IPO Valuation. Chapter Introduction. The International Evidence. Sanjai Bhagat, Jun Lu, and Srinivasan Rangan

Transcription

1 Chapter 5 IPO Valuation The International Evidence Sanjai Bhagat, Jun Lu, and Srinivasan Rangan 5.1. Introduction Initial public offerings (IPOs) are an economically significant contributor to the amount of capital raised by firms around the world. Over US$4.1 trillion was raised by over 36,000 firms that completed an IPO between 1998 and 2015 (Thomson SDC Global New Issues database). Further, IPO firms are also considered the drivers of innovation and employment in the economy, and hence their success is critical to economic growth. Beginning with Loughran, Ritter, and Rydqvist (1994), several studies have documented cross- country differences in the IPO process and how these differences influence the decision to go public and subsequent pricing (Henderson, Jegadeesh, Weisbach, 2006; Kim and Weisbach, 2008; Caglio, Hanley, Westburg, 2010). While a few studies have focused on underpricing (change in price on the first day of trading) of IPOs around the world, no previous study has examined cross- country differences in determinants of the level of the IPO value. 1 The valuation of IPOs occupies an important place in finance, perhaps because an IPO provides public capital market participants their first opportunity to value a set of corporate assets and growth opportunities. The valuation of IPOs is also quite relevant from an economic efficiency perspective: the IPO is the first opportunity that managers of such (usually young) companies get to observe price signals from the public capital markets. Such signals can either affirm or repudiate management s beliefs regarding the firm s future growth opportunities, which have obvious implications for real economic activity for example, employment and corporate investment. The finance literature has several hundred papers on IPO underpricing, which is the difference between the market price at the end of the offer day and the offer price. However, there is little in the literature on the offer price itself. It is the offer price that is of greater interest to investors, issuing companies, regulators, and investment bankers; investors are Part-1.indd Jul-18 10:11:09 PM

2 IPO Valuation: The International Evidence 109 more interested in IPO valuation, which is the offer price times the number of shares outstanding. In light of this, it is especially surprising that the finance literature has devoted so much intellectual effort to understanding IPO underpricing but so little to understanding the more primitive economic variable: the offer price, or the value of the IPO. In this chapter, we study the valuation of 6,199 IPOs during for the following countries: Australia, Canada, China, Germany, India, Japan, the United Kingdom, and the United States. We construct a model to value IPOs. Conceptually, the value of an asset is the sum of the present value of its expected future cash flows. Estimating the expected future cash flow of most any company is nontrivial; this task is even more challenging for IPO firms given that, in general, less is known about their past performance, and there is greater uncertainty about their future prospects. Instead of directly estimating future cash flows and the cost of capital to discount these cash flows, we construct proxies for the cash flows and discount rates; these proxies involve financial and nonfinancial variables. We discuss these proxies in the following. In a seminal article, Ohlson (1990) develops a model that expresses the market value of equity as a linear function of earnings, book value of equity, and other information. Subsequently, a sizable body of empirical research (see, for example, Penman, 1998, and Francis and Schipper, 1999) has used this model to motivate empirical investigations of the value relevance of earnings and book value of equity. Myers (1977) has suggested that the market value of a firm is positively related to its growth opportunities. Growth opportunities are especially critical for IPOs given that more, if not most, of an IPO s value is based on them. More recently, Abel and Eberly (2005) propose a valuation model that explicitly incorporates the possibility that firms may upgrade to or adopt a new technology. In their model, the value of the firm is composed of three components: the replacement cost of the firm s physical capital, the net present value of the firm s expected future cash flows from assets in place, and the value of growth options associated with future technological upgrades. There is a substantive literature in financial accounting on the value relevance of accounting variables in explaining cross- sectional variation in the market value of stocks; Kothari (2001) and Barth, Beaver, and Landsman (2001) review this literature. Collins, Maydew, and Weiss (1997) document the increasing importance of net income in explaining the cross- sectional variation in market value. Lev (2001) has argued that the value- relevance of book value has declined over time, since intangible assets comprise a larger fraction of the value of most firms today; this would be especially true for IPOs whose value, as noted earlier, would be drawn from their growth opportunities (intangible assets). Rhodes- Kropf, Robinson, and Viswanathan (2005) (RRV) draw on this literature and propose a simple model that includes only book value and net income to explain the cross- sectional variation in stock market values. We adopt and extend the RRV model in our analysis. The main variables that prior literature on accounting and finance has shown to influence IPO prices in the United States are earnings, book value, expected growth, pre- IPO insider ownership, underwriter reputation, and auditor reputation (Ohlson, 1990; Aggarwal, Bhagat, and Rangan, 2009). The accounting and finance literatures document that earnings opacity influences a range of capital market outcomes Part-1.indd Jul-18 10:11:09 PM

