treated (i.e., is an IPO firm). λ is a dummy equal to one if y,k,c year y is equal to k and zero otherwise, where k =0 is the year in which the IPO occurs, α i,c are firm/cohort fixed and δ are time/cohort fixed effects. Year zero contains the quarter in t,c which the firm IPOs. Standard errors are clustered by firm. 31

Figure 2: Asset Dynamics Note: In this figure, we analyze the dynamics of firm assets, i.e., log(assets), before and after the IPO using a matched sample. We form a matched sample by estimating propensity scores based on our ex-ante regression (1) with date by industry by CBSA fixed effects, but only including IPO firms’ last quarter in the year prior to the IPO. The matched sample includes the three closest firms in terms of propensity scores that are in the same two-digit NAICS industry and have the same VC-backing status. The dots are point estimates of the interaction coefficients between treated (IPO firms) and year dummies, and the bars are the 90% confidence intervals from the following regression: 4 (cid:88) Log(Assets) = β (d ×λ )+α +δ +u , i,c,t k i,c y,k,c i,c t,c i,c,t k=−3 where i, c, and t index firm, cohort (matched group), and time, respectively, d is a dummy i,c that equals one if the firm is treated (i.e., is an IPO firm). λ is a dummy equal to one if y,k,c year y is equal to k and zero otherwise, where k =0 is the year in which the IPO occurs, α i,c are firm/cohort fixed and δ are time/cohort fixed effects. Year zero contains the quarter in t,c which the firm IPOs. Standard errors are clustered by firm. 32

Figure 3: Intangible Assets Dynamics Note: In this figure, we analyze the dynamics of intangible assets, i.e., log(intangible assets), before and after the IPO using a matched sample. We form a matched sample by estimating propensity scores based on our ex-ante regression (1) with date by industry by CBSA fixed effects, but only including IPO firms’ last quarter in the year prior to the IPO. The matched sample includes the three closest firms in terms of propensity scores that are in the same twodigit NAICS industry and have the same VC-backing status. The dots are point estimates of the interaction coefficients between treated (IPO firms) and year dummies, and the bars are the 90% confidence intervals from the following regression: 4 (cid:88) Log(Intangible Assets) = β (d ×λ )+α +δ +u , i,c,t k i,c y,k,c i,c t,c i,c,t k=−3 where i, c, and t index firm, cohort (matched group), and time, respectively, d is a dummy i,c that equals one if the firm is treated (i.e., is an IPO firm). λ is a dummy equal to one if y,k,c year y is equal to k and zero otherwise, where k =0 is the year in which the IPO occurs, α i,c are firm/cohort fixed and δ are time/cohort fixed effects. Year zero contains the quarter in t,c which the firm IPOs. Standard errors are clustered by firm. 33

Figure 4: Debt Dynamics Note: Inthisfigure,weanalyzethedynamicsoftotaldebt(inlogs),beforeandaftertheIPO, using a matched sample. We form a matched sample by estimating propensity scores based on our ex-ante regression (1) with date by industry by CBSA fixed effects, but only including IPO firms’ last quarter in the year prior to the IPO. The matched sample includes the three closest firms in terms of propensity scores that are in the same two-digit NAICS industry and have the same VC-backing status. The dots are point estimates of the interaction coefficients betweentreated(IPOfirms)andyeardummies, andthebarsarethe90%confidenceintervals from the following regression: 4 (cid:88) Log(Debt) = β (d ×λ )+α +δ +u , i,c,t k i,c y,k,c i,c t,c i,c,t k=−3 where i, c, and t index firm, cohort (matched group), and time, respectively, d is a dummy i,c that equals one if the firm is treated (i.e., is an IPO firm). λ is a dummy equal to one if y,k,c year y is equal to k and zero otherwise, where k =0 is the year in which the IPO occurs, α i,c are firm/cohort fixed and δ are time/cohort fixed effects. Year zero contains the quarter in t,c which the firm IPOs. Standard errors are clustered by firm. 34

Figure 5: Leverage Dynamics Note: In this figure, we analyze the dynamics of firm leverage, i.e., debt/assets, before and after the IPO using a matched sample. We form a matched sample by estimating propensity scores based on our ex-ante regression (1) with date by industry by CBSA fixed effects, but only including IPO firms’ last quarter in the year prior to the IPO. The matched sample includes the three closest firms in terms of propensity scores that are in the same two-digit NAICS industry and have the same VC-backing status. The dots are point estimates of the interaction coefficients between treated (IPO firms) and year dummies, and the bars are the 90% confidence intervals from the following regression: 4 (cid:88) Debt/Assets = β (d ×λ )+α +δ +u , i,c,t k i,c y,k,c i,c t,c i,c,t k=−3 where i, c, and t index firm, cohort (matched group), and time, respectively, d is a dummy i,c that equals one if the firm is treated (i.e., is an IPO firm). λ is a dummy equal to one if y,k,c year y is equal to k and zero otherwise, where k =0 is the year in which the IPO occurs, α i,c are firm/cohort fixed and δ are time/cohort fixed effects. Year zero contains the quarter in t,c which the firm IPOs. Standard errors are clustered by firm. 35

Figure 6: PD Dispersion Dynamics Note: Inthisfigure,weanalyzethedynamicsofthedispersioninbanks’probabilityofdefault (PD) estimates, measured as the cross-sectional standard deviation in PD within firm/time, using a matched sample. We form a matched sample based on the Mahalanobis distance measure using PD and the number of banks as non-exact matching variables. Each IPO firm is matched to three control firms that are in the same two-digit NAICS industry and have the same VC-backing status. The dots are point estimates of the interaction coefficients between treated (IPO firms) and year dummies, and the bars are the 90% confidence intervals from the following regression: 4 (cid:88) SD(PD) = β (d ×λ )+α +δ +u , i,c,t k i,c y,k,c i,c t,c i,c,t k=−3 where i, c, and t index firm, cohort (matched group), and time, respectively, d is a dummy i,c that equals one if the firm is treated (i.e., is an IPO firm). λ is a dummy equal to one if y,k,c year y is equal to k and zero otherwise, where k =0 is the year in which the IPO occurs, α i,c are firm/cohort fixed and δ are time/cohort fixed effects. Year zero contains the quarter in t,c which the firm IPOs. Standard errors are clustered by firm. 36

Figure 7: Number of Banks Dynamics Note: In this figure, we analyze the dynamics of the number of banks the firm borrows from before and after the IPO using a matched sample. We form a matched sample by estimating propensity scores based on our ex-ante regression (1) with date by industry by CBSA fixed effects, but only including IPO firms’ last quarter in the year prior to the IPO. The matched sample includes the three closest firms in terms of propensity scores that are in the same twodigit NAICS industry and have the same VC-backing status. The dots are point estimates of the interaction coefficients between treated (IPO firms) and year dummies, and the bars are the 90% confidence intervals from the following regression: 4 (cid:88) NOB = β (d ×λ )+α +δ +u , i,c,t k i,c y,k,c i,c t,c i,c,t k=−3 where i, c, and t index firm, cohort (matched group), and time, respectively, d is a dummy i,c that equals one if the firm is treated (i.e., is an IPO firm). λ is a dummy equal to one if y,k,c year y is equal to k and zero otherwise, where k =0 is the year in which the IPO occurs, α i,c are firm/cohort fixed and δ are time/cohort fixed effects. Year zero contains the quarter in t,c which the firm IPOs. Standard errors are clustered by firm. 37

