SEARCH

SEARCH BY CITATION

Keywords:

  • going public;
  • initial public offering;
  • product market competition;
  • confidential information

Abstract

  1. Top of page
  2. Abstract
  3. 1. INTRODUCTION
  4. 2. LITERATURE AND HYPOTHESES
  5. 3. DATA AND METHODOLOGY
  6. 4. EMPIRICAL RESULTS
  7. 5. CONCLUSION
  8. REFERENCES

Abstract:  This paper investigates the effect of product market characteristics on the decision to go public. When firms decide to go public or remain private, they trade off product market related costs and benefits. Costs arise from the loss of confidential information to competitors, e.g., in the IPO prospectus and subsequent mandated public disclosures, while benefits emerge from raising capital allowing the firm to strengthen its position in the product market. Our results show that UK firms are more likely to go public when they operate in a more profitable industry and in an industry with lower barriers to entry. These firms are more likely to go public in order to improve their position in the product market and to deter new entrants into the industry. However, firms from more competitive industries and firms with smaller market share are less likely to go public. For these firms the loss of confidential information to rivals outweighs the benefits of going public.


1. INTRODUCTION

  1. Top of page
  2. Abstract
  3. 1. INTRODUCTION
  4. 2. LITERATURE AND HYPOTHESES
  5. 3. DATA AND METHODOLOGY
  6. 4. EMPIRICAL RESULTS
  7. 5. CONCLUSION
  8. REFERENCES

The decision to go public is one of the most important decisions in the life of a firm (Jain and Kini, 1999). Going public not only offers the advantage of lower-cost equity financing (Chemmanur and Fulghieri, 1999; and Maksimovic and Pilcher, 2001) but also allows the firm to improve its position in the product market at the expense of its competitors that stay private (Chemmanur and He, 2011; and Chod and Lyandres, 2010). These product market benefits arise because shareholders in public firms are better able to diversify risk and tolerate higher idiosyncratic profit variability than owners of private firms (Shah and Thakor, 1988). This enables public firms to commit to riskier and more aggressive product market strategies than otherwise similar private firms.

However, firms that go public also face a loss of confidential information, because public firms have to disclose their strategy, profitability and product innovations, which is valuable information for their product market rivals (Campbell, 1979; Bhattacharya and Ritter, 1983; Yosha, 1995; and Marra and Suijs, 2004). This information allows competitors to copy more easily the firm's product innovation (Maksimovic and Pichler, 2001; and Spiegel and Tookes, 2008). Moreover, the decision to go public itself signals higher product quality to competitors (Stoughton et al., 2001). The decision to go public therefore depends on product market characteristics, and the position of the firm within the product market.

In this paper we investigate the impact of ex ante product market characteristics of a firm on the decision to go public. Our sample consists of 337 firms that went public on the Official List of the London Stock Exchange during the period 1994–2006 and a sample of 18,386 firms that decided to stay private but were eligible to go public. Our study contributes to the literature in two ways. First, there are few empirical studies of the decision to go public, mainly due to the difficulty of obtaining information about private firms (Ritter and Welch, 2002). Existing studies therefore involve one industry (Lerner, 1994) or small samples dominated by equity carve-outs (Pagano et al., 1998) or large, highly leveraged firms (Helwege and Packer, 2009). Our study adds to this literature by providing the first large sample study on the decision to go public in the United Kingdom.

Second, this paper is one of the first to empirically test theoretical models that provide product market explanations for going public. The only other paper we are aware of is that of Chemmanur et al. (2010) who show, for a sample of US firms, that a firm's market share, productivity and industry concentration have positive effects on the likelihood of going public. The authors conclude that these results are consistent with confidential information theories. However, their sample is limited to manufacturing firms and therefore their conclusions may not apply to services firms (including service-oriented high technology firms). In this paper, we investigate a sample of UK firms from both manufacturing and services industries. We have access to audited financial statement data of private firms that are obliged to file financial information with Companies House and are eligible to go public on the Official List of the London Stock Exchange.

We confirm the US results of Chemmanur et al. (2010) using our UK sample and find that, after controlling for other determinants of going public, confidential information theories are important for explaining the decision to go public. However, this paper also extends their work by investigating the impact of industry margins and entry barriers on the decision to go public. We report that firms in more profitable industries and firms from industries with lower entry barriers are more likely to seek a public listing. We argue that firms in these industries are more likely to benefit from an Initial Public Offering (IPO) which enables them to pursue more aggressive product market strategies to deter entry into their profitable industry and to grab market share from their private competitors. This finding is consistent with the theoretical models of Chod and Lyandres (2010) and Chemmanur and He (2011). We also provide interesting results for other variables. For example, Pagano et al. (1998) and Baker and Wurgler (2002) argue that companies time financial decisions and may go public to exploit high share valuations. We indeed find that companies are more likely to go public when the equity valuation of their industries is relatively high, which suggest that companies take advantage of this window of opportunity.

We further examine the product market related effects of going public in the years directly following the IPO. Firms that go public increase their market share and operational risk in the post-IPO period to the year before the IPO. We also find that firms increase their scale of operations in the post-IPO years. IPO-firms significantly increase (in real value terms) their capital expenditures, fixed assets, total assets, sales and profitability in comparison to the year before the IPO. These findings are consistent with theories that predict that firms go public to increase the scale of their investments (Clementi, 2002) and/or to engage in riskier product market strategies to improve their position in the product market (Shah and Thakor, 1988; Chod and Lyandres, 2010; and Chemmanur and He, 2011).

The remainder of this paper is organized as follows. In Section 2 we survey the literature and develop our hypotheses. Section 3 describes our data and methodology. In Section 4 we present our empirical results. Section 5 concludes our study.

2. LITERATURE AND HYPOTHESES

  1. Top of page
  2. Abstract
  3. 1. INTRODUCTION
  4. 2. LITERATURE AND HYPOTHESES
  5. 3. DATA AND METHODOLOGY
  6. 4. EMPIRICAL RESULTS
  7. 5. CONCLUSION
  8. REFERENCES

There exists only a handful of studies that empirically investigate why firms decide to go public. Lerner (1994) studies the timing of going public of privately held venture-backed biotechnology firms in the United States. He reports that biotech firms decide to go public when equity valuations are high. Pagano et al. (1998) show that the industry market-to-book ratio and firm size positively impact the likelihood of going public using a sample of 5,350 Italian firms, of which only 69 went public. More recent evidence comes from Helwege and Packer (2009) who investigate a sample of 181 large, highly levered US firms that continue to file with the SEC because they had public bonds outstanding before the IPO. They find that family-controlled firms and firms with greater managerial autonomy are less likely to go public because they want to protect their private benefits of control. However, none of these studies examines the role of product market characteristics on the decision to go public.

This lack of empirical work contrasts with the extensive theoretical literature related to the IPO decision that focuses on explanations based on the product market. Firms require access to external capital to exploit valuable opportunities in their product markets once their internal financing is used up. External capital can be attracted from either public or private capital markets. A general assumption is that public capital is less expensive, because of improved diversification options and increased liquidity (Chemmanur and Fulghieri, 1999; Maksimovic and Pilcher, 2001; and Boehmer and Ljungqvist, 2004). However, the fixed cost of publicly issuing securities is greater which makes going public more likely for firms with large scale investment projects (Clementi, 2002). Investing on a large scale offers the opportunity to expand product market positions more effectively and to deter new entrants to the industry.

Chemmanur and He (2011) develop a theoretical model and argue that firms that go public can seize market share from their private rivals or prevent entry by new firms. They mention several mechanisms through which public firms can improve their position in the product market: they gain additional credibility with customers and suppliers; they can hire and retain better quality staff by compensating them via stock and stock options; and they are able to acquire other firms in the same industry using their own publicly traded stock as an acquisition currency. Shah and Thakor (1988) and Chod and Lyandres (2010) theoretically show that public firms can pursue more aggressive product market strategies because their shareholders put up with more idiosyncratic profit volatility than owners of private companies. Firms that go public are therefore able to credibly commit to a more aggressive product market strategy which helps to deter new entrants to their industry.

We therefore hypothesize that firms from more profitable industries are more likely to go public. These firms have most to gain from grabbing market share from their private competitors and from deterring further entry into the industry which could erode their profitability.

  • H1a: Firms from more profitable industries are more likely to go public.

We also conjecture that companies from industries with low barriers to entry are more likely to go public in order to adopt more aggressive product market strategies that will deter new entrants.

  • H2a: Firms from industries with low entry barriers are more likely to go public.