3 110 Sanjai Bhagat, Jun Lu, and Srinivasan Rangan (Bhattacharya, Daouk, and Welker, 2003; Gelos and Wei, 2005; Jin and Myers, 2006). We expect that the impact of pre- IPO data on IPO prices will be lower in countries where earnings are more opaque. Recently, several countries, especially in the European Union, have adopted International Financial Reporting Standards (IFRS). An open question is whether the adoption of these standards has influenced the quality of accounting data in these countries and if market participants perceive differences in quality. In a series of influential papers, La Porta et al. (1997, 2000) analyze the role of a country s legal system in protecting investor rights. They argue that diverse elements of countries financial systems as the breadth and depth of their capital markets, the pace of new security issues, corporate ownership structures, dividend policies, and the efficiency of investment allocation appear to be explained both conceptually and empirically by how well the laws in these countries protect outside investors. (2000, p. 4.) Further, they postulate that the commercial legal codes of most countries are based on four legal traditions: the English common law, the French civil law, the German civil law, and the Scandinavian law. They find that common law countries provide the most protection to investors and that they have the deepest stock markets and most dispersed corporate ownership structures. They also document that countries develop substitute mechanisms for poor investor protection, such as mandatory dividends and greater ownership concentration. Finally, they find that investor protection is positively correlated with valuation across countries. To the extent that IPO valuations are correlated with valuations of financial securities, in general, investor protection would be a determinant of IPO valuation. In a recent paper, Cumming et al. (2016) document a positive correlation between IPO valuations and backing of such IPOs by international venture capital syndicates. Additionally, listing requirements across exchanges and countries have been found to be correlated with IPO valuation; see Carpentier et al. (2012); Johan (2010); Carpentier et al. (2012); and Cumming et al. (2016). The remainder of the chapter is organized as follows. The next section (5.2) describes our sample and data, and section 5.3 discusses descriptive statistics of several pre- IPO financial and offering variables. Section 5.4 details our empirical results for the determinants of IPO valuation across countries. The final section (5.5) concludes with a summary Sample and Data Sample An initial sample of 21,577 observations for the years is obtained from the Thomson Reuters SDC Platinum New Issues database (SDC) for eight countries: Australia, Canada, China, Germany, India, Japan, the United Kingdom, and the Part-1.indd Jul-18 10:11:09 PM

4 IPO Valuation: The International Evidence 111 United States. In Table 5.1, we report the type and number of IPOs that are excluded to arrive at our final sample, by country and for all countries combined. Our filters are similar to those applied in prior research on IPO pricing we exclude financial firms, privatizations, unit offerings, private placements, spinoffs, rights offerings, offerings by limited partnerships, offerings that do not involve common shares (for example, loan Table 5.1 Sample Selection Screens for International IPOs, All Australia Canada China Germany India Japan United Kingdom United States Start: 21,527 2,205 4,061 4, ,193 1,890 4,588 ( ) Financial firms 5, , ,304 ( ) Privatizations ( ) Offerings of units (shares + warrants) ( ) Private placements ( ) Spinoffs 1, ( ) Rights offerings ( ) Limited partnerships ( ) Loan stock, options, or flow through shares ( ) Not underwritten ( ) Missing SEDOL and CUSIP ( ) Follow- on offerings ( ) Multiple tranches of same IPO ( ) Missing prospectus or IPOs not qualifying per screens , , ,534 1, , , ,321 4, Final Sample: 6, , ,069 Part-1.indd Jul-18 10:11:09 PM