Figure 8: Syndicated Bank Loan Dynamics Note: In this figure, we analyze the dynamics of syndicated bank debt usage before and after the IPO using a matched sample. The dependent variable, Syndicated Debt/Debt, equals the ratio of syndicated loans outstanding over total debt. We form a matched sample by estimating propensity scores based on our ex-ante regression (1) with date by industry by CBSA fixed effects, but only including IPO firms’ last quarter in the year prior to the IPO. The matched sample includes the three closest firms in terms of propensity scores that are in thesametwo-digitNAICSindustryandhavethesameVC-backingstatus. Thedotsarepoint estimates of the interaction coefficients between treated (IPO firms) and year dummies, and the bars are the 90% confidence intervals from the following regression: 4 (cid:88) Syndicated Debt/Debt = β (d ×λ )+α +δ +u , i,c,t k i,c y,k,c i,c t,c i,c,t k=−3 where i, c, and t index firm, cohort (matched group), and time, respectively, d is a dummy i,c that equals one if the firm is treated (i.e., is an IPO firm). λ is a dummy equal to one if y,k,c year y is equal to k and zero otherwise, where k =0 is the year in which the IPO occurs, α i,c are firm/cohort fixed and δ are time/cohort fixed effects. Year zero contains the quarter in t,c which the firm IPOs. Standard errors are clustered by firm. 38

Figure 9: Bond Dynamics Note: In this figure, we analyze the dynamics of the bond usage before and after the IPO usingamatchedsample. Thedependentvariable,Bonds/TotalDebt,equalstheratioofbonds outstanding to total debt. We form a matched sample by estimating propensity scores based on our ex-ante regression (1) with date by industry by CBSA fixed effects, but only including IPO firms’ last quarter in the year prior to the IPO. The matched sample includes the three closest firms in terms of propensity scores that are in the same two-digit NAICS industry and have the same VC-backing status. The dots are point estimates of the interaction coefficients betweentreated(IPOfirms)andyeardummies, andthebarsarethe90%confidenceintervals from the following regression: 4 (cid:88) Bonds/Debt = β (d ×λ )+α +δ +u , i,c,t k i,c y,k,c i,c t,c i,c,t k=−3 where i, c, and t index firm, cohort (matched group), and time, respectively, d is a dummy i,c that equals one if the firm is treated (i.e., is an IPO firm). λ is a dummy equal to one if y,k,c year y is equal to k and zero otherwise, where k =0 is the year in which the IPO occurs, α i,c are firm/cohort fixed and δ are time/cohort fixed effects. Year zero contains the quarter in t,c which the firm IPOs. Standard errors are clustered by firm. 39

Table 1: Firm Level Summary Statistics: IPO vs. Non-IPO Firms This table contains summary statistics comparing IPO firm-quarters to non-IPO firm-quarters. Appendix B contains variable definitions. IPO Firms Non-IPO Firms N Mean Median SD N Mean Median SD Diff w.r.t. IPO firms Sales ($mm) 2898 1424.45 365.29 5414.61 1439472 414.60 66.37 4530.40 -1009.851∗∗∗ Assets ($mm) 2898 2019.63 549.01 6301.45 1439472 278.99 31.38 2601.11 -1740.649∗∗∗ Capex/Assets 2680 0.08 0.03 0.13 1235073 0.05 0.02 0.10 -0.026∗∗∗ Sales Growth 2684 0.42 0.16 0.73 1367765 0.15 0.07 0.42 -0.271∗∗∗ EBITDA/Assets 2706 0.08 0.09 0.26 1372216 0.16 0.12 0.22 0.085∗∗∗ Positive Profits 2706 0.76 1.00 0.43 1372216 0.89 1.00 0.31 0.130∗∗∗ Financing Deficit 2679 -0.00 -0.05 0.29 1231510 -0.12 -0.09 0.23 -0.120∗∗∗ Debt/Assets 2898 0.33 0.32 0.26 1439472 0.31 0.26 0.26 -0.021∗∗∗ Cash/Assets 2895 0.16 0.06 0.22 1435851 0.12 0.07 0.15 -0.039∗∗∗ VC-Backed 2898 0.29 0.00 0.45 1439472 0.02 0.00 0.12 -0.276∗∗∗ Silicon Valley 2898 0.14 0.00 0.35 1438585 0.02 0.00 0.15 -0.122∗∗∗ Tech/Life Science Firm 2898 0.23 0.00 0.42 1439472 0.06 0.00 0.23 -0.170∗∗∗ 40

Table 2: Summary Statistics: Loan-Level Sample This table contains summary statistics for the data used in our loan-level analyses. Appendix B contains variable definitions. N Mean Median SD P5 P95 Interest Rate (%) 58064 4.11 3.75 1.98 1.78 7.60 PD (%) 66620 1.56 0.88 2.89 0.14 4.54 LGD (%) 65034 33.21 34.97 15.00 7.00 54.60 PD × LGD (%) 64549 0.50 0.25 0.98 0.03 1.49 Maturity 73952 48.47 55.87 40.37 5.73 109.57 Loan Amount ($mm) 77944 10.91 3.80 32.37 1.00 40.00 Floating Rate 58822 0.69 1.00 0.46 0.00 1.00 Syndicated Loan 77944 0.13 0.00 0.33 0.00 1.00 41

Table 3: Cross-Sectional Determinants of Firms’ IPO Decisions This table tests which firm characteristics predict whether firms go public within the next three years. The dependent variable IPO is a dummy that equals one if the firm goes public in the next three years, which we multiply by 100. Standard errors are shown below the parameter estimatesinparenthesesandareclusteredbyfirm. *,**,and***indicatestatisticalsignificance at the 10%, 5%, and 1% levels, respectively. IPO (1) (2) (3) (4) Log(1+Sales) 0.240*** 0.238*** 0.251*** 0.253*** (0.019) (0.018) (0.020) (0.023) Sales Growth 0.507*** 0.313*** 0.491*** 0.284*** (0.058) (0.041) (0.060) (0.051) Debt/Assets 0.081 0.279*** 0.123* 0.295*** (0.061) (0.063) (0.064) (0.084) Financing Deficit 0.673*** 0.637*** 0.679*** 0.582*** (0.080) (0.073) (0.081) (0.096) Capex/Assets 0.154 -0.106 0.183 -0.079 (0.156) (0.124) (0.163) (0.190) NAICS4 MTB 0.109*** 0.109*** (0.014) (0.014) Date FE Y N N N Date × NAICS4 FE N Y N N Date × CBSA FE N N Y N Date × NAICS4 × CBSA FE N N N Y N 967151 1226028 960914 919120 R2 0.007 0.031 0.020 0.251 42