Alternatively, confidential information theories predict that firms in more profitable industries and industries with low barriers to entry are less likely to go public. Going public forces the firm to inform investors publicly, thereby running the risk of simultaneously informing competing firms. Ball and Shivakumar (2005 and 2008) support this view and show that public UK firms supply higher quality financial reports demanded by public investors, because these investors face higher information asymmetry than private investors. Moreover, public firms have to disclose their strategy, profitability and product innovations, which is valuable information for their product market rivals and for new entrants to the industry. The confidential information allows competitors to copy the firm's product innovation more easily (Maksimovic and Pichler, 2001; and Spiegel and Tookes, 2008). This loss of confidential information imposes larger costs on firms going public in industries that are more profitable and that are easier to enter. We therefore also formulate two alternative hypotheses.

  • H1b: Firms from more profitable industries are less likely to go public.

  • H2b: Firms from industries with low entry barriers are less likely to go public.

The confidential information theories of Bhattacharya and Ritter (1983) and Marra and Suijs (2004) also predict that the loss of confidential information is more of a concern when product markets are more competitive. We therefore hypothesize that firms that are active in a more concentrated (i.e., less competitive) industry are more likely to go public.

  • H3a: Firms from less competitive industries are more likely to go public.

On the other hand, firms in less competitive industries stand to benefit less from going public. Public firms are able to adopt more risky and aggressive product market strategies than otherwise similar private firms which enable them to expand their market share (Chemmanur and He, 2011; and Chod and Lyandres, 2010). These benefits are positively related to the level of product market competition in the industry. Firms from less competitive industries are therefore less likely to go public.

  • H3b: Firms from less competitive industries are less likely to go public.

We also expect that firms with a higher market share have a higher propensity of going public. These firms are established industry players that are expected to be less concerned with the loss of confidential information (Chemmanur et al., 2010). Guo et al. (2004) show that biotech firms that go public disclose more product-related information at a time when they have secured patent protection and the firm is at a later stage of development. This suggests that firms with more secure product market positions are less concerned with a loss of confidential information. Hence:

  • H4a: Firms with higher market shares are more likely to go public.

At the same time, firms with higher market shares may be less likely to go public. Firms with smaller market shares have more to gain from going public, as it allows them to increase their market share and to strengthen their position in the product market through aggressive product market strategies (Chemmanur and He, 2011; and Chod and Lyandres, 2010). We therefore conjecture that the firm's market share and the likelihood of going public are inversely related.

  • H4b: Firms with higher market shares are less likely to go public.

3. DATA AND METHODOLOGY

  1. Top of page
  2. Abstract
  3. 1. INTRODUCTION
  4. 2. LITERATURE AND HYPOTHESES
  5. 3. DATA AND METHODOLOGY
  6. 4. EMPIRICAL RESULTS
  7. 5. CONCLUSION
  8. REFERENCES

(i)Data Sources and Sample

The main source of data used in this study is the FAME database, which provides financial and other information on approximately 3.4 million UK companies.1 The FAME database contains detailed financial information and industry classifications for both public and private companies. For our analysis, we use three sets of companies. The first set contains firms that went public on the Official List of the London Stock Exchange in the period 1994–2006. The second set consists of firms that were in a position to go public, but did not do so during this period. Our third set comprises almost all UK firms that are registered in the FAME database in every year during the period of our research.

To create our set of IPO firms we started with all 456 UK companies that went public between 1994 and 2006 on the Official List and that did not previously have a listing on another stock market.2 We include both public offers and placings (Goergen et al., 2006). We identify IPO firms from the SDC New Issues database and from New Issues Statistics provided to us by the London Stock Exchange. We exclude 56 financial companies because their accounting information is difficult to compare with that of non-financial companies. Another 74 firms are omitted due to incomplete information. Our final sample comprises 337 IPO firms.

The second set consists of companies that were in a position to go public in the period of study. The selection of these companies is based on the availability of audited financial statements. Companies with annual sales in excess of £350,000 before 2000 and £1,000,000 thereafter are required to file audited financial statements at Companies House. This ensures that the rules concerning financial reporting are the same for the public and private companies in our sample. Firms are assumed to be eligible to go public in every year that they fulfill these requirements in the period 1994–2006. We do not use other selection criteria. The London Stock Exchange is flexible in applying its listing requirements and allows companies to go public without its formally required three-year track record.3 Later we test the robustness of our results in relation to the possible selection bias induced by our selection criterion (see Section 4(iv)). In addition, we check whether our results change when we add the condition that private firms have a five-year trading record. The sample of potential IPO firms contains 192,514 observations for the period 1994–2006, corresponding, on average, to more than 14,800 potential IPO firms per year.

Finally, the third set of firms consists of all firms with a registered office in the UK that fulfill the following criteria: (i) qualified independent, (ii) non-financial, (iii) private or public, (iv) active or inactive, and (v) turnover and operating profit available. We use this set of firms to estimate our product market characteristics. This set comprises 1,428,727 observations, which implies an average of 109,902 firms per year.

(ii)Empirical Model

In order to assess whether our product market characteristics and control variables can discriminate between firms that go public and eligible firms that remain private we test logistic regressions using the following specification:

  • image(1)

where yijt is the logit of the odds that firm i from industry j has gone public in period t; INDMARGINjt-1 is the weighted average of the ratio of operating profit over sales for industry j in period t−1; CAPINTjt−1 is a dummy that equals 1 if both the ratio of property, plant and equipment over number of employees in industry j and the ratio of net sales over number of firms in industry j exceed the median industry values in period t−1, otherwise 0; HERFjt−1 is the Herfindahl index of industry j in period t−1; MSHAREit−1 is the market share of firm i in period t−1; CAPEXit−1 is the ratio of capital expenditure over property, plant and equipment of firm i in period t−1; GROWTHit−1 is the growth in net sales of firm i in period t−1; LNSALESit−1 is the natural logarithm of total sales of firm i in period t−1; LNAGEit is the natural logarithm of firm i's age in period t; LEVERAGEit−1 is the debt ratio of firm i in period t−1; BTMjt−1 is the median book-to-market value of equity of firms in the same industry j which are traded on the Official List of the London Stock Exchange in period t−1; INDEXjt−1 is the buy-and-hold return of the FTSE UK All Share ICB industry Index for industry j over the year t−1; PROFITit−1 is earnings before interest and taxes over sales of firm i in period t−1; OWNit−1 is the percentage of shares held by senior managers and other insiders in firm i in period t−1 and OWNit−1* OWNit−1 is its square; and YEARt is a calendar year dummy. Industries are defined at the three-digit level of the SIC industry classification code.4 IPO firms are excluded from the sample after the year they went public. Table 1 provides more detailed definitions of the variables.

Table 1.  Variable Definitions
VariableDescriptionSource
  1. Notes: Industries are classified by the three-digit SIC code and firms are assigned to the industry in which they had the largest proportion of sales.

Product market characteristics
 INDMARGINIndustry margin computed as the weighted average margin of firms within industry j using the firm's market share as weights. In order to correct for extreme values, we excluded the 5% lowest and highest profit margins in an industry before computing the weighted industry margin. We measure industry margin in industry j as: inline image where MARGINi is firm i's profit margin defined as earnings before interest and taxes (EBIT) divided by sales; and MSHAREi is firm i's market share.FAME
 CAPINTCapital intensity captured as a dummy variable that equals 1 if both the ratio of property, plant and equipment over number of employees and the ratio of net sales over number of firms in an industry exceed the median values of these ratios across all industries.FAME
 HERFHerfindahl index measured as the sum of the squared market shares in industry j. We measure the Herfindahl index as: inline image where MSHAREi is firm i's market share.FAME
 MSHAREMarket share defined as the ratio of the firm's sales over total sales in the firm's industry.FAME
Control variables  
 CAPEXCapital expenditures over property, plant and equipment.FAME, IPO prospectus
 GROWTHGrowth in net sales over the year.FAME, IPO prospectus
 SALESTotal sales (in millions of pounds).FAME, IPO prospectus
 AGEYear of observation minus the year the company was founded.FAME, IPO prospectus
 LEVERAGEBook value of long term debt plus short term loans and overdrafts divided by book value of total assets.FAME, IPO prospectus
 BTMMedian book-to-market value of equity of firms in the same industry which traded on the London stock exchange. In order to correct for extreme values, we excluded the 1% lowest and highest book-to-market values.FAME, IPO prospectus
 INDEXBuy-and-hold return of the FTSE UK All Share ICB industry Index during the past year. In order to correct for extreme values, we excluded the 1% lowest and highest buy-and-hold returns.Datastream
 PROFITOperating profit over sales.FAME, IPO prospectus
 OWNPercentage of shares held by senior management and other insiders.FAME, IPO prospectus