5 112 Sanjai Bhagat, Jun Lu, and Srinivasan Rangan stock), offerings that are not underwritten, offerings by firms with missing CUSIPS and SEDOLs, and follow- on offerings. 2 These exclusions enable us to achieve a relatively homogenous sample of underwritten offerings of common shares to new investors that result in a cash infusion to the firm or its pre- IPO shareholders. The basic unit of observation on SDC is an offering tranche; while some IPOs have a single tranche, others have more than one tranche. To achieve a sample of unique IPOs (one observation per firm), we also eliminate 2,417 tranches that relate to multiple- tranche IPOs. For each of the remaining 10,534 IPOs (except for Chinese IPOs), we attempt to obtain final IPO prospectuses from the following sites: for US firms, www. sedar.com for Canadian firms, for Indian firms, and the Filings Library of the Thomson One Banker database for Australian, German, Japanese, and UK firms. For Chinese firms, we obtain pre- IPO data from the Guo Tai An (GTA) database, and not from prospectuses. When reading prospectuses, we discovered firms that would be eliminated based on the aforementioned filters. Our final filter- compliant sample for which we have prospectuses consists of 6,199 IPOs. China and the United States have the highest number of IPOs among the eight countries, and Canada and Germany the lowest. Figure 5.1 presents a pie chart of the country- wise shares of the number of completed IPOs. Table 5.2 presents the time series of offering frequencies in percentages for the eight countries. In general, IPO volumes display cyclicality, with peak volumes varying across countries. IPO frequencies rose during the years before dropping in 2001 the year following the crash of the dot- com bubble. Subsequently, IPO markets Australia 8% Canada 5% USA 33% China 27% UK 7% Japan 8% India 8% Germany 4% Australia Canada China Germany India Japan UK USA Figure 5.1. Distribution of 6,199 IPOs in countries during Part-1.indd Jul-18 10:11:09 PM

6 Table 5.2 Distribution of IPOs over Time ( ) Year Frequency Frequency % All All Australia Canada China Germany India Japan United Kingdom United States % 1% 6% 5% 6% 0% 2% 3% 9% % 5% 6% 3% 18% 0% 4% 1% 19% % 6% 9% 6% 32% 3% 9% 14% 15% % 1% 1% 3% 5% 0% 7% 6% 3% % 2% 3% 2% 1% 0% 4% 6% 2% % 5% 1% 3% 0% 1% 6% 4% 2% % 10% 6% 5% 1% 4% 6% 15% 7% % 10% 6% 1% 2% 10% 7% 13% 6% % 13% 6% 3% 11% 14% 9% 9% 6% % 18% 12% 6% 11% 16% 5% 7% 6% % 3% 6% 4% 1% 7% 2% 1% 1% % 4% 4% 6% 0% 3% 0% 0% 2% % 5% 8% 19% 2% 10% 3% 2% 3% % 3% 9% 15% 4% 6% 4% 1% 3% % 2% 6% 5% 3% 3% 6% 2% 3% % 3% 5% 0% 0% 4% 6% 4% 4% % 5% 2% 5% 1% 6% 9% 8% 5% % 5% 3% 9% 1% 12% 11% 4% 3% 6, , ,069 Part-1.indd Jul-18 10:11:09 PM

7 114 Sanjai Bhagat, Jun Lu, and Srinivasan Rangan around the world recovered to grow in volume from 2002 to With the advent of the financial crisis in the years , IPO volume fell for all countries. Thereafter, during the most recent period, , volumes returned to pre- crisis levels, but never exceeded levels attained between 2002 and The peak offering frequency year varies across countries and is as follows: Australia 2007, Canada 2007, China 2010, Germany 2000, India 2007, Japan 2015, United Kingdom 2004, and United States Figure 5.2 plots the time series of offering frequency for the eight countries combined Data Sources and Definitions SDC is our primary data source for the following 13 offering- related variables: offer price in local currency, offer date, offering proceeds in US dollars, primary shares offered, secondary shares offered, total shares offered, shares outstanding after the offering, shares outstanding before the offering, name of the auditor, names of the lead underwriters, exchange, currency in which shares are offered, and four- digit SIC code. When these variables are missing, we filled in their values from prospectuses. We verify the accuracy of these SDC- sourced variables (except for SIC codes, offering proceeds, and underwriter names) by comparing their values with those in prospectuses; whenever we find a difference, we replace SDC data with values from the prospectuses. As stated earlier, for Chinese firms, we rely on the GTA database rather than SDC for offering data. We obtain founding years from IPO prospectuses, except for China and the United States. Founding years for Chinese firms are from the GTA database. For US firms, we obtain founding years from Professor Jay Ritter s IPO database; 3 when founding years were not available on that database, we obtain them from prospectuses. We define firm age as the difference between the offering year based on the offer date and the founding year Figure 5.2. Aggregate number of IPOs for Australia, Canada, China, Germany, India, Japan, the United Kingdom, and the United States. Part-1.indd Jul-18 10:11:09 PM