Table 4: Cross-Sectional Determinants of Firms’ IPO Decisions: Investment and Profitability This table tests which firm characteristics predict whether firms go public within the next three years. The dependent variable IPO is a dummy that equals one if the firm goes public within thenextthreeyears, whichwemultiplyby100. Standarderrorsareshownbelowtheparameter estimatesinparenthesesandareclusteredbyfirm. *,**,and***indicatestatisticalsignificance at the 10%, 5%, and 1% levels, respectively. IPO (1) (2) (3) (4) Capex/Assets × EBITDA/Assets -1.491*** -0.767*** -1.492*** -1.156*** (0.368) (0.276) (0.385) (0.397) Capex/Assets 1.259*** 0.755*** 1.294*** 0.839*** (0.239) (0.181) (0.249) (0.266) EBITDA/Assets -0.556*** -0.568*** -0.563*** -0.486*** (0.075) (0.072) (0.077) (0.097) Log(1+Sales) 0.241*** 0.238*** 0.252*** 0.253*** (0.019) (0.018) (0.020) (0.023) Sales Growth 0.507*** 0.313*** 0.492*** 0.283*** (0.059) (0.041) (0.060) (0.051) Debt/Assets 0.089 0.282*** 0.131** 0.299*** (0.060) (0.063) (0.063) (0.084) NAICS4 MTB 0.109*** 0.109*** (0.014) (0.014) Date FE Y N N N Date × NAICS4 FE N Y N N Date × CBSA FE N N Y N Date × NAICS4 × CBSA FE N N N Y N 967151 1226028 960914 919120 R2 0.007 0.031 0.020 0.251 43

Table 5: Cross-Sectional Determinants of Firms’ IPO Decisions (VC-Backed Sample) This table tests which firm characteristics predict whether firms go public within the next three years, restricting the sample to VC-backed firms. The dependent variable IPO is a dummy that equals one if the firm goes public in the next three years, which we multiply by 100. Standard errors are shown below the parameter estimates in parentheses and are clustered by firm. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. IPO (1) (2) (3) (4) Log(1+Sales) 2.393*** 2.809*** 2.441*** 2.945*** (0.322) (0.382) (0.347) (0.525) Sales Growth 1.043*** 0.864*** 1.009*** 0.510* (0.328) (0.307) (0.337) (0.288) Debt/Assets -3.170** -2.507* -2.940* -2.355 (1.455) (1.448) (1.660) (2.054) Financing Deficit 5.061*** 6.367*** 2.644* 4.143** (1.257) (1.370) (1.453) (1.719) Capex/Assets 3.183 0.787 6.007* 7.039* (2.858) (2.916) (3.207) (3.787) Date FE Y N N N Date × NAICS4 FE N Y N N Date × CBSA FE N N Y N Date × NAICS4 × CBSA FE N N N Y N 20450 18279 18710 11216 R2 0.069 0.224 0.179 0.358 44

Table 6: Going Public and Firms’ Borrowing Costs This table tests whether firms’ borrowing costs drop after the IPO. The sample includes only new loans. Standard errors are shown below the parameter estimates in parentheses and are clustered by firm. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. Interest Rate (%) (1) (2) (3) (4) IPO Firm × Post -0.511** -0.435** -0.427** -0.384** (0.208) (0.184) (0.170) (0.179) Log(Assets) -0.075*** -0.061** -0.048** -0.056** (0.026) (0.024) (0.024) (0.026) Sales Growth 0.030 0.034 0.047 0.055 (0.037) (0.036) (0.034) (0.040) Debt/Assets 0.423*** 0.444*** 0.414*** 0.368*** (0.092) (0.082) (0.081) (0.093) EBITDA/Assets -0.393*** -0.332*** -0.350*** -0.260*** (0.100) (0.078) (0.076) (0.073) Capex/Assets 0.208* 0.148 0.154 0.189* (0.114) (0.107) (0.098) (0.112) PD (%) -0.002 (0.012) LGD (%) 0.001 (0.001) PD × LGD (%) 0.141*** (0.036) Date FE Y Y Y Y Bank/Firm FE Y Y Y Y Bank/Date FE N N Y Y Loan Controls N Y Y Y N 37029 35822 35771 29723 R2 0.803 0.866 0.877 0.880 45

Appendix A. Additional Data Details In this section, we present additional details primarily relating to our assembly of our sample of private firms from the Y-14 data and our merging processes. A.1. Filtering the Y-14 raw data: additional details We apply several filtering measures to the Y-14 raw data. Specifically, we drop firms with missing taxpayer identification numbers (TINs), firms headquartered outside the US, firms with loans denominated in foreign currencies, borrowers that appear to be high-net-worth individuals, financial firms (NAICS code 52), real estate firms (NAICS code 92), and public administration and government entities (NAICS code 53). Some financial and non-profit firms have different industry classifications and are not dropped after this first pass, so we also drop any firms that have phrases in the firm names such as: “School of”, “CLO”, and similar. We also we exclude firms with the following terms in their names: real estate, subsidiary, properties, investment, newco, credit, family, acquisition, merger, series, holdco, finco, funding, trust, bank, banc mortgage, government, commonwealth, school, university, college, township, financing, finance, lease, leasing, foundation, insurance, retirement, church, temple, jewish, christian, muslim, bible, ymca, yeshiva, methodist, episcopalian, community, jesus, israel, redevelopment, partners, partnership, citigroup, citicorp, jpmorgan, metlife, airport, hathaway, museum, nonprofit, non-profit, public, china, usa, securitization, ubs ag, north america, receivables company, distribution company, client services inc., institutional fund, reit, clo, spv, iii, ii, iv, viii, vii, vi, county of, counties of, city of, town of, state of, board of, district of, borough of, society of, college of, council of, council for, center of, center for, educational estate, national association, non profit, indian tribe, development auth, development and auth, developmentauth, building auth, and housing dev. We use the name filters in order to exclude records in which industries are incorrectly classified or missing. The Y-14 data includes the date indicating the period-end for each corresponding borrower firm’s latest financial data, which we use to construct our panel of borrower financial data. For smaller private firms, the financial data are generally updated on an annual basis, while for larger public firms, the financial data are generally updated quarterly. Throughout our analysis, we fill-down intra-year borrower financial data, by at most three quarters, for firms with financial data only reported at annual frequency. Our results are robust to the removal of the within-year fill-down process; however, the fill-down increases the power of our time-series tests. For variables constructed from loan-level information, we use the dates that correspond to the borrower’s loan record, rather than the date of the borrower’s most recent financials. Becauseofthedifferentdatefields, constructingapanelthatcontainsboththe 46

private firm’s financial data and its banking characteristics requires constructing separate panels using the two sets of dates and then merging the loans dataset with the firm-level dataset. This process ensures that our panel of borrower financial data and bank debt characteristics are synced correctly. We make several additional adjustments to clean the data. First, some banks mistakenly record borrowers’ capital expenditures as a negative number. Therefore, we replace all capex records with their absolute value, prior to taking the median across various bank loan records. Prior to aggregating the data to the firm/quarter level, we also drop observations in which assets are less than 50% or greater than 150% of the median value across all loan records for the same firm for the same date, to exclude a small number of records that are reported in different units (i.e., thousands vs. millions). We also drop observations with negative debt.58 Finally, for categorical variables such as NAICS, zip code, borrower firm name, CUSIP, ticker, and year established, we take the mode across loan observations within each firm-quarter. A.2. Merging the private firms and the IPO firms samples: additional details We identify public firms, and subsidiaries of private firms, in the Y-14 data using a process involving a merge with Compustat, which is similar to the process in Beyhaghi et al. (2024). First, we merge the Y-14 panel by TIN and quarter for the borrower, and the borrower’s guarantor (if populated), with the panel of firms from Compustat that have non-missing stock prices. We assign all firms in the Y-14 data with TINs that match firms in the Compustat panel in the same quarter, or that have guarantors with TINs associated with public firms.59 We also assign all firms in the Y-14 data as public if any of the firm’s loan records, within the same bank, are associated with a non-missing CUSIP or ticker. Finally, we also exclude the top 1% of the largest firms, by assets, to ensure that we exclude miscoded outliers or large subsidiaries of public firms that were not properly identified. A.3. Identifying the IPO Firms in Y-14: additional details Our process for identifying the IPO firms in the Y-14 data involves several steps. First, we merge the IPO firms from the SDC Platinum database with Compustat, using CUSIP identifiers in both databases, to find TIN identifiers for each IPO firm, which are only included in the Compustat data. We then use TIN identifiers to merge the IPO firms 58We drop roughly 4,100 loan-level observations due to these filters, which represent about 0.1% of the loan-level debt observations. 59We designate firms that have public guarantors as public, as these are either subsidiaries of public firmsorotherwiselinkedtoapublicfirminsuchawaythatthelenderlikelyassociatesthecharacteristics of the public guarantor with the otherwise private firm. 47