In any year t, the sample consists of firms that went public in that year and private firms that were eligible to go public. As a result, our sample contains several (year) observations of a single firm, since private firms can be eligible IPO candidates for more than one year during the period of research. To control for dependencies in the standard errors of the firm year observations in our sample that relate to a single firm, we use a logistic regression technique that clusters such related observations.5

(a) Product Market Characteristics

Our regression model includes industry margin (INDMARGIN) to account for differences in industry profitability and to test our first hypothesis. We also measure how difficult it is for new firms to enter the industry in order to test hypothesis two. Main barriers to entry are the capital, marketing and R&D intensity of an industry. However, the available data does not allow us to measure marketing and R&D intensity. We therefore measure the barriers to entry using a dummy variable (CAPINT) that is one if both the average value of the ratio of property, plant, and equipment over number of employees in an industry and the average sales per firm in an industry exceed the median values of these two measures across all industries in the sample. We use this dummy variable because both the relative investments in plant, property and equipment and the firm's absolute size can indicate the capital intensity of an industry. To test our third hypothesis, we measure the level of competition in the industry by the Herfindahl index (HERF). The higher the Herfindahl index the lower the level of competition in the industry. We use the market share of the firm (MSHARE), defined as the sales of the firm over the sum of all sales in the firm's industry, to test hypothesis four.

(b) Control Variables

Firms do not only go public to improve their position in the product market. We, therefore, control for other possible determinants of going public. Firms that invest heavily and grow fast need capital and therefore are more likely to seek external financing (Jain and Kini, 1999). We measure current investments as capital expenditures over property, plant and equipment (CAPEX), and growth as one-year growth in sales (GROWTH). Both variables are capped at 1 (and CAPEX also at −1), to prevent (small) companies with extreme investments or growth distorting our findings.

An IPO involves considerable fixed costs, such as fees for the underwriters, auditors and lawyers, as well as stock exchange fees. The fixed costs suggests that the likelihood of an IPO should be positively associated with size (Pagano et al., 1998). We measure size as the natural logarithm of net sales (LNSALES). We include the age of the firm as a proxy for information costs. Investors more heavily discount prices for shares in firms for which less information is available (Leland and Pyle, 1977). Such information costs are more likely to affect the going public decision of relatively small and young companies, which have a limited track record and low visibility (Chemmanur and Fulghieri, 1999). This may mean that public financing is less attractive for younger firms. We measure age as the natural logarithm of the age of the firm (LNAGE).

A high debt burden may force a firm to rebalance its capital structure by issuing (public) equity. We use the debt ratio, which is defined as book value of long term debt plus short term loans and overdrafts over total assets, as a proxy for leverage (LEVERAGE). Companies may go public to exploit high valuations of their shares in certain periods (Pagano et al., 1998; Baker and Wurgler, 2002; Benninga et al., 2005; and Gregory et al., 2010). To control for the possibility that firms time their IPOs immediately following a strong run-up in stock prices in their industry, we include in our empirical model the median industry book-to-market ratio (BTM) and the buy-and-hold return of the FTSE UK All Share ICB industry index during the previous year.

As Pagano et al. (1998) argue, firm profitability can affect the likelihood of an IPO in two opposite ways. First, tapping public equity markets is an infrequent event. Firms are, therefore, likely to time an IPO in a period when their prospects are favorable and their profits are high. Second, more profitable firms can more easily finance their projects internally and therefore have less need for external capital. This implies a negative relationship between profitability and the likelihood of an IPO. We measure profitability (PROFIT) as the earnings before interest and taxes scaled by sales.

Managerial incentives in privately owned companies can also influence the decision to go public (Wagner, 2010). For example, if managers of a private firm have a relatively large ownership stake, then there might be a strong incentive to ‘cash out' this investment by taking the firm public. We therefore include the percentage of shares held by senior management and other insiders (OWN) in our empirical model. Because the effect of insider ownership on the IPO decision might be nonlinear (i.e., insider ownership might only play a role when the insiders’ stake is very large) we also include the square of insider ownership (OWN*OWN) in the regression.6

4. EMPIRICAL RESULTS

  1. Top of page
  2. Abstract
  3. 1. INTRODUCTION
  4. 2. LITERATURE AND HYPOTHESES
  5. 3. DATA AND METHODOLOGY
  6. 4. EMPIRICAL RESULTS
  7. 5. CONCLUSION
  8. REFERENCES

In this section we present our empirical results. We start with descriptive statistics (4(i)) followed by a presentation of our logistic regression results (4(ii)). We continue with a discussion of interaction effects (4(iii)), robustness checks (4(iv)) and an analysis of product market related effects of going public in the post-IPO period (4(v)).

(i)Univariate Comparisons

Table 2 reports summary statistics for the sample of IPO and potential IPO firms. The univariate comparisons of the means and medians of these samples provide a first indication of factors that discriminate between IPO and private firms. All variables, except firm age, are lagged one period so as to align the firm and industry characteristics more closely with the period when the decision to go public or stay private was presumably made.

Table 2.  Summary Statistics
VariableMeanTest of DifferenceMedianTest of DifferenceStandard DeviationMinimumMaximum
IPOEligibleIPOEligibleIPOEligibleIPOEligibleIPOEligible
  1. Notes:

  2. This table reports summary statistics for the sample of firms that went public and the firms that were eligible to go public on the Official List of the London Stock Exchange between 1994 and 2006. All variables are one year lagged except for AGE. The sample of eligible firms consists of firms that file audited financial statements with Companies House (i.e., net sales in excess of £350,000 before 2000 and £1,000,000 thereafter) for every year they fulfill this requirement in the period 1994–2006. The sample consists of 75,708 firm-year observations of which 337 are IPOs. Managerial ownership data is only available for 46,271 firm-year observations of which 329 are IPOs. INDMARGIN denotes industry margin computed as the weighted average profit margin of firms within industry j using the firm's market share as weights. We measure industry margin in industry j as:

    • image
  3. where MARGINi is firm i's profit margin defined as earnings before interest and taxes (EBIT) divided by sales; and MSHAREi is firm i's market share. CAPINT is capital intensity measured as a dummy variable that equals 1 if both the ratio of property, plant and equipment over number of employees and the ratio of net sales over number of firms in an industry exceed the median values of these ratios across all industries. HERF is the Herfindahl index measured as the sum of the squared market shares in industry j. We measure the Herfindahl index as:

    • image
  4. where MSHAREi is firm i's market share. MSHARE refers to market share defined as the ratio of the firm's sales over total sales in the firm's industry. CAPEX is capital expenditures over property, plant and equipment. GROWTH denotes growth in net sales over the year. SALES equals total sales (in millions of pounds). AGE is the year of observation minus the year the company was founded. LEVERAGE is measured as book value of long term debt plus short term loans and overdrafts divided by book value of total assets. BTM is the median book-to-market value of equity of firms in the same industry which traded on the London stock exchange. INDEX refers to the buy-and-hold return of the FTSE UK All Share ICB industry Index during the past year. PROFIT is operating profit over sales. OWN equals the percentage of shares held by senior management and other insiders. Industries are classified by the three-digit SIC code and firms are assigned to the industry in which they had the largest proportion of sales. All ratio variables are winsorized to be no greater than one in absolute value, except for BTM, INDEX and INDMARGIN. BTM and INDEX are winsorized at the first and 99th percentile and INDMARGIN excludes profit margins that are less (greater) than the fifth (95th) percentile. One, two and three asterisks indicate that the coefficients are significantly different from zero at the 10%, 5% and 1% level or better, respectively.