8 IPO Valuation: The International Evidence 115 Besides age, we construct four measures that we expect to influence IPO valuation: auditor reputation, underwriter reputation, industry Price- Earnings (PE) ratio, and pre- IPO market returns. Consistent with a large body of auditing literature, we measure auditor reputation as a 1 0 dummy variable based on whether or not the firm s financials were audited by a Big- N (N largest auditors, where N varies from 4 to 8) auditor. To measure underwriter reputation, we employ the method of Megginson and Weiss (1991). For each underwriter j and for every year t, we define x jt as the three- year moving average (t- 3, t- 2, t- 1) of IPO proceeds. Then, for the set of underwriters I, for the year t, the Megginson- Weiss rank for underwriter j is: UWRANK jt log( x jt ) = Max j I log x ( jt ) This measure of underwriter reputation is market- share based and is a continuous variable on the interval [0,1]. When an IPO has multiple lead underwriters, we divide IPO proceeds by the number of underwriters. To measure industry PE, for each IPO firm, we obtain the market capitalizations at the end of the month before the offering date for all firms in that firm s country that had the same two- digit SIC code (industry peers). We chose two- digit SIC codes to minimize data loss because of industries having too few firms. For these industry peers, we obtain the income before extraordinary items for the most recent year relative to the month- end at which market capitalization is measured. We compute industry medians of PE, market capitalization divided by income before extraordinary items. Market capitalization and income before extraordinary items are from The Center for Research on Security Prices (CRSP) and North America COMPUSTAT for US firms, Global Compustat for firms from Australia, Canada, China, Germany, Japan, and United Kingdom, and from the CMIE Prowess database for Indian firms. To compute three- month market return before each firm s offering date, we sum daily market returns over 63 trading days ending on the day before the offering date. Daily market returns are measured as the value- weighted market- wide return for US stocks from CRSP and as percent changes in daily market index values from CMIE prowess for Indian firms and from Global Compustat for the other six countries (Appendix 5.1 details the specific market indices used for the seven countries). Turning to the financial statement variables, we manually collect the following eight financial statement variables for years 1 to 3 relative to the offering date from prospectuses, whenever available: sales, research and development (R&D) costs, income before extraordinary items, cash flows from operations, capital expenditures, long- term debt, book value of equity, and total assets. All numbers were coded in millions of the local currency. For the empirical work, we converted flow numbers such as sales into US dollars by using the average of the daily exchange rates over the fiscal year. Stock numbers such as total assets were converted into US dollars based on the exchange rate on the last day of the fiscal year. When an exchange rate was unavailable on the last date, we (5.1) Part-1.indd Jul-18 10:11:09 PM

9 116 Sanjai Bhagat, Jun Lu, and Srinivasan Rangan used the most recent exchange rate before the fiscal year end date. Daily exchange rates are from All financial data items obtained are audited historical numbers. 4 In general, all flow numbers were measured over 12 months; however, to minimize data loss, when a firm reported either 13 months or 11 months of data, we converted those numbers into 12- month values. If a firm reported flow numbers for more than 13 months or less than 11 months, we set those numbers to missing. Additionally, if a firm changed its fiscal year, data for the year of the fiscal year change was set to missing. For stock numbers, we collected the numbers reported on the fiscal year end date; if the only balance sheet numbers available were reported on a date other than the fiscal year end date, we set those numbers to missing. Detailed definitions of financial statement variables are provided in Appendix Descriptive Statistics Table 5.3 presents descriptive statistics for several offering variables. All variables, except for the auditor reputation dummy variable, are winsorized at the 1% and 99% levels. Because most of the variables are skewed, we discuss only median values. Our primary dependent variable is post- IPO market capitalization, measured as the product of offer price and shares outstanding on completion of the offering (MCAP). To achieve comparability, MCAP is expressed in US dollars using exchange rates on the offer date. 5 US IPOs record the highest median MCAP ($312.8 million), followed by China ($256.3 million), and then by Canada at a distant third ($50.4 million). Median offering proceeds are largest for the United States ($75.0 million), and smallest for Australia ($4.8 million). Median offer prices range from $770 for Japan to $0.32 for Australia. Prior literature has documented that insider retention (percentage of offering retained by insiders) is a significant determinant of IPO valuation. We compute insider retention by subtracting secondary shares sold during the IPO from shares outstanding before the IPO and dividing that difference by post- IPO shares outstanding. Table 5.3 indicates that median post- IPO insider retention varies from 61% in Australia to 80% in Japan. We also report data on the frequency of secondary offerings. Secondary offerings are very frequent in Japan, with 89% of Japanese IPOs having a secondary component. At the other extreme, only 3% of Chinese IPOs have a secondary component. A large body of evidence documents that IPOs are underpriced and that first- day prices are higher than offer prices, on average. Table 5.3 indicates that this is true of our sample as well, with two additional insights. First, the mean percentage change in price from offer to open is indeed positive for all eight countries, ranging from 1% for Japan to 182% for the United Kingdom. However, medians are negative for Australia ( 18%), Germany ( 6%) and Japan ( 8%). Second, the underpricing evidence is largely driven by the change from the offer to the first- day open. The magnitude of price changes from open to close on the first day are relatively small. Mean open- to- close percent change Part-1.indd Jul-18 10:11:09 PM