from the SDC Platinum database with the private firms in the Y-14 data. Some firms in the Y-14 data that change TIN identifiers after going public. Therefore, we also hand-match IPO firms from the SDC Platinum database using each firm’s name, location, and industry. For each IPO firm from SDC, we compile all private firms in the Y-14 data with the same state code and two-digit NAICS industry codes. Among these potential matches, we use the FEDMATCH string-matching algorithm to identify the ten closest name matches. We manually select the closest firm, or firms, among these. For instances where we determine that there are no good match candidates, we do not designate a match. Through our hand-matching process, we also identify subsidiaries of IPO companies that were missed by our initial filters and drop them from the sample. A.4. Identifying firms with VC-backing, merger targets, and bonds in Y-14: additional details We next explain in more detail our approach to matching the Y-14 with three other datasets: Preqin, the SDC Platinum M&A dataset, and FISD. Our method for merging the Preqin data involves several steps. For each firm in Preqin, we assemble all firms in Y-14 with matching state codes and the 2-digit NAICS codes. Secondly, among each Preqin firm’s location and industry peers, we select the Y-14 firm with the most similar firm name as ranked by Jaro–Winkler distance (after cleaning the firm names in each database to exclude terms like ‘corp’ and ‘inc’) if the best match has a Jaro–Winkler distance above 0.85. Using this method, we match about 15% of the firms in the Preqin database to our panel of private firms in Y-14, which translates to over 8,000 unique VC-backed private firms. Our method for identifying private firms that are acquired based on the SDC Platinum mergers and acquisitions dataset involves the same string-matching procedure using cleaned firm-names, and the same requirement that matches share the same location, and 2-digit NAICS industry codes. Our method for identifying private firms that have bonds outstanding using the FISD involves two steps. The FISD data have CUSIP identifiers, which we use as a first-stage merge to Y-14. For the firms in FISD that do not merge using CUSIPS, we employ the same matching method as described above: we use a string-match using cleaned firm names, and require that matches share the same location, and 2-digit NAICS industry codes. 48

Appendix B. Variable Definitions Assets: Total assets, from Y-14. Bonds/Debt: The ratio of total bonds outstanding to total debt, trimmed if > 1, from FISD and Y-14. Capex: Funds used to acquire a long-term asset resulting in depreciation deductions over the life of the acquired asset, from Y-14. Capex/Assets: Funds used to acquire a long-term asset resulting in depreciation deductions over the life of the acquired asset divided by total assets, winsorized at [1%, 99%], from Y-14. Debt: Total debt, from Y-14. EBITDA/Assets: EBITDA/Assets, winsorized at [1%, 99%], from Y-14. Financing Deficit: (Capex-EBITDA)/Assets, winsorized at [1%, 99%], from Y-14. Intangible Assets: Total assets minus tangible assets, from Y-14. Interest Rate: Annual interest rate of the loan, trimmed if negative, in percentages, from Y-14. IPO: Dummy variable that equals one if the firm goes public within the next three years, multiplied by 100, from SDC. IPO Firm: Dummy variable that equals one if the firm goes public at all during the sample period, from SDC. Leverage: Debt/Assets, winsorized at [1%, 99%], from Y-14. Loss Given Default (LGD): The bank’s estimated loss given default per unit of loan weight by the committed dollar amount of each loan at the bank/firm/quarter level, from Y-14 trimmed if LGD = 0 or LGD = 1. Maturity: Remaining maturity in months weight by the committed dollar amount of each loan at the bank/firm/quarter level, from Y-14. NOB: The number of banks the firm borrows from as of the current quarter, from Y-14. 49

NAICS4 MTB: The median market-to-book ratio of publicly traded companies for a given four-digit NAICS industry within the given quarter, from Compustat. Positive Profits: DummyvariablethatequalsoneifthefirmhasapositiveEBITDA/Assets, from Y-14. Post: Dummy variable that equals one if the firm has IPOed as of the current quarter, from Y-14. Probability of Default (PD): Thebank’sexpectedannualdefaultrateoverthelifeofthe loan weighted by the committed dollar amount of each loan at the bank/firm/quarter level, trimmed if PD = 0 or PD = 1, from Y-14. Sales Growth: Annual sales growth, winsorized at [1%, 99%], from Y-14. Silicon Valley: Dummy variable that equals one if the firm is located in Silicon Valley defined as CBSA San Francisco-Oakland-Hayward (code 41860) or San Jose-Sunnyvale- Santa Clara (code 41940), from Y-14 and HUD. Standard Deviation of PD: The cross-sectional standard deviation of PD across banks within firm/time, from Y-14. Syndicated Debt/Debt: The ratio of total syndicated bank debt to total debt, trimmed if > 1, from Y-14. Technology and Life Sciences Firm: Dummy variable that equals one if the firm has one of the following NAICS codes: 3254, 3341, 3342, 3344, 3345, 3364, 5122, 5182, 5191, 5413, 5415, 5417, from Y-14. VC-Backed: Dummy variable that equals one if the firm has received funding from a private equity fund in the Preqin VC funding dataset, from Preqin. 50

Appendix C. Additional Tests: For Online Publication Figure C1: PD Dispersion Dynamics (Fixed Set of Banks) Note: Inthisfigure,weanalyzethedynamicsofthedispersioninbanks’probabilityofdefault (PD) estimates, measured as the cross-sectional standard deviation in PD within firm/time, using a matched sample that maintains a fixed set of banks for each firm throughout the analysisperiod. WeformamatchedsamplebasedontheMahalanobisdistancemeasureusing PD and the number of banks as non-exact matching variables. Each IPO firm is matched to three control firms that are in the same two-digit NAICS industry and have the same VCbacking status. The dots are point estimates of the interaction coefficients between treated (IPOfirms)andyeardummiesandthebarsarethe90%confidenceintervalsfromthefollowing regression: 4 (cid:88) SD(PD) = β (d ×λ )+α +δ +u , i,c,t k i,c y,k,c i,c t,c i,c,t k=−3 where i, c and t index firm, cohort (matched group) and time respectively, d is a dummy i,c that equals one if the firm is treated (i.e., is an IPO firm). λ is a dummy equal to one if y,k,c year y is equal to k and zero otherwise, where k =0 is the year in which the IPO occurs, α i,c are firm/cohort fixed and δ are time/cohort fixed effects. Year zero contains the quarter in t,c which the firm IPOs. Year zero contains the quarter in which the firm IPOs. Standard errors are clustered by firm. 51