INDMARGIN0.0640.044(−5.87)***0.0640.043(−8.09)***0.0620.039−0.429−0.5440.2670.299
CAPINT0.2700.298(1.12)0.0000.000(1.120)0.4450.4570.0000.0001.0001.000
HERF0.2270.120(−8.64)***0.1380.048(−10.04)***0.2270.1570.0030.0020.9050.993
MSHARE0.0890.006(−7.29)***0.0150.001(−19.93)***0.2080.0250.0000.0001.0001.000
CAPEX0.3940.107(−13.83)***0.3400.015(−19.09)***0.3800.239−1.000−1.0001.0001.000
GROWTH0.3360.122(−10.42)***0.2460.070(−13.24)***0.3760.280−0.986−0.9811.0001.000
SALES140.31511.893(−5.28)***24.6474.600(−17.60)***446.68335.6870.0030.3515,286.0003,406.761
AGE32.11630.001(−0.85)14.00024.000(9.46)***45.79320.3341.0004.000277.000141.000
LEVERAGE0.3220.192(−7.91)***0.2460.147(−7.28)***0.3010.1840.0000.0001.0000.979
BTM0.7711.234(18.60)***0.6891.095(14.72)***0.4550.6910.2160.2164.1804.180
INDEX0.1660.035(−9.64)***0.1800.077(−8.77)***0.2170.247−0.663−0.6630.6320.632
PROFIT−0.0240.043(3.36)***0.0890.032(−9.69)***0.3640.093−1.000−1.0000.4671.000
OWN42.33377.472(19.16)***36.27099.000(18.23)***33.16931.4660.0000.000100.000100.000

Our sample comprises 203 different 3-digit US SIC industries (not tabulated). The 337 IPOs in our sample are well-distributed over the industries, as 95 out of the 203 industries have at least one IPO. The Services-Computer Programming, Data Processing (SIC 737) industry has the largest number of IPOs (61 firms). At the industry level, we find that both the average and median weighted margins (INDMARGIN) are significantly higher in the industries which have IPOs, indicating that the IPO firms operate in more profitable industries than the eligible firms. Our measure of barriers to entry into the industry (CAPINT) is not significantly different between the firms that go public and those that remain private. However, the Herfindahl index (HERF) exhibits a significant difference: firms that go public operate in more concentrated (i.e., less competitive) industries. Furthermore, firms that go public have significantly higher market shares (MSHARE) than firms that decide to stay private.

Table 2 also reveals that firms that go public have larger capital expenditures (CAPEX), faster one-year sales growth (GROWTH), more sales (SALES) and higher leverage (LEVERAGE) compared to firms that remain private. In terms of the median, IPO firms are younger than firms that stay private. The industry book-to-market ratio (BTM) of IPO firms is significantly lower than for private firms. This indicates that firms go public in industries with higher equity valuations (i.e., higher market-to-book ratios). Similarly, the industry buy-and-hold return (INDEX) is also significantly higher for IPO firms than for private companies, further supporting the idea that firms time their access to public equity markets. The mean and median profitability (PROFIT) for IPO firms are respectively −2.4% and 8.9%, while for eligible firms the mean and median are 4.3% and 3.2%. Because of the skewed distribution for IPO firms, the mean and median tests yield opposite results. Also note that 15.8% (11,942) of all observations concern firm-years with negative operating profit. We include these loss reporting firms in our sample as the London Stock Exchange sometimes allows loss making firms to go public without the formally required three-year trading record. Share ownership by senior management and other insiders (OWN) is less concentrated for IPO firms than for firms that remain private. Table 3 shows the correlations between the variables we use in the subsequent regression analyses. The most extreme correlation is the one between the variables OWN and LNSALES. However, multi-collinearity is not a problem for the regression analyses per se. A standard way to gauge whether multi-collinearity may cause problems is to calculate Variance Inflation Factors (VIF). The VIF provides an index that measures how much the variance of an estimated regression coefficient is increased because of collinearity. A common rule is that a VIF higher than 5 indicates high multi-collinearity. We find that none of the VIFs is higher than 5 indicating no severe multi-collinearity problem.7

Table 3.  Correlation Matrix
 INDMARGINCAPINTHERFMSHARECAPEXGROWTHLNSALESLNAGELEVERAGEBTMINDEXPROFITOWN
  1. Notes:

  2. This table reports Pearson correlations for the variables used in our logistic regressions. All variables are one year lagged except for LNAGE. The sample of eligible firms consists of firms that file audited financial statements with Companies House (i.e., net sales in excess of £350,000 before 2000 and £1,000,000 thereafter) for every year they fulfill this requirement in the period 1994–2006. The sample consists of 75,708 firm-year observations of which 337 are IPOs. Managerial ownership data is only available for 46,271 firm-year observations of which 329 are IPOs INDMARGIN denotes industry margin computed as the weighted average profit margin of firms within industry j using the firm's market share as weights. We measure industry margin in industry j as:

    • image
  3. where MARGINi is firm i's profit margin defined as earnings before interest and taxes (EBIT) divided by sales; and MSHAREi is firm i's market share. CAPINT is capital intensity measured as a dummy variable that equals 1 if both the ratio of property, plant and equipment over number of employees and the ratio of net sales over number of firms in an industry exceed the median values of these ratios across all industries. HERF is the Herfindahl index measured as the sum of the squared market shares in industry j. We measure the Herfindahl index as:

    • image
  4. where MSHAREi is firm i's market share. MSHARE refers to market share defined as the ratio of the firm's sales over total sales in the firm's industry. CAPEX is capital expenditures over property, plant and equipment. GROWTH denotes growth in net sales over the year. LNSALES equals the natural logarithm of total sales (in millions of pounds). LNAGE is the natural logarithm of the year of observation minus the year the company was founded. LEVERAGE is measured as book value of long term debt plus short term loans and overdrafts divided by book value of total assets. BTM is the median book-to-market value of equity of firms in the same industry which traded on the London stock exchange. INDEX refers to the buy-and-hold return of the FTSE UK All Share ICB industry Index during the past year. PROFIT is operating profit over sales. OWN equals the percentage of shares held by senior management and other insiders. Industries are classified by the three-digit SIC code and firms are assigned to the industry in which they had the largest proportion of sales. All ratio variables are winsorized to be no greater than one in absolute value, except for BTM, INDEX and INDMARGIN. BTM and INDEX are winsorized at the first and 99th percentile and INDMARGIN excludes profit margins that are less (greater) than the fifth (95th) percentile. p-values are reported in parentheses. One, two and three asterisks indicate that the coefficients are significantly different from zero at the 10%, 5% and 1% level or better, respectively.

INDMARGIN1.0000            
CAPINT−0.1252***1.0000           
 (<0.0001)            
HERF0.1705***0.2027***1.0000          
 (<0.0001)(<0.0001)           
MSHARE−0.0033**−0.00330.0612***1.0000         
 (0.0184)(0.3604)(<0.0001)          
CAPEX0.0066−0.00470.0298***0.0421***1.0000        
 (0.3610)(0.1963)(<0.0001)(<0.0001)         
GROWTH0.0273***−0.0596***−0.0301***0.00110.0918***1.0000       
 (<0.0001)(<0.0001)(<0.0001)(0.7592)(<0.0001)        
LNSALES−0.0916***0.2437***0.0191***0.2677***0.2227***0.00701.0000      
 (<0.0001)(<0.0001)(<0.0001)(<0.0001)(<0.0001)(0.0534)       
LNAGE−0.0170***0.1398***0.0554***0.0750***−0.0200***−0.2332***0.2134***1.0000     
 (<0.0001)(<0.0001)(<0.0001)(<0.0001)(<0.0001)(<0.0001)(<0.0001)      
LEVERAGE0.0519***0.0690***0.0141***0.0158***−0.0282***0.00150.0523***−0.0412***1.0000    
 (<0.0001)(<0.0001)(0.0001)(<0.0001)(<0.0001)(0.6753)(<0.0001)(<0.0001)     
BTM−0.2138***0.1585***−0.1323***0.0533***−0.0556***−0.0524***0.0684***0.1042***0.0414***1.0000   
 (<0.0001)(<0.0001)(<0.0001)(<0.0001)(<0.0001)(<0.0001)(<0.0001)(<0.0001)(<0.0001)    
INDEX0.1503***0.0219***0.0130***0.0206***0.0030−0.00040.0327−0.0108***0.0156***−0.1323***1.0000  
 (<0.0001)(<0.0001)(0.0004)(<0.0001)(0.4165)(0.9200)(<0.0001)(0.0029)(<0.0001)(<0.0001)   
PROFIT0.1336***−0.0798***−0.00170.0076**−0.00470.1626***−0.0868***−0.1138***−0.0140***−0.0401***0.0194***1.0000 
 (<0.0001)(<0.0001)(0.6319)(0.0357)(0.1920)(<0.0001)(<0.0001)(<0.0001)(0.0001)(<0.0001)(<0.0001)  
OWN0.0662***−0.2185***−0.0118**−0.2097***−0.2090***−0.0533***−0.8363***−0.1582***−0.0592***−0.0654***−0.018***0.0557***1.0000

(ii)Regression Results

Table 4 shows the results of our logistic regressions. Model (1) includes the industry margin (INDMARGIN) and control variables that are predicted to influence the IPO decision that are derived from the previous literature (see Section 3(ii)(b)). We find that firms are more likely to seek a public listing when their industry is more profitable. A one standard deviation increase in the weighted average industry margin increases the sample average probability of an IPO with about 47%.8 This finding is in line with hypotheses 1a. Firms from profitable industries have an incentive to go public in order to strengthen their position in the product market and to deter further entry into their industry. These firms are less concerned with the loss of confidential information to rivals.