10 Table 5.3 Offering Descriptive Statistics Australia Canada China Germany India Japan United Kingdom United States Market Cap. at Offer Price (Millions of US$) Mean Median No. of obs , ,069 Offer Proceeds (Millions of US$) Mean Median No. of obs , ,069 Offer Price (US$) Mean Median No. of obs , ,069 Insider Retention Mean 61% 66% 73% 67% 69% 79% 63% 73% Median 61% 71% 75% 71% 72% 80% 66% 75% No. of obs , ,069 Secondary Offerings Frequency 12% 12% 3% 65% 17% 89% 34% 29% No. of obs , ,069 First- Day Close (US$) Mean Median No. of obs , ,987 (Continued) Part-1.indd Jul-18 10:11:09 PM

11 Table 5.3 (Contd.) Australia Canada China Germany India Japan United Kingdom United States First- Day Open (US$) Mean Median No. of obs , ,970 % Change from Offer to First- Day Open Mean 7% 30% 68% 15% 18% 1% 182% 31% Median 18% 6% 43% 6% 8% 8% 199% 11% No. of obs , ,970 % Change from First Day Open to Close Mean 1% 1% 4% 1% 3% 0% 4% 3% Median 2% 0% 2% 0% 0% 0% 1% 0% No. of obs , ,970 Age Mean Median No. of obs , ,069 Underwriter Reputation Mean Median No. of obs , ,069 Part-1.indd Jul-18 10:11:09 PM

12 Big- N Auditor Frequency (%) 32% 55% 3% 40% 15% 85% 42% 87% No. of obs , ,067 Industry PE Mean Median No. of obs , ,067 Pre- Offering 63- Day Return Mean 2.9% 2.8% 3.3% 2.5% 5.6% 3.4% 2.5% 3.8% Median 2.1% 1.9% 3.0% 2.8% 4.7% 2.6% 1.6% 4.5% No. of obs , ,069 Note: Variable Definitions are in Appendix 5.1. Part-1.indd Jul-18 10:11:09 PM

13 120 Sanjai Bhagat, Jun Lu, and Srinivasan Rangan ranges from 1% for Australia and Canada to 4% for China and the United Kingdom; medians range from 2% to 2%. Table 5.3 also reports descriptive statistics for our three nonfinancial value drivers: age, underwriter reputation, and auditor reputation. IPO firms in Australia, Canada, and the United Kingdom are relatively young, with a median age of one to three years. Japanese and Indian IPOs have the highest median ages of 11 and 19 years, respectively. Chinese, German, and US firms have a median age of seven years. Median underwriter reputation values are highest for Japan and the United States (0.89, 0.81). For the other countries, median reputation values range from 0.55 for Australia to 0.75 for China. The frequency with which IPO firms employ Big- N auditors varies considerably across countries, ranging from a mere 3% in China to 85% in Japan and 87% in the United States. Overall, it appears that US and Japanese firms prefer high reputation auditors and underwriters. Alternately stated, the audit markets in the United States and Japan are more oligopolistic than those of other countries. Finally, we report means and medians for pre- IPO industry PE ratios and three- month market returns. Median industry multiples, except for China, range from 10 to 25; these numbers are comparable with the median PE ratio of 24 for the S&P 500 over our sample period, ( table). Median Chinese industry PEs are relatively high at 44. Pre- IPO market returns are in general positive, with median returns ranging from 1.6% to 4.7%. Table 5.4 reports, via a series of panels, the mean and median values of several financial indicators over years 3 to 1. Again, all variables are winsorized at the 1% and 99% levels. Panel A reports that median sales increases monotonically for six of the eight countries; Australian and Canadian IPOs record average declining sales. The surprising decline for these two countries can be explained by the fact that their sample sizes increases from year 3 to 1 and the firms that enter the sample for the first time in year 1 tend to have low sales. 6 Panel B presents sales growth rates in year 2 and 1. Median sales growth rates in year 2 range from 17% for Japanese firms to 37% for US firms. In year 1, growth rates decline for China, Germany, India, and the United Kingdom, but rise for the remaining countries. Panels C and D in Table 5.4 provide information on pre- IPO trends in return on assets (ROA) and loss frequency. ROA is defined as income before extraordinary items divided by ending total assets. While the median Australian, Canadian, and US firm is in the red or fails to achieve profitability for all three pre- IPO years, the median firms in the other five countries generate positive ROA in all three years. Further, median ROA trends upward in India and Japan. In data in Panel D, suggest that pre- IPO loss frequencies in the three Asian countries China, India, and Japan, are relatively low, ranging from 0% to 6%. In contrast, close to two- thirds of the US and Canadian firms are unable to turn a profit in year 1. To gain further understanding of the profit differentials across countries, we present in Panels E and F, a decomposition of ROA into operating cash flows scaled by total assets and its complement, accruals scaled by total assets (accruals = income before extraordinary items operating cash flows). The poor profitability of US and Canadian Part-1.indd Jul-18 10:11:09 PM