Figure C2: IPO Number of Banks Dynamics (Poisson Regression) Note: In this figure, we analyze the dynamics of the number of banks firms borrow from before and after the IPO using a matched sample one quarter prior to IPO using a Poisson regression. We form a matched sample by estimating propensity scores based on our ex-ante regression (1) with date by industry by CBSA fixed effects, but only including IPO firms’ last quarter in the year prior to the IPO. The matched sample includes the three closest firms in terms of propensity scores that are in the same two-digit NAICS industry and have the same VC-backing status. The dots are point estimates of the interaction coefficients between treated (IPO firms) and year dummies and the bars are the 90% confidence intervals from the following regression: 4 (cid:88) E[NOB |X ]= β (d ×λ )+α +δ , i,c,t i,c,t k i,c y,k,c i,c t,c k=−3 where i, c and t index firm, cohort (matched group) and time respectively, X is the set i,c,t of all predictors, d is a dummy that equals one if the firm is treated (i.e., is an IPO firm). i,c λ is a dummy equal to one if the year y is equal to k and zero otherwise, where k = 0 y,k,c is the year in which the IPO occurs, α are firm/cohort fixed and δ are time/cohort fixed i,c t,c effects. Year zero contains the quarter in which the firm IPOs. Standard errors are clustered by firm. 52

Figure C3: PD Dispersion Dynamics (Non-Syndicated Loans) Note: Inthisfigure,weanalyzethedynamicsofthedispersioninbanks’probabilityofdefault (PD) estimates among non-syndicated loans, measured as the cross-sectional standard deviation in PD within firm/time, using a matched sample. We form a matched sample based on the Mahalanobis distance measure using PD and the number of banks as non-exact matching variables. Each IPO firm is matched to three control firms that are in the same two-digit NAICS industry and have the same VC-backing status. The dots are point estimates of the interaction coefficients between treated (IPO firms) and year dummies and the bars are the 90% confidence intervals from the following regression: 4 (cid:88) SD(PD) = β (d ×λ )+α +δ +u , i,c,t k i,c y,k,c i,c t,c i,c,t k=−3 where i, c, and t index firm, cohort (matched group), and time, respectively, d is a dummy i,c that equals one if the firm is treated (i.e., is an IPO firm). λ is a dummy equal to one if y,k,c year y is equal to k and zero otherwise, where k =0 is the year in which the IPO occurs, α i,c are firm/cohort fixed and δ are time/cohort fixed effects. Year zero contains the quarter in t,c which the firm IPOs. Standard errors are clustered by firm. 53

Table C1: IPO Sample Composition This table displays differences in the composition of our sample of IPO firms in the Y-14 data and the full sample of IPOs in SDC. IPO firms are considered in our sample if there is at least one observation within a three-year window prior to the IPO. See Section 2.2 for details regarding our matching and filtering processes. (1) (2) (3) SDC Assets Y14 IPO Below IPO NAICS 4-Digit Industry: Sample $10mm Sample Panel A: Number of Unique IPO Firms Pharmaceutical and Medicine Manufacturing 284 69 21 Scientific Research and Development Services (Biotech) 234 37 16 Software Publishers 161 35 72 Medical Equipment and Supplies Manufacturing 56 9 12 Computer Systems Design and Related Services 36 4 15 Electronic Instrument Manufacturing 32 8 7 Oil and Gas Extraction 21 0 15 Data Processing, Hosting and Related Services 20 6 10 Full Service Restaurants 18 2 11 Semiconductor and Electronic Component Manufacturing 17 4 6 Other 511 54 238 Total 1,390 228 423 Share of SDC Total 16% 30% Panel B: Aggregate IPO Proceeds ($bn) Total IPO Proceeds $346.2 $3.9 $211.8 Share of SDC Total 1% 61% 54

Table C2: Summary Statistics (IPO Firms in Y-14 vs. IPO Firms in SDC Only) This table compares our sample of IPO firms in the Y-14 data with IPO firms in SDC, but not in our sample. We include all firm-quarter observations within three years prior to the IPO. We obtain pre-IPO financial data from Compustat. Appendix B contains variable definitions. Merged IPO Firms Unmerged IPO Firms N Mean Median SD N Mean Median SD Diff w.r.t. Merged IPO Firms Sales (mm) 2898 1424.45 365.29 5414.61 2932 381.95 0.20 1724.33 -1042.500∗∗∗ Assets (mm) 2898 2019.63 549.01 6301.45 2932 573.42 59.91 1896.86 -1446.216∗∗∗ Capex/Assets 2680 0.08 0.03 0.13 991 0.07 0.02 0.13 -0.005 Sales Growth 2684 0.42 0.16 0.73 535 0.30 0.21 0.44 -0.124∗∗∗ EBITDA/Assets 2706 0.08 0.09 0.26 693 -0.19 -0.29 0.32 -0.263∗∗∗ Positive Profits 2706 0.76 1.00 0.43 693 0.39 0.00 0.49 -0.372∗∗∗ Debt/Assets 2898 0.33 0.32 0.26 2925 0.34 0.24 0.34 0.011 Cash/Assets 2895 0.16 0.06 0.22 2925 0.43 0.42 0.33 0.265∗∗∗ VC-Backed 2898 0.29 0.00 0.45 2932 0.19 0.00 0.39 -0.104∗∗∗ Silicon Valley 2898 0.14 0.00 0.35 2928 0.19 0.00 0.39 0.043∗∗∗ Tech/Life Science Firm 2898 0.23 0.00 0.42 2932 0.62 1.00 0.48 0.396∗∗∗ 55

Table C3: Determinants of IPO Firms Being Present in Y-14 Data This table tests which firm characteristics predict whether IPO firms are in our sample of Y-14 data or only in SDC. The dependent variable is a dummy that equals one if the firm is in our sample of Y-14 data within a three-year window prior to the IPO. We obtain pre-IPO financial datafromCompustat. Standarderrors areshownbelow theparameterestimates inparentheses and are clustered by firm. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. In Y-14 Sample (1) (2) Pharma & Medicine Manufacturing -0.325*** 0.037 (0.040) (0.049) Scientific R&D (Biotech) -0.393*** 0.010 (0.042) (0.051) Software Publishers 0.075* 0.093** (0.043) (0.039) Medical Equipment Manufacturing -0.210*** -0.030 (0.080) (0.072) Computer Systems Design 0.009 0.048 (0.098) (0.093) Electronic Instrument Manufacturing -0.134 0.037 (0.107) (0.093) Oil and Gas Extraction 0.073 0.022 (0.084) (0.096) Data Processing & Hosting -0.015 0.012 (0.119) (0.094) Full-Service Restaurants 0.087 0.079 (0.101) (0.079) Semis & Components Manufacturing -0.095 0.018 (0.152) (0.130) Log(Sales) 0.091*** (0.007) Year FE Y Y N 1036 1036 R2 0.219 0.329 F-statistic (industry dummies) 17.642 0.764 F-test p-value 0.000 0.664 56