Table 4.  The Determinants of the Decision to Go Public
Variables123456
  1. Notes:

  2. This table reports the results of the estimated logistic regression model on the association between the likelihood of going public and firm and industry specific product market variables, controlling for other variables. It uses the following equation (see also equation (1) in Section 3(ii)):

    • image
  3. The dependent variable is 1 for IPO-firms and zero otherwise. An IPO-firm is excluded from the sample after the year it went public. In any year the sample consists of firms that went public in that year and private firms that are eligible to go public. INDMARGIN denotes industry margin computed as the weighted average profit margin of firms within industry j using the firm's market share as weights. We measure industry margin in industry j as:

    • image
  4. where MARGINi is firm i's profit margin defined as earnings before interest and taxes (EBIT) divided by sales; and MSHAREi is firm i's market share. CAPINT is capital intensity measured as a dummy variable that equals 1 if both the ratio of property, plant and equipment over number of employees and the ratio of net sales over number of firms in an industry exceed the median values of these ratios across all industries.

  5. HERF is the Herfindahl index measured as the sum of the squared market shares in industry j. We measure the Herfindahl index as:

    • image
  6. where MSHAREi is firm i's market share. MSHARE refers to market share defined as the ratio of the firm's sales over total sales in the firm's industry. CAPEX is capital expenditures over property, plant and equipment. GROWTH denotes growth in net sales over the year. LNSALES equals the natural logarithm of total sales (in millions of pounds). LNAGE is the natural logarithm of the year of observation minus the year the company was founded. LEVERAGE is measured as book value of long term debt plus short term loans and overdrafts divided by book value of total assets. BTM is the median book-to-market value of equity of firms in the same industry which traded on the London stock exchange. INDEX refers to the buy-and-hold return of the FTSE UK All Share ICB industry Index during the past year. PROFIT is operating profit over sales. OWN equals the percentage of shares held by senior management and other insiders. Industries are classified by the three-digit SIC code and firms are assigned to the industry in which they had the largest proportion of sales. All ratio variables are winsorized to be no greater than one in absolute value, except for BTM, INDEX and INDMARGIN. BTM and INDEX are winsorized at the first and 99th percentile and INDMARGIN excludes profit margins that are less (greater) than the fifth (95th) percentile. We use lagged values of all the variables except for LNAGE to align the firm and industry characteristics with the period of time the decision to go public was presumably made. The regression includes a constant and calendar year dummies (not reported) and standard errors are adjusted for clustering in firms. One, two and three asterisks indicate that the coefficients are significantly different from zero at the 10%, 5% and 1% level or better, respectively. The joint-test which indicates that all coefficients are zero is rejected by the likelihood ratio test at a level lower than 0.001 for all models. For goodness-of-fit we report the pseudo R2 and the Coefficient of Concordance (CoC): the percentage of randomly drawn pairs of one IPO firm and one eligible IPO firm for which it is true that the likelihood estimate of the model is higher for the IPO firm.

INDMARGIN9.87***   7.32***1.89
 (3.98)   (3.03)(0.85)
CAPINT −0.69**  −1.01***−0.72***
  (−2.48)  (−4.11)(−3.42)
HERF  2.16*** 2.25***2.42***
   (5.18) (5.04)(4.94)
MSHARE   6.86***5.83***6.35***
    (6.13)(5.07)(5.68)
CAPEX1.87***1.76***1.89***2.00***1.94***2.08***
 (7.67)(7.31)(7.92)(8.14)(7.69)(8.76)
GROWTH0.99***0.94***1.01***0.95***1.02***1.27***
 (4.07)(3.98)(4.09)(3.87)(4.13)(5.76)
LNSALES1.15***1.18***1.12***0.93***1.03***0.78***
 (15.28)(14.77)(15.23)(11.78)(11.65)(4.98)
LNAGE−1.48***−1.53***−1.52***−1.59***−1.46***−1.12***
 (−4.99)(−5.16)(−5.17)(−5.27)(−5.00)(−4.90)
LEVERAGE3.13***3.28***3.15***3.15***3.27***2.96***
 (7.24)(7.50)(7.30)(7.23)(7.54)(6.90)
BTM−1.72***−1.95***−1.88***−2.07***−1.68***−1.61***
 (−5.05)(−5.37)(−5.41)(−6.61)(−5.73)(−6.02)
INDEX2.50***2.61***3.49***2.65***3.10***1.36**
 (2.67)(2.88)(3.64)(2.73)(3.23)(2.19)
PROFIT−3.84***−3.73***−3.59***−3.43***−3.77***−2.90***
 (−6.42)(−6.69)(−6.53)(−6.11)(−6.96)(−5.82)
OWN     3.32***
      (4.31)
OWN*OWN     −3.50***
      (−5.53)
N75,70875,70875,70875,70875,70846,271
pseudo R20.5660.5620.5670.5780.5970.485
CoC0.9730.9710.9720.9750.9770.965

In Model (2) we include barriers to entry, measured by the industry's capital intensity (CAPINT). In support of hypothesis 2a, we find that firms from industries with lower entry barriers are more likely to go public. In particular, operating in a capital intense industry decreases the sample average probability of an IPO by roughly 50%.9 For firms from non capital intense industries going public offers an opportunity to adopt more aggressive product market strategies in order to deter potential new entrants into their industry. This effect dominates possible concerns about the loss of confidential information to industry rivals. The positive sign of the coefficient of the Herfindahl index (HERF) in Model (3) shows that firms from more concentrated industries are more likely to go public. If greater concentration means less competition in an industry, this result implies that less product market competition positively affects the decision to go public, as predicted by hypothesis 3a. When product market competition is relatively weak, firms are less concerned about the loss of confidential information about their strategies and products. This result is also economically significant with a one standard deviation increase in the Herfindahl index increasing the sample average probability of going public by almost 41%.

Model (4) in Table 4 reveals that the firm's market share (MSHARE) positively influences the decision to go public. Firms that have a strong position in the product market seem to be less worried about the loss of confidential information to industry rivals. This supports hypothesis 4a. If the firm's market share increases by one standard deviation, the sample average probability of seeking a public listing increases with almost 22%. We include all product market variables in Model (5) and observe that all these variables remain statistically significant. Again, we find support for hypotheses 1a, 2a, 3a and 4a and thus fail to support our alternative hypotheses 1b, 2b, 3b and 4b.

To complete our discussion of the results, we turn to the control variables in our regression models. We find that firms are more likely to go public when they have larger investment projects to fund (CAPEX) and they experience faster one-year sales growth (GROWTH). This indicates that growing and heavily investing firms seek to finance their growth with public equity capital. Our results further show that firms that go public are larger (LNSALES) than private firms. A plausible explanation is that larger firms can more easily bear the fixed costs associated with an IPO. Surprisingly, we find that firms going public are younger (LNAGE) than firms that stay private. This is inconsistent with prior literature which suggests that younger firms are less likely to go public because of their higher information costs (Chemmanur and Fulghieri, 1999).

Furthermore, firms going public have a higher debt ratio (LEVERAGE) than firms that remain private. Such firms may need to rebalance their capital structure or have exhausted their debt capacity. Firms also go public at a time when the equity valuation of their industry is relatively high (i.e., their industry book-to-market ratio (BTM) is relatively low) and their industry index return (INDEX) has increased. This suggests that firms attempt to take advantage of the relative high equity valuation of their industries (Pagano et al., 1998). We find that more profitable firms (PROFIT) have a lower propensity to go public. More profitable firms seem to be able to finance their investment projects with retained earnings, instead of seeking public equity capital. In Model (6) we also include the ownership of senior managers and other insiders (OWN) and its square (OWN*OWN). The ownership data for IPO firms are from the pre-IPO year. The ownership data for the sample of eligible firms is collected for the year 2002. Our motivation for this one-year cross section is that ownership structures are typically stable over time and the collection of ownership data for all eligible firms from our sample is very time consuming. As a result, we have 46,271 firm-year observations for which we have pre-IPO or 2002 ownership data. We find that ownership by senior management is positively related to the IPO decision. However, the relation is non-linear because the square of OWN is negatively related to the decision to go public. This suggests that insiders with a large ownership stake prefer to stay private and keep in full control of the firm.10 All other results remain as before with the exception of the coefficient of industry margin (INDMARGIN) which becomes insignificant.