14 Table 5.4 Descriptive Statistics of Pre- IPO Financial Variables Panel A: Sales (Millions of US$) Means Medians No. of obs. Year 3 Year 2 Year 1 Year 3 Year 2 Year 1 Year 3 Year 2 Year 1 Australia Canada China ,612 1,632 1,643 Germany India Japan United Kingdom United States ,537 1,899 2,012 Panel B: Sales Growth Means Medians No. of obs. Year - 3 Year 2 Year 1 Year 3 Year 2 Year 1 Year 3 Year 2 Year 1 Australia 52% 40% 20% 24% Canada 108% 86% 27% 41% China 35% 29% 24% 21% 1,612 1,631 Germany 85% 64% 32% 31% India 106% 105% 37% 31% Japan 45% 49% 17% 18% United Kingdom 90% 63% 22% 21% United States 128% 169% 37% 44% 1,389 1,730 Panel C: Return on Assets = Income before Extraordinary Items Divided by Total Assets Means Medians No. of obs. Year 3 Year 2 Year 1 Year 3 Year 2 Year 1 Year 3 Year 2 Year 1 Australia 10% 18% 31% 3% 4% 0% Canada 60% 36% 24% 10% 3% 7% China 14% 14% 14% 12% 13% 13% 1,599 1,619 1,642 Germany 11% 4% 0% 9% 3% 5% India 6% 7% 8% 4% 5% 7% Japan 6% 9% 10% 6% 7% 10% United Kingdom 22% 29% 24% 2% 2% 1% United States 47% 50% 44% 10% 15% 14% 1,324 1,865 2,008 (Continued) Part-1.indd Jul-18 10:11:10 PM

15 Table 5.4 (Contd.) Panel D: Loss Frequency Means No. of obs. Year 3 Year 2 Year 1 Year 3 Year 2 Year 1 Australia 17.0% 31.5% 44.9% Canada 59.2% 55.1% 65.7% China 0.2% 0.0% 0.0% 1,612 1,620 1,643 Germany 27.2% 29.0% 26.7% India 11.2% 7.6% 5.9% Japan 10.1% 8.7% 6.1% United Kingdom 44.2% 44.1% 45.6% United States 65.1% 66.3% 65.4% 1,529 1,893 2,010 Panel E: Operating Cash Flows/Assets Means Medians No. of obs. Year 3 Year 2 Year 1 Year 3 Year 2 Year 1 Year 3 Year 2 Year 1 Australia 10% 8% 6% 0% 0% 6% Canada 32% 23% 18% 6% 2% 5% China 11% 13% 12% 10% 11% 11% 1,174 1,235 1,409 Germany 10% 2% 2% 11% 8% 8% India 4% 3% 3% 4% 3% 5% Japan 6% 9% 9% 7% 8% 9% United Kingdom 8% 14% 10% 6% 7% 7% United States 30% 33% 24% 2% 5% 4% 1,270 1,847 1,999 Panel F: Accruals / Assets; Accruals = Income before Extraordinary Items minus Operating Cash Flows before Discontinued Operations Means Medians No. of obs. Year 3 Year 2 Year 1 Year 3 Year 2 Year 1 Year 3 Year 2 Year 1 Australia 1% 8% 18% 2% 5% 5% Canada 24% 13% 8% 7% 5% 5% China 3% 2% 2% 3% 2% 2% 1,174 1,229 1,409 Germany 1% 7% 2% 0% 3% 2% India 2% 3% 4% 1% 2% 2% Japan 3% 0% 2% 1% 0% 1% United Kingdom 14% 13% 14% 6% 7% 7% United States 18% 17% 18% 8% 8% 10% 1,270 1,845 1,997 Part-1.indd Jul-18 10:11:10 PM