Table C4: Industry Composition of IPO Firms This table displays the distribution of industries, based on four-digit NAICS codes, in our sample of private firms that ultimately go public. Industry # of Firms % of Total Software Publishers 41 11.02 Computer Systems Design & Related Services 19 5.11 Data Processing, Hosting, & Related Service 16 4.30 Pharmaceutical & Medicine Manufacturing 15 4.03 Oil and Gas Extraction 13 3.49 Restaurants & Other Eating Places 11 2.96 Electronic Shopping 9 2.42 Support Activities for Mining 8 2.15 Lumber & Other Construction Materials Wholesalers 8 2.15 Other Information Services 7 1.88 Electric Power Gen, Transmission and Distribution 6 1.61 Miscellaneous Durable Goods Manufacturing 6 1.61 Scientific Research & Development Services 6 1.61 Traveler Accommodation 6 1.61 Navigation, Measuring, Electromed, & Control Instruments 5 1.34 Clothing Stores 5 1.34 Architectural, Engineering, & Related 5 1.34 Management, Scientific, & Technical Consulting 5 1.34 Other Amusement & Recreation Industries 5 1.34 Residential Building Construction 4 1.08 Semiconductor & Other Component Manufacturing 4 1.08 Professional & Commercial Equipment & Supplies Wholesalers 4 1.08 Grocery & Related Product Merchant Wholesalers 4 1.08 Advertising Agencies 4 1.08 Business Support Services 4 1.08 Investigation and Security Services 4 1.08 Other Wood Product Manufacturing 3 0.81 Basic Chemical Manufacturing 3 0.81 Soap, Cleaning Comp, and Toilet Prep Manufacturing 3 0.81 Plastics Product Manufacturing 3 0.81 57

Table C5: Location Composition of IPO Firms This table displays the distribution of firms’ headquarters CBSA in our sample of private firms that ultimately IPO. Industry # of Firms % of Total San Francisco-Oakland-Hayward 43 11.56 New York-Newark-Jersey City 25 6.72 Boston-Cambridge-Newton 23 6.18 Los Angeles-Long Beach-Anaheim 22 5.91 Dallas-Fort Worth-Arlington 17 4.57 San Jose-Sunnyvale-Santa Clara 16 4.30 Chicago-Naperville-Elgin 15 4.03 Washington-Arlington-Alexandria 13 3.49 Philadelphia-Camden-Wilmington 12 3.23 Phoenix-Mesa-Scottsdale 11 2.96 Orlando-Kissimmee-Sanford 10 2.69 Indianapolis-Carmel-Anderson 9 2.42 Austin-Round Rock 8 2.15 Atlanta-Sandy Springs-Roswell 7 1.88 Minneapolis-St. Paul-Bloomington 6 1.61 Raleigh 6 1.61 Cleveland-Elyria 5 1.34 Detroit-Warren-Dearborn 5 1.34 Virginia Beach-Norfolk-Newport News 5 1.34 Bridgeport-Stamford-Norwalk 4 1.08 San Antonio-New Braunfels 4 1.08 Denver-Aurora-Lakewood 4 1.08 Las Vegas-Henderson-Paradise 4 1.08 Miami-Fort Lauderdale-West Palm Beach 4 1.08 Portland-Vancouver-Hillsboro 4 1.08 Salt Lake City 4 1.08 Riverside-San Bernardino-Ontario 4 1.08 Seattle-Tacoma-Bellevue 4 1.08 Cape Coral-Fort Myers 3 0.81 Provo-Orem 3 0.81 58

Table C6: Industry Composition of Private Firm Sample This table displays the distribution of industries, based on four-digit NAICS codes, in our sample of private firms. Industry # of Firms % of Total Automobile Dealers 13,478 13.40 Restaurants & Other Eating Places 2,782 2.77 Wholesale Distribution 2,415 2.40 Computer Systems Design & Related Services 1,939 1.93 Grocery & Related Product Merchant Wholesalers 1,682 1.67 Nonresidential Building Construction 1,574 1.56 Building Equipment Contractors 1,514 1.51 Architectural, Engineering, & Related 1,472 1.46 General Freight Trucking 1,463 1.45 Software Publishers 1,406 1.40 Management, Scientific, & Technical Consulting 1,372 1.36 Other Motor Vehicle Dealers 1,297 1.29 Misc Durable Goods Merchant Wholesalers 1,215 1.21 Plastics Product Manufacturing 1,064 1.06 Offices of Physicians 1,007 1.00 Electric Power Gen, Transmission and Distribution 988 0.98 Apparel & Accessories, Not Elsewhere 926 0.92 Other Amusement & Recreation Industries 919 0.91 Motor Vehicle Parts & Supplies Wholesalers 895 0.89 Professional & Commercial Equipment & Supplies Wholesalers 876 0.87 Household Appliances & Electrical Goods Wholesalers 872 0.87 Lumber & Other Construction Materials Wholesalers 871 0.87 General Medical & Surgical Hospitals 867 0.86 Legal Services 859 0.85 Highway, Street, and Bridge Construction 858 0.85 Management of Companies and Enterprises 852 0.85 Nursing Care Facilities 817 0.81 Support Activities for Mining 774 0.77 Miscellaneous Nondurable Goods Wholesalers 746 0.74 Oil and Gas Extraction 744 0.74 59

Table C7: Location Composition of Private Firm Sample Thistabledisplaysthedistributionoffirms’headquartersCBSAinoursampleofprivatefirms. Industry # of Firms % of Total New York-Newark-Jersey City 8,948 8.93 Los Angeles-Long Beach-Anaheim 4,644 4.63 Chicago-Naperville-Elgin 4,030 4.02 Philadelphia-Camden-Wilmington 2,692 2.69 Indianapolis-Carmel-Anderson 2,602 2.60 Non-Metro Area 2,417 2.41 Dallas-Fort Worth-Arlington 2,301 2.30 Washington-Arlington-Alexandria 2,248 2.24 San Francisco-Oakland-Hayward 2,209 2.20 Miami-Fort Lauderdale-West Palm Beach 2,177 2.17 Boston-Cambridge-Newton 2,168 2.16 Detroit-Warren-Dearborn 2,120 2.12 Atlanta-Sandy Springs-Roswell 1,960 1.96 Seattle-Tacoma-Bellevue 1,439 1.44 Minneapolis-St. Paul-Bloomington 1,348 1.35 Cleveland-Elyria 1,247 1.24 Phoenix-Mesa-Scottsdale 1,192 1.19 Denver-Aurora-Lakewood 1,122 1.12 Charlotte-Concord-Gastonia 1,117 1.11 Orlando-Kissimmee-Sanford 1,035 1.03 Portland-Vancouver-Hillsboro 1,035 1.03 Sacramento-Roseville-Arden-Arcade 957 0.96 San Antonio-New Braunfels 954 0.95 Riverside-San Bernardino-Ontario 925 0.92 Tampa-St. Petersburg-Clearwater 899 0.90 Columbus, OH 894 0.89 Milwaukee-Waukesha-West Allis 830 0.83 Indianapolis-Carmel-Greenwood 822 0.82 San Jose-Sunnyvale-Santa Clara 800 0.80 Grand Rapids-Wyoming 756 0.75 60

Table C8: Cross-Sectional Determinants of Firms’ IPO Decisions (Technology and Life Science Firms Only) This table tests which firm characteristics predict whether firms go public within the next three years among technology and life sciences firms. The dependent variable IPO is a dummy that equals one if the firm goes public in the next three years, which we multiply by 100. Standard errors are shown below the parameter estimates in parentheses and are clustered by firm. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. IPO (1) (2) (3) (4) Log(1+Sales) 1.082*** 0.861*** 1.061*** 0.875*** (0.173) (0.143) (0.179) (0.173) Sales Growth 1.138*** 0.748*** 1.085*** 0.436** (0.263) (0.189) (0.279) (0.178) Debt/Assets 0.932 0.385 0.785 -0.243 (0.607) (0.495) (0.622) (0.553) Financing Deficit 2.672*** 2.398*** 2.359*** 1.839*** (0.427) (0.355) (0.435) (0.370) Capex/Assets 0.049 -0.671 -0.053 -0.323 (1.303) (1.024) (1.262) (1.110) NAICS4 MTB 0.749*** 0.738*** (0.210) (0.239) Date FE Y N N N Date × NAICS4 FE N Y N N Date × CBSA FE N N Y N Date × NAICS4 × CBSA FE N N N Y N 62664 75547 59370 61176 R2 0.032 0.041 0.119 0.218 61