(iii)Interaction Effects

Our current analysis omits potentially important interaction effects between industry and firm-level variables.11 This is particularly true for product market characteristics. We, therefore, add interactions between potentially correlated independent variables to our main model. The results are reported in Table 5. It is important to note that all our previous findings remain intact after inclusion of the interactions.

Table 5.  The Determinants of the Decision to Go Public: Interaction Effects
Variables1234567
  1. Notes:

  2. This table reports the effects of several interactions between product market and control variables on our main results. INDMARGIN denotes industry margin computed as the weighted average profit margin of firms within industry j using the firm's market share as weights. We measure industry margin in industry j as:

    • image
  3. where MARGINi is firm i's profit margin defined as earnings before interest and taxes (EBIT) divided by sales; and MSHAREi is firm i's market share. CAPINT is capital intensity measured as a dummy variable that equals 1 if both the ratio of property, plant and equipment over number of employees and the ratio of net sales over number of firms in an industry exceed the median values of these ratios across all industries.

  4. HERF is the Herfindahl index measured as the sum of the squared market shares in industry j. We measure the Herfindahl index as:

    • image
  5. where MSHAREi is firm i's market share. MSHARE refers to market share defined as the ratio of the firm's sales over total sales in the firm's industry. CAPEX is capital expenditures over property, plant and equipment. GROWTH denotes growth in net sales over the year. LNSALES equals the natural logarithm of total sales (in millions of pounds). LNAGE is the natural logarithm of the year of observation minus the year the company was founded. LEVERAGE is measured as book value of long term debt plus short term loans and overdrafts divided by book value of total assets. BTM is the median book-to-market value of equity of firms in the same industry which traded on the London stock exchange. INDEX refers to the buy-and-hold return of the FTSE UK All Share ICB industry Index during the past year. PROFIT is operating profit over sales. Industries are classified by the three-digit SIC code and firms are assigned to the industry in which they had the largest proportion of sales. All ratio variables are winsorized to be no greater than one in absolute value, except for BTM, INDEX and INDMARGIN. BTM and INDEX are winsorized at the first and 99th percentile and INDMARGIN excludes profit margins that are less (greater) than the fifth (95th) percentile. We use lagged values of all the variables except for LNAGE to align the firm and industry characteristics with the period of time the decision to go public was presumably made. The regression includes a constant and calendar year dummies (not reported) and standard errors are adjusted for clustering in firms. One, two and three asterisks indicate that the coefficients are significantly different from zero at the 10%, 5% and 1% level or better, respectively. The joint-test which indicates that all coefficients are zero is rejected by the likelihood ratio test at a level lower than 0.001 for all models. For goodness-of-fit we report the pseudo R2 and the Coefficient of Concordance (CoC): the percentage of randomly drawn pairs of one IPO firm and one eligible IPO firm for which it is true that the likelihood estimate of the model is higher for the IPO firm.

INDMARGIN7.33***7.88***7.18***7.19***7.17***7.21***7.32***
 (3.04)(3.44)(2.99)(2.68)(2.93)(3.00)(3.03)
CAPINT−1.01***−1.13***−1.01***−1.05***−1.02***0.44−0.85***
 (−4.11)(−4.49)(−4.13)(−4.33)(−4.22)(0.24)(−2.68)
HERF2.25***2.36***2.24***2.24***2.65***2.21***2.28***
 (5.05)(5.22)(5.03)(4.97)(6.05)(4.84)(5.10)
MSHARE5.84***5.13***5.87***5.74***9.95***5.93***5.81***
 (5.09)(4.95)(5.15)(5.02)(6.17)(5.20)(5.10)
CAPEX2.03***1.96***1.94***1.90***1.97***1.91***2.05***
 (5.69)(7.61)(7.76)(7.81)(8.00)(7.73)(8.12)
GROWTH1.12***0.89***1.15***1.08***1.00***1.02***1.02***
 (3.07)(4.06)(5.62)(4.40)(4.11)(4.11)(4.12)
LNSALES1.03***−1.40***1.03***1.04***1.01***1.07***1.02***
 (11.65)(−4.64)(12.13)(11.93)(11.58)(11.22)(11.67)
LNAGE−1.46***−10.40***−1.46***−1.45***−1.49***−1.47***−1.47***
 (−5.01)(−8.28)(−5.08)(−4.88)(−5.10)(−5.03)(−5.01)
LEVERAGE3.27***3.17***3.27***3.29***3.32***3.27***3.28***
 (7.56)(7.66)(7.58)(7.64)(7.56)(7.55)(7.57)
BTM−1.68***−1.70***−1.68***−1.67***−1.70***−1.69***−1.68***
 (−5.72)(−5.33)(−5.67)(−5.60)(−5.40)(−5.72)(−5.69)
INDEX3.10***3.13***2.99***3.07***3.24***3.14***3.09***
 (3.24)(3.23)(3.07)(3.08)(3.34)(3.29)(3.23)
PROFIT−3.78***−2.30***−4.40***−2.22−3.72***−3.77***−3.76***
 (−6.87)(−3.45)(−6.64)(−1.12)(−6.89)(−7.02)(−6.92)
CAPEX*0.23      
 GROWTH(0.37)      
LNSALES* 0.85***     
 LNAGE (7.95)     
GROWTH*  1.45    
 PROFIT  (1.58)    
PROFIT*   19.56   
 INDMARGIN   (0.97)   
MSHARE*    −14.77***  
 HERF    (−2.93)  
LNSALES*     −0.14 
 CAPINT     (−0.79) 
CAPEX*      −0.49
 CAPINT      (−0.74)
N75,70875,70875,70875,70875,70875,70875,708
pseudo R20.5970.6300.5980.5980.6000.5970.597
CoC0.9780.9780.9780.9780.9780.9780.977

Since we use a nonlinear estimation model, the coefficients of the interaction terms cannot be interpreted as in a linear model.12 There is some controversy about the correct approach to analyzing interaction effects in nonlinear models. We follow Greene (2010) who argues that interaction effects in nonlinear models are too complex to describe numerically by simple summary statistics.13 One cannot assess the statistical significance of the interaction effect with the standard significance tests on the coefficient on the interaction term. Instead, Greene suggests using graphical analysis in which the values of the interacted variables are plotted against the probabilities predicted by the model.

Figures 1 to 7 display the relationships between the interacted variables presented in the models 1 to 7 in Table 5. These figures display the predicted probability of an IPO for every single model in relation to the two independent variables of interest, where the other variables are held constant at their means. For interactions between two continuous variables we plot five lines where each line shows the relationship between, respectively, the 10th, 25th, 50th, 75th and 90th percentile value of the one variable and a range of values of the other variable. The choice of the x-axis variable as well as the range of values of both interacted variables is arbitrary.14 For interactions between one continuous and one dummy variable we plot two lines: one for each value of the dummy. Greene (2010) interprets the interaction effect as the change in the distance between the different sets of predicted probabilities, which implies that the more the slopes of the plotted lines will diverge, the higher the impact of the interaction effect on the probability of interest, the probability of an IPO in our case. For lines that run parallel we can say that there is no interaction effect.

Figure 1 shows that the interaction between CAPEX and GROWTH broadens the distance between predicted probabilities for low versus high growth firms as CAPEX increases. Hence, the positive effect of firm's capital expenditures on the likelihood of an IPO increases with its growth in sales. The effect of the interaction on the change in the predicted probability, however, is rather low (and occurs mainly in the right tail of the distribution). Hence, the economic significance of the interaction effect of firm's capital expenditures and growth on the probability of an IPO is low.

Figure 2 shows that the interaction effect between LNSALES and LNAGE is ambivalent. LNAGE first decreases the probability of an IPO to a certain point of LNSALES after which it increases the probability of an IPO by increasing LNSALES. The interaction effect of size and age on the probability of an IPO reverses at about LNSALES of 12.24 (i.e., sales of 206.9 millions of pounds). The existence of a complex interaction effect between size and age may appear from the change of sign for LNSALES in model 2 with respect to all other estimated logistic models in our paper.

Figure 3 shows that the positive effect of growth in sales on the likelihood of an IPO decreases the more profitable the firm. The interaction effect between GROWTH and PROFIT appears to be weak because the gap between the predicted probabilities for the five different levels of PROFIT hardly changes when GROWTH increases.