16 Table 5.4 (Contd.) Panel G: Research & Development (R&D)/ Assets (When R&D is Non Zero) Means Medians No. of obs. Year 3 Year 2 Year 1 Year 3 Year 2 Year 1 Year 3 Year 2 Year 1 Australia 9% 8% 34% 9% 7% 22% Canada 41% 34% 22% 18% 20% 16% China Germany 27% 15% 15% 10% India 26% 18% 14% 2% 1% 1% Japan 10% 28% 1% 2% United Kingdom 30% 27% 30% 14% 14% 12% United States 49% 48% 41% 27% 29% 26% 806 1,171 1,248 Panel H: Capital Expenditures / Assets Means Medians No. of obs. Year 3 Year 2 Year 1 Year 3 Year 2 Year 1 Year 3 Year 2 Year 1 Australia 6% 7% 11% 2% 3% 4% Canada 6% 11% 21% 3% 5% 11% China 6% 7% 9% 5% 6% 8% 1,173 1,224 1,368 Germany 6% 9% 13% 3% 5% 8% India 6% 8% 12% 3% 5% 8% Japan 2% 4% 6% 2% 2% 3% United Kingdom 6% 7% 9% 3% 3% 4% United States 7% 8% 12% 3% 4% 7% 1,453 1,853 1,968 Panel I: Book Value of Equity (Millions of US$) Means Medians No. of obs. Year 3 Year 2 Year 1 Year 3 Year 2 Year 1 Year 3 Year 2 Year 1 Australia Canada China ,364 1,380 1,393 Germany India Japan United Kingdom United States ,416 1,936 2,059 (Continued) Part-1.indd Jul-18 10:11:10 PM

17 124 Sanjai Bhagat, Jun Lu, and Srinivasan Rangan Table 5.4 (Contd.) Panel J: Long- Term Debt by Total Assets (%) Means Medians No. of obs. Year - 3 Year - 2 Year - 1 Year - 3 Year - 2 Year - 1 Year - 3 Year - 2 Year - 1 Australia 11% 13% 8% 0% 2% 0% Canada 13% 13% 11% 0% 1% 0% China 7% 7% 7% 2% 3% 3% 1,551 1,575 1,581 Germany 11% 17% 17% 0% 9% 10% India 26% 26% 26% 23% 23% 24% Japan 23% 19% 17% 25% 14% 11% United Kingdom 24% 20% 20% 4% 5% 3% United States 20% 21% 23% 4% 6% 7% 849 1,816 1,981 firms holds up even with operating cash flows; median numbers are negative for all three years. The median IPO firm in the remaining six countries are generating positive operating cash flows in all three years; however, only Japanese firms report a strictly monotonic trend. Because accruals tend to be dominated by a large negative accrual depreciation we expect average pre- IPO accruals to be negative. Surprisingly, median accruals divided by assets are positive for all three years for China and India, ranging from 1% to 3%. The unusual nature of these positive accruals becomes more stark, when contrasted with the median accruals divided by assets for all US firms on COMPUSTAT for our sample period. This number is 6%. While a portion of the positive accruals can be attributed to high sales growth rates, the possibility that Chinese and Indian firms are managing their pre- IPO accruals upward remains. For the other six countries, median accruals scaled by assets are in general negative, as expected, and do not display any particular trend. We also examine the investing patterns of IPO firms before they conduct their offerings. Panel G and H report averages for R&D deflated by assets, when R&D is non- zero, and capital expenditures deflated by assets. US and Canadian firms are the most R&D intensive, with R&D rates ranging from 26% to 29% this partly explains the poor profitability of IPOs in these two countries. India and Japanese IPOs fall on the low end of the R&D investing spectrum, with R&D rates of only 1% 2% of assets in the year before the IPO. With respect to median capital expenditure spending rates, Panel H indicates that IPO firms in all eight countries display an increasing trend over years 3 to 1. Panel I and J report data on two balance sheet characteristics book value of equity (in millions of $) and long- term debt to assets ratio. Except for US firms that have a small negative median book value, IPO firms in general report a positive book Part-1.indd Jul-18 10:11:10 PM