Table C9: Cross-Sectional Determinants of Firms’ IPO Decisions (VC-Backed Technology and Life Science Firms Only) This table tests which firm characteristics predict whether firms go public within the next three years among VC-backed, technology and life science firms. The dependent variable IPO is a dummythatequalsoneifthefirmgoespublicinthenextthreeyears, whichwemultiplyby100. Standard errors are shown below the parameter estimates in parentheses and are clustered by firm. *, **, and***indicatestatisticalsignificanceatthe10%, 5%, and1%levels, respectively. IPO (1) (2) (3) (4) Log(1+Sales) 3.301*** 3.532*** 3.174*** 3.814*** (0.503) (0.541) (0.536) (0.671) Sales Growth 1.333*** 1.062*** 1.183*** 0.773** (0.360) (0.328) (0.347) (0.344) Debt/Assets -4.033** -4.808*** -4.668** -3.828 (1.705) (1.607) (1.829) (2.511) Financing Deficit 5.230*** 6.238*** 2.118 5.962*** (1.779) (1.717) (2.100) (2.063) Capex/Assets 5.587 1.772 12.563** 5.311 (4.183) (3.997) (5.122) (4.959) Date FE Y N N N Date × NAICS4 FE N Y N N Date × CBSA FE N N Y N Date × NAICS4 × CBSA FE N N N Y N 10797 10402 9687 7628 R2 0.097 0.188 0.224 0.329 62

Table C10: Cross-Sectional Determinants of Firms’ IPO Decisions (Controlling for VC Backing) This table tests which firm characteristics predict whether firms go public within the next three years, controlling for whether the firm is VC-backed. The dependent variable IPO is a dummy that equals one if the firm goes public in the next three years, which we multiply by 100. Standard errors are shown below the parameter estimates in parentheses and are clustered by firm. *, **, and***indicatestatisticalsignificanceatthe10%, 5%, and1%levels, respectively. IPO (1) (2) (3) (4) Log(1+Sales) 0.255*** 0.247*** 0.264*** 0.261*** (0.020) (0.018) (0.021) (0.023) Sales Growth 0.291*** 0.203*** 0.295*** 0.187*** (0.052) (0.038) (0.054) (0.048) Debt/Assets 0.227*** 0.355*** 0.248*** 0.364*** (0.058) (0.062) (0.062) (0.083) Financing Deficit 0.214*** 0.271*** 0.249*** 0.240*** (0.067) (0.064) (0.069) (0.089) Capex/Assets 0.441*** 0.072 0.430*** 0.092 (0.157) (0.125) (0.164) (0.193) NAICS4 MTB 0.062*** 0.065*** (0.012) (0.013) VC-Backed 5.174*** 3.458*** 5.095*** 3.317*** (0.555) (0.414) (0.569) (0.481) Date FE Y N N N Date × NAICS4 FE N Y N N Date × CBSA FE N N Y N Date × NAICS4 × CBSA FE N N N Y N 967151 1226028 960914 919120 R2 0.020 0.037 0.032 0.256 63

Table C11: Cross-Sectional Determinants of Firms’ IPO Decisions (Excluding Merger Targets) This table tests which firm characteristics predict whether firms go public within the next three years, excluding firms that were acquired within the next three years. The dependent variable IPO is a dummy that equals one if the firm goes public in the next three years, which we multiply by 100. Standard errors are shown below the parameter estimates in parentheses and are clustered by firm. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. IPO (1) (2) (3) (4) Log(1+Sales) 0.232*** 0.228*** 0.242*** 0.240*** (0.019) (0.018) (0.020) (0.022) Sales Growth 0.510*** 0.318*** 0.498*** 0.299*** (0.059) (0.042) (0.060) (0.051) Debt/Assets 0.050 0.241*** 0.089 0.253*** (0.060) (0.061) (0.063) (0.082) Financing Deficit 0.672*** 0.633*** 0.682*** 0.601*** (0.081) (0.074) (0.082) (0.097) Capex/Assets 0.179 -0.079 0.205 -0.085 (0.157) (0.124) (0.164) (0.189) NAICS4 MTB 0.109*** 0.110*** (0.014) (0.014) Date FE Y N N N Date × NAICS4 FE N Y N N Date × CBSA FE N N Y N Date × NAICS4 × CBSA FE N N N Y N 961535 1219660 955315 913705 R2 0.007 0.030 0.020 0.250 64

Table C12: Cross-Sectional Determinants of Firms’ IPO Decisions (Excluding Private Firms in the Sample for Less than Three Years) This table tests which firm characteristics predict whether firms go public within the next three years, excluding private firms that remain in the Y-14 data for less than three years. The dependent variable IPO is a dummy that equals one if the firm goes public in the next three years, which we multiply by 100. Standard errors are shown below the parameter estimates in parentheses and are clustered by firm. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. IPO (1) (2) (3) (4) Log(1+Sales) 0.443*** 0.515*** 0.459*** 0.454*** (0.038) (0.041) (0.040) (0.046) Sales Growth 1.212*** 1.039*** 1.202*** 0.825*** (0.145) (0.133) (0.149) (0.157) Debt/Assets 0.187* 0.814*** 0.298** 0.829*** (0.113) (0.139) (0.120) (0.171) Financing Deficit 1.252*** 1.680*** 1.300*** 1.407*** (0.159) (0.194) (0.163) (0.251) Capex/Assets 0.464 -0.622** 0.486 -0.553 (0.320) (0.301) (0.338) (0.450) NAICS4 MTB 0.190*** 0.200*** (0.027) (0.029) Date FE Y N N N Date × NAICS4 FE N Y N N Date × CBSA FE N N Y N Date × NAICS4 × CBSA FE N N N Y N 494040 537783 488297 367318 R2 0.115 0.120 0.109 0.342 65

Table C13: Cross-Sectional Determinants of Firms’ IPO Decisions (Asset Decile Fixed Effects) This table tests which firm characteristics predict whether firms go public within the next three years, interacting the main set of fixed effects from Table 3 with asset decile fixed effects. The dependent variable IPO is a dummy that equals one if the firm goes public in the next three years, which we multiply by 100. Standard errors are shown below the parameter estimates in parentheses and are clustered by firm. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. IPO (1) (2) (3) (4) Log(1+Sales) 0.006 0.115*** 0.015 0.180*** (0.018) (0.026) (0.021) (0.057) Sales Growth 0.337*** 0.302*** 0.348*** 0.258*** (0.041) (0.040) (0.045) (0.063) Debt/Assets 0.045 0.065 0.089 0.190* (0.051) (0.059) (0.061) (0.114) Financing Deficit 0.376*** 0.479*** 0.404*** 0.576*** (0.060) (0.070) (0.067) (0.119) Capex/Assets 0.086 0.099 0.187 0.207 (0.122) (0.125) (0.142) (0.225) Date × Assets Decile FE Y N N N Date × NAICS4 × Assets Decile FE N Y N N Date × CBSA × Assets Decile FE N N Y N Date × NAICS4 × CBSA × Assets Decile FE N N N Y N 1226248 1210242 1156429 562407 R2 0.012 0.137 0.077 0.410 66