The effect of the interaction between the firm sales margin (PROFIT) and the industry sales margin (INDMARGIN) is depicted in Figure 4. It shows that the predicted probabilities for firms from relatively high profitable industries vis-à-vis relatively low profitable industries decreases when the profitability of the firm decreases. Stated differently, the negative association between profitability of the firm and the likelihood of an IPO is increasing with the profitability of the industry.

Figure 5 displays the interaction effect of the firm's market share (MSHARE) and industry concentration (HERF). Both variables positively influence the likelihood of a firm going public. The joint effect of these variables on the likelihood of an IPO is ambivalent. HERF first increases the effect of MSHARE on the likelihood of an IPO to a certain value of MSHARE after which HERF decreases the effect of MSHARE. The change in influence turns at a value of MSHARE at about 0.16. Hence, the negative effect of HERF on the propensity to go public when MSHARE increases only occurs for firms with very high market shares.

Figure 6 shows that the impact of the interaction between LNSALES and CAPINT is to widen the gap between predicted probabilities for non capital intense industries and capital intense industries as firm size increases. Also, Figure 7 shows that the impact of the interaction between CAPEX and CAPINT is to widen the predicted probability gap for non capital intense and capital intense industries when firm capital expenditures increase. Both Figure 6 and Figure 7 show that the positive effect of, respectively, firm's size and capital expenditures on the likelihood of an IPO decreases when a firm operates in a capital intense industry.

Figure 8 shows the effect of adding the squared value of OWN to the estimation model. Both OWN and OWN squared have significant coefficients suggesting that both variables are statistical significant determinants of the likelihood of an IPO. The signs of the coefficients of OWN and OWN squared suggest that the relationship between OWN and the dependent variable is inverted U-shaped, as we contend in Section 4(ii). Figure 8 supports this presumption. It shows that ownership by senior management increases the likelihood of an IPO for low values and decreases it for high values of ownership. The turning point occurs around an ownership value of 53%.

(iv)Robustness Checks

We conduct several robustness checks. We first test how sensitive our results are to the sample criteria we used to identify eligible IPO firms. We based the selection of potential IPOs on the availability of audited financial statements, i.e., companies with annual sales in excess of £350,000 before 2000 and £1,000,000 thereafter. To test the robustness of our results, we re-estimate all models with the requirement that (i) annual sales is at least £1 million, (ii) the company exists for at least five years, (iii) a combination of requirements (i) and (ii), (iv) annual sales is at least £350,000 and (v) a combination of requirements (ii) and (iv). All our findings are similar except for profitability (PROFIT) which becomes insignificant when we apply requirement (iii).15

The largest industry in our IPO sample is the industry Services-Computer Programming, Data Processing (SIC 737) with 61 IPO firms. To test whether our results are driven by this industry, we re-estimated all models excluding firms from this industry. We find that our results are not changed. As a final robustness check, we measure the product market characteristics at the 4-digit SIC code level, instead of the 3-digit SIC code level. Again, we find similar results.

(v)Product Market Related Effects in the Post-IPO Period

In this section we further investigate the product market related effects of going public. Table 6 shows dynamic IPO firm characteristics for one year before up to three years after the IPO. The event year of the IPO is labeled year 0. The first (second) row shows the median (mean) values. We compare the median (mean) ex-post IPO characteristics to the year before the IPO using a two sample Wilcoxon rank-sum (Mann-Whitney) test for differences in medians and a t-test for differences in means. In our discussion we focus on the differences in medians because these are less influenced by extreme observations.

Table 6.  Dynamic IPO Firm Characteristics (Ratios)
VariableEvent YearsTest of Difference (p-values)
Year −1Year 0Year 1Year 2Year 3−1 to 0−1 to +1−1 to +2−1 to +3
  1. Notes:

  2. This table reports dynamic firm characteristics of firms that went public on the London Stock Exchange between 1994 and 2006. Table values are for the mean and median level. The first (second) row per variable shows median (mean) values, and the third row depicts (within parentheses) the number of observations. The last four columns report p-values using a two sample Wilcoxon rank-sum (Mann-Whitney) test for differences in medians and a t-test for differences in means. MSHARE refers to market share defined as the ratio of the firm's sales over total sales in the firm's industry. CAPEX is capital expenditures over property, plant and equipment. PPE is property, plant and equipment divided by total assets. GROWTH denotes growth in net sales over the year. PROFIT is operating profit over sales. VOLATILITY is the standard deviation of PROFIT. VOLATILITY before the IPO is measured for the three years period preceding the year of the IPO. VOLATILITY after the IPO is measured for the period starting in the year of the IPO (year 0) and ending three years thereafter. All observations are firm year observations except for VOLATILITY. All ratio variables are winsorized to be no greater than one in absolute value. VOLATILITY is winsorized at the 5th and 95th percentile.

MSHARE0.0150.0190.0210.0210.0180.1270.066*0.058*0.161
0.0890.0980.0930.0810.0810.5760.8000.6050.667
(337)(335)(304)(261)(232)(335)(304)(261)(232)
CAPEX0.3400.3800.3310.2540.2230.5100.7340.001***0.001***
0.3940.4190.3780.2940.2970.3670.5790.001***0.001***
(337)(335)(312)(279)(254)(335)(312)(279)(254)
PPE0.3600.3230.3650.4030.4210.098*0.6110.2770.093*
0.4030.3720.4130.4260.4360.1370.6280.2990.143
(336)(335)(313)(281)(254)(335)(313)(281)(254)
GROWTH0.2460.2660.2350.1870.1620.9440.5140.001***0.000***
0.3360.3270.3190.2230.2170.7620.5680.001***0.001***
(337)(333)(314)(280)(252)(333)(314)(280)(252)
PROFIT0.0890.0920.0770.0700.0610.5860.3890.2680.009***
−0.024−0.032−0.039−0.025−0.0220.7910.5990.9850.933
(337)(335)(312)(280)(254)(335)(312)(280)(254)
VOLATILITYBeforeAfterTest of Difference (p-values)      
 Mean0.0670.0880.001***      
 Median0.0230.0280.001***      
    (257)    (257)       

We find that the median market share (MSHARE) increases directly after an IPO. Market share in post-IPO years 1 and 2 is significantly higher than in the year before the IPO. However, the median market share in year 3 is not significantly higher than in the year before the IPO. This result suggests that in the years directly following an IPO the effect of having access to public equity at least initially dominates the potential loss of confidential information (which would predict a decline in market share). This result supports the models of Chemmanur and He (2011) and Chod and Lyandres (2010).

Table 6 shows no evidence that capital expenditures (CAPEX) increase immediately after the IPO. In contrast, there is a significant reduction in capital expenditures in the post-IPO years 2 and 3 relative to the year before the IPO. The ratio of fixed assets to total assets (PPE) decreases in the year of going public (year 0) relative to the year before. However, in post-IPO year 3 this ratio is significantly higher than in the year before going public.

We find that sales continue to increase during all post-IPO years (median sales growth is always positive) but that sales growth is declining over time. In post-IPO years 2 and 3 we find that sales growth (GROWTH) is significantly lower than in year −1. This finding is consistent with the US findings of Chemmanur et al. (2010) who attribute it to firms timing their IPO when a positive productivity shock occurs. The strong growth in sales in the year before the IPO reflects a positive shock in productivity. The financing of the increase in sales after the productivity shock, increases the likelihood of going public. This argument is also in line with the models of Clementi (2002) and Spiegel and Tookes (2008). Clementi argues that firms primarily go public to raise external financing to increase the scale of their operations. Spiegel and Tookes (2008) predict that firms before going public first (privately) finance the projects that will likely generate most revenues, such that only modest growth remains after the IPO. We also find that the ratio operating profits divided by sales (PROFIT) declines over time with the median profit margin in post-IPO year 3 being significantly lower than in the year before the IPO. This result is in line with Coakley et al. (2007) who report post-IPO operating underperformance for UK-IPOs in the period 1985–2003.

As a final point, we compare the standard deviation of PROFIT before and after the IPO (VOLATILITY). We find that PROFIT becomes more volatile after the IPO. (Since we need at least two years to be able to compute VOLATILITY, both before and after the IPO, we are left with a lower number of observations.) The increase in the volatility of profits is consistent with firms pursuing more aggressive product market strategies after their IPO resulting in an increase in operational risk (Chemmanur and He, 2011; and Chod and Lyandres, 2010).

One important reason for firms to go public is to have access to external equity markets and to increase the size of their operations (Clementi, 2002). We follow Chemmanur et al. (2010) and also examine the absolute values in real terms (using the 2009 CPI index from the Office of National Statistics) of capital expenditures, fixed assets, total assets, sales and operating profits. Table 7 shows the results. Again, we focus on median values in our discussion of the results because medians are less influenced by extreme observations.

Table 7.  Dynamic IPO Firm Characteristics (Values)
VariableEvent YearsTest of Difference (p-values)
Year −1Year 0Year 1Year 2Year 3−1 to 0−1 to +1−1 to +2−1 to +3
  1. Notes:

  2. This table reports dynamic firm characteristics of firms that went public on the London Stock Exchange between 1994 and 2006. The first (second) row per variable shows median (mean) values, and the third row depicts (within parentheses) the number of observations. The last four columns report p-values using a two sample Wilcoxon rank-sum (Mann-Whitney) test for differences in medians and a t-test for differences in means. All variables are values in thousands of pounds (in 2009 real terms using the CPI index from the Office of National Statistics). CAPEX is capital expenditures, PPE is property, plant and equipment, ASSETS refers to total assets, SALES equals sales and PROFIT is operating profit.

CAPEX£1,364£4,135£4,914£4,867£5,4620.000***0.000***0.000***0.000***
 (£000)£16,290£22,899£32,606£30,203£59,4860.2050.035**0.057*0.017**
 (337)(335)(312)(279)(254)(335)(312)(279)(254)
PPE£7,861£12,029£16,313£22,300£25,4640.002***0.000***0.000***0.000***
 (£000)£104,024£114,979£147,150£174,510£277,8660.7270.2480.095*0.027**
 (336)(335)(313)(281)(254)(335)(313)(281)(254)
ASSETS£22,302£42,195£51,451£59,623£65,8500.000***0.000***0.000***0.000***
 (£000)£166,102£200,850£222,206£260,419£398,2190.4180.2320.074*0.018**
 (337)(335)(314)(281)(254)(335)(314)(281)(254)
SALES£30,966£41,075£50,422£63,992£74,2250.034**0.000***0.000***0.000***
 (£000)£169,011£184,091£209,227£256,893£342,8970.7040.3650.1070.027**
 (337)(333)(314)(280)(252)(333)(314)(280)(252)
PROFIT£3,185£4,725£4,723£6,135£5,4180.081*0.1410.026**0.098*
(£000)£12,981£13,674£14,065£20,478£33,1460.8510.7720.1500.061*
 (337)(335)(312)(280)(254)(335)(312)(280)(254)

We find that firms spend substantially more on capital expenditures in the year of the IPO and directly thereafter than in the year before the IPO. The median amount spent on capital expenditures increases significantly from £1.3 million before going public to £5.4 million in post-IPO year 3. IPO firms also substantially increase their fixed assets, total assets, sales and profits in value terms as of the year of the IPO. This is consistent with the idea that firm go public to increase their scale of operations (Clementi, 2002; and Chemmanur et al., 2010).

Overall, our results suggest that firms go public at a peak in their sales growth and continue to expand their scale of operations directly following the IPO. After the IPO, firms show higher volatility in profitability which is consistent with firms being able to pursue more aggressive product market strategies. The increase in market share by the median IPO firm in post-IPO years 1 and 2 suggests that the effect of having access to lower-cost equity financing for increasing the scale of operations at least initially dominates the effect of having to disclose sensitive information to product market competitors (which would predict a smaller market share).

5. CONCLUSION

  1. Top of page
  2. Abstract
  3. 1. INTRODUCTION
  4. 2. LITERATURE AND HYPOTHESES
  5. 3. DATA AND METHODOLOGY
  6. 4. EMPIRICAL RESULTS
  7. 5. CONCLUSION
  8. REFERENCES

This paper investigates the influence of product market characteristics on the decision to go public. When making this decision, firms trade off the product market related costs and benefits of a public listing. Costs arise from the disclosure of confidential information which comes with the public status. This loss of confidential information may benefit product market rivals. At the same time, benefits come from the opportunity to grow and improve the firm's position in the product market using the newly obtained access to financing.

We examine the decision to go public using a sample of UK firms that went public and that stayed private during 1994–2006. We find that the likelihood of going public is positively related to the average profit margin in the industry. This result implies that firms from more profitable industries have more incentives to go public in order to grow and improve their position in the product market. They can do so by committing to riskier and more aggressive product market strategies that otherwise similar private firms or new entrants would not be able to adopt (Shah and Thakor, 1988). The increase in volatility in operating profits and increase in market share and scale of operations after the firm went public provides further support for the product market related benefits of going public. We also find that firms from industries with low barriers to entry are more likely to go public. This suggests that a public listing can yield an important competitive advantage which may deter new entrants.

Another key finding of this study is that firms from less competitive industries are more likely to obtain a public listing. This result implies that the loss of confidential information is less important for the going public decisions of firms that face less competition in their industry. We also find that a higher market share increases the likelihood of going public. Firms with a higher market share have a more secure competitive position and are therefore less worried about the loss of confidential information.

In conclusion, we find that going public involves a trade-off between competitive benefits and costs. Our results show that both confidential information theories (e.g., Campbell, 1979; Bhattacharya and Ritter, 1983; and Marra and Suijs, 2004) and more recent theories that focus on product market benefits of obtaining a public listing (e.g., Chemmanur and He, 2011; and Chod and Lyandres, 2010) are relevant for explaining the complex decision to go public.

Footnotes
  • 1

    FAME is an acronym for Financial Analysis Made Easy and is marketed by Bureau van Dijk Electronic Publishing.

  • 2

    We do not include firms that go public on the Alternative Investment Market (AIM) of the London Stock Exchange, because data for these small and medium sized private firms is difficult to obtain from the FAME database. The 1981 Companies Act permits small and medium sized private companies to file abridged financial statements (which do not need to include sales) at Companies House (Ball and Shivakumar, 2005). This means that we cannot obtain data for the small private firms that are eligible to list on the AIM.

  • 3

    The listing requirements of the Official List of the London Stock Exchange include, among others: (i) at least 25% of the share capital should be in the hands of the public so that shares can be actively traded, (ii) the anticipated market capitalization should be at least £700,000 and (iii) the company should have at least three years of accounts. However, the London Stock Exchange allowed companies (e.g., Freeserve plc in 1999) to go public on the Official List without a three-year trading record. These companies have to be ‘innovative high-growth companies’ which can demonstrate an ‘ability to attract funds from sophisticated investors’ and be raising more than £20 million. At the time of listing, they should have a market capitalization of more than £50 million. They must also detail their business plan and ‘risk factors'. The problem with using these listing criteria to identify potential IPO firms is that market capitalization is unobservable for private companies.

  • 4

    We considered using the four-digit SIC code but this would increase the difficulty of determining the main product market of the firm, especially for multi-product market firms (we do, however, check the robustness of our results by using a four-digit SIC demarcation). Using the two-digit or even one-digit SIC industry classification code would lead to too coarse product market boundaries.

  • 5

    The standard errors in our analysis are adjusted for 18,723 clusters. Since the sample contains 337 unique IPO firms, there were a total of 18,386 unique private firms that were in a position to go public in the period 1994–2006.

  • 6

    We thank an anonymous referee for this suggestion.

  • 7

    The results are available on request from the authors.

  • 8

    We calculate the marginal effect of a continuous variable as the percentage change in the estimated probability of an IPO by adding one standard deviation to the mean value of the variable of interest, while leaving all other variables constant at their mean values and all time dummies zero.

  • 9

    We calculate the marginal effect of CAPINT as the percentage change in the estimated probability of an IPO with and without this dummy variable, while leaving all other variables constant at their mean values and all time dummies zero.

  • 10

    The relationship between insider ownership and the probability of an IPO is further discussed in Section 4(iii) and depicted in Figure 8.

  • 11

    We thank an anonymous referee for this suggestion.

  • 12

    This is, of course, true for all coefficients in a logit model. In a linear model, the coefficients measure the average marginal effect of an independent variable on the dependent variable. The marginal effect of an independent variable in a logit model depends on the values of the other independent variables.

  • 13

    Huang and Shields (2000) also recommend using graphical displays for interpreting interactive effects in logit and probit analysis.

  • 14

    According to Greene (2010) one of the problems of analyzing interaction effects for continuous variables is accommodating the units of measurement.

  • 15

    All results are available upon request from the authors.

REFERENCES

  1. Top of page
  2. Abstract
  3. 1. INTRODUCTION
  4. 2. LITERATURE AND HYPOTHESES
  5. 3. DATA AND METHODOLOGY
  6. 4. EMPIRICAL RESULTS
  7. 5. CONCLUSION
  8. REFERENCES