Table C14: Cross-Sectional Determinants of Firms’ IPO Decisions (Including Firms With Less Than $10mm in Assets) This table tests which firm characteristics predict whether firms go public within the next three years, similar to Table 3 but including firms with less than $10mm in assets. The dependent variable IPO is a dummy that equals one if the firm goes public in the next three years, which we multiply by 100. Standard errors are shown below the parameter estimates in parentheses and are clustered by firm. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. IPO (1) (2) (3) (4) Log(1+Sales) 0.116*** 0.116*** 0.122*** 0.129*** (0.010) (0.010) (0.010) (0.011) Sales Growth 0.184*** 0.111*** 0.177*** 0.101*** (0.023) (0.017) (0.023) (0.021) Debt/Assets 0.128*** 0.196*** 0.150*** 0.212*** (0.030) (0.031) (0.031) (0.039) Financing Deficit 0.050*** 0.042*** 0.053*** 0.040*** (0.008) (0.007) (0.009) (0.009) Capex/Assets 0.175*** 0.104*** 0.170*** 0.123** (0.046) (0.035) (0.048) (0.051) NAICS4 MTB 0.053*** 0.053*** (0.007) (0.008) Date FE Y N N N Date × NAICS4 FE N Y N N Date × CBSA FE N N Y N Date × NAICS4 × CBSA FE N N N Y N 1669900 2075598 1664722 1663366 R2 0.004 0.018 0.012 0.210 67

Table C15: Cross-Sectional Determinants of Firms’ IPO Decisions (Excluding Firms with Less than $50mm in Assets) This table tests which firm characteristics predict whether firms go public within the next three years,butincludesonlyfirmswithmorethan$50mminassets. ThedependentvariableIPOisa dummythatequalsoneifthefirmgoespublicinthenextthreeyears, whichwemultiplyby100. Standard errors are shown below the parameter estimates in parentheses and are clustered by firm. *, **, and***indicatestatisticalsignificanceatthe10%, 5%, and1%levels, respectively. IPO (1) (2) (3) (4) Log(1+Sales) 0.307*** 0.379*** 0.339*** 0.431*** (0.036) (0.036) (0.039) (0.058) Sales Growth 0.975*** 0.583*** 0.943*** 0.514*** (0.119) (0.084) (0.122) (0.115) Debt/Assets 0.387** 0.543*** 0.468*** 0.560** (0.160) (0.147) (0.176) (0.240) Financing Deficit 1.325*** 1.367*** 1.364*** 1.463*** (0.204) (0.203) (0.213) (0.331) Capex/Assets 0.214 -0.101 0.438 0.086 (0.407) (0.317) (0.442) (0.496) NAICS4 MTB 0.234*** 0.226*** (0.034) (0.036) Date FE Y N N N Date × NAICS4 FE N Y N N Date × CBSA FE N N Y N Date × NAICS4 × CBSA FE N N N Y N 358406 462282 351184 295632 R2 0.007 0.057 0.039 0.313 68

Table C16: Cross-Sectional Determinants of Firms’ IPO Decisions (Excluding Tech/SV) This table tests which firm characteristics predict whether firms go public within the next three years, excluding technology and life science firms as well as those from Silicon Valley. The dependent variable IPO is a dummy that equals one if the firm goes public in the next three years, which we multiply by 100. Standard errors are shown below the parameter estimates in parentheses and are clustered by firm. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. IPO (1) (2) (3) (4) Log(1+Sales) 0.176*** 0.178*** 0.181*** 0.175*** (0.017) (0.016) (0.018) (0.018) Sales Growth 0.342*** 0.220*** 0.349*** 0.252*** (0.053) (0.038) (0.055) (0.050) Debt/Assets 0.129** 0.266*** 0.143** 0.260*** (0.052) (0.056) (0.056) (0.075) Financing Deficit 0.387*** 0.392*** 0.419*** 0.451*** (0.069) (0.065) (0.072) (0.092) Capex/Assets 0.269* 0.002 0.261* -0.071 (0.140) (0.111) (0.148) (0.182) NAICS4 MTB 0.054*** 0.058*** (0.010) (0.011) Date FE Y N N N Date × NAICS4 FE N Y N N Date × CBSA FE N N Y N Date × NAICS4 × CBSA FE N N N Y N 887196 1127309 881053 839071 R2 0.005 0.027 0.018 0.258 69

Table C17: Cross-Sectional Determinants of Firms’ IPO Decisions (Firm Fixed Effects) This table tests which firm characteristics predict whether firms go public within the next three years, controlling for firm fixed effects. The dependent variable IPO is a dummy that equals one if the firm goes public in the next three years, which we multiply by 100. Standard errors are shown below the parameter estimates in parentheses and are clustered by firm. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. IPO (1) (2) (3) (4) Log(1+Sales) 0.240*** 0.238*** 0.251*** 0.253*** (0.019) (0.018) (0.020) (0.023) Sales Growth 0.507*** 0.313*** 0.491*** 0.284*** (0.058) (0.041) (0.060) (0.051) Debt/Assets 0.081 0.279*** 0.123* 0.295*** (0.061) (0.063) (0.064) (0.084) Financing Deficit 0.673*** 0.637*** 0.679*** 0.582*** (0.080) (0.073) (0.081) (0.096) Capex/Assets 0.154 -0.106 0.183 -0.079 (0.156) (0.124) (0.163) (0.190) NAICS4 MTB 0.109*** 0.109*** (0.014) (0.014) Date FE Y N N N Firm FE Y Y Y Y Date × NAICS4 FE N Y N N Date × CBSA FE N N Y N Date × NAICS4 × CBSA FE N N N Y N 967151 1226028 960914 919120 R2 0.007 0.031 0.020 0.251 70

Table C18: Going Public and Firms’ Borrowing Costs (Controlling for Syndication) This table tests whether firms’ borrowing costs drop after the IPO by estimating the same regressions in Table 6, but includes an additional control for whether the loan is syndicated. The sample includes only new loans. Standard errors are shown below the parameter estimates in parentheses and are clustered by firm. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. Interest Rate (%) (1) (2) (3) (4) IPO Firm × Post -0.485** -0.480** -0.470*** -0.432** (0.203) (0.190) (0.178) (0.186) Log(Assets) -0.070*** -0.070*** -0.056** -0.063** (0.027) (0.025) (0.024) (0.026) Sales Growth 0.028 0.038 0.050 0.058 (0.037) (0.036) (0.034) (0.040) Debt/Assets 0.431*** 0.431*** 0.399*** 0.355*** (0.092) (0.081) (0.080) (0.092) EBITDA/Assets -0.394*** -0.329*** -0.348*** -0.260*** (0.100) (0.077) (0.076) (0.073) Capex/Assets 0.206* 0.153 0.159 0.189* (0.114) (0.106) (0.098) (0.112) PD (%) -0.002 (0.012) LGD (%) 0.001 (0.001) PD × LGD (%) 0.142*** (0.036) Date FE Y Y Y Y Syndication FE Y Y Y Y Bank/Firm FE Y Y Y Y Bank/Date FE N N Y Y Loan Controls N Y Y Y N 37029 35822 35771 29723 R2 0.803 0.866 0.877 0.880 71