18 IPO Valuation: The International Evidence 125 value of equity. In terms of leverage, Indian firms have the highest levels, 24% in the year before the offering. The median German and Japanese firm funds about 10% 11% of their assets with long- term debt. Pre- IPO leverage is relatively low for other countries. Figures 5.3 through 5.9 illustrate some of the key descriptive statistics Australia Canada China Germany India Japan UK US Median Market Capitalization at Offer Price (Millions of US$) Median Offer Proceeds (Millions of US$) Figure 5.3. Median IPO market capitalization, median IPO offer proceeds (US$ millions) Australia Canada China Germany India Japan UK US Figure 5.4. IPO company median age (years) Australia Canada China Germany India Japan UK US Figure 5.5. Median sales of IPO companies during the past 12 months (US$ millions). Part-1.indd Jul-18 10:11:10 PM

19 50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% Australia Canada China Germany India Japan UK US Figure 5.6. Median sales growth of IPO companies during the past 12 months % 60.00% 50.00% 40.00% 30.00% 20.00% 10.00% 0.00% Australia Canada China Germany India Japan UK US Figure 5.7. Percentage of IPOs with an earnings loss during the past 12 months. 30% 25% 20% 15% 10% 5% 0% Australia Canada China Germany India Japan UK US Figure 5.8. Median R&D/ assets during the past 12 months (when R&D is non- zero). Part-1.indd Jul-18 10:11:10 PM

20 IPO Valuation: The International Evidence % 10% 8% 6% 4% 2% 0% Australia Canada China Germany India Japan UK US Figure 5.9. Median capital expenditure/ assets during the past 12 months IPO Valuation Cross- Country Differences in Relation of Financial Variables to IPO Valuation Our research objective is to examine the relation between IPO market valuations and financial variables and to document the existence and magnitude of cross- country differences in the same. To do so, we estimate a simple and parsimonious valuation model estimated for US mergers by Rhodes- Kropf, Robinson, and Viswanathan (2005; RRV hereafter). The model, which is an adaptation of the seminal work of Ohlson (1990), is specified for each country j as follows: LOGMCAPj = α0j + α1jbvj + α2jlogabsni j + α3jlossdum * LOGABSNI j + α2 LEV j +α j (5.2) j where LOGMCAP is the log of market value of equity on the offer date, BV is the book value of common equity at the end of the year before the offering (year 1), LOGABSNI is the log of the absolute value of income before extraordinary items in year 1, LOSSDUM is a dummy variable that equals one when income before extraordinary items in year 1 is negative and 0 otherwise, and LEV is the ratio of long- term debt to total assets as the end of year 1. We employ log values for market capitalization and absolute value of income before extraordinary items considering the significant skewness in these variables. By interacting LOSSDUM and LOGABSNI, we allow for the valuation of income to differ based on its sign. We expect the coefficients of BV and ABSNI, α 1 and α 2, to be positive. Because loss firms tend to have income that is less persistent, we expect α 3 to be negative. RRV interpret leverage as a proxy for cost of capital and expect that higher leverage will be associated with lower market values. Alternately, leverage can also be viewed a source of monitoring that could benefit shareholders (Jensen, 1989). Under this interpretation, the sign of the coefficient on leverage would be positive. Part-1.indd Jul-18 10:11:10 PM

21 128 Sanjai Bhagat, Jun Lu, and Srinivasan Rangan We first estimate Eq. (5.1) for each country j, individually. Then, we estimate a global valuation model that combines data for all eight countries and allows us to test for differences in coefficients of value- drivers across countries. All variables, except dummy variables, are winsorized at the 1% and 99% levels. Standard errors are adjusted for heteroscedasticity and clustering across firms within each year. For the country- level models, financial variables are in local currencies; for the global valuation model, financial data are expressed in US dollars. Table 5.5, Panel A, presents the results for the RRV valuation model for IPOs in the eight countries. As expected, income before extraordinary items (LOGABSNI) is positively related to IPO valuation in each of the eight countries. The economic impact of net income is largest for Chinese IPOs and smallest for Australian IPOs. Specifically, a 1% increase in net income in China is associated with a 0.82% increase in IPO valuation. A 1% increase in net income in Australia is associated with a mere 0.07% increase in IPO valuation. In contrast to income, book value of equity that reflects assets in place is valued significantly in only four countries: Canada, Germany, India, and the United States. We do not observe a significant relation between book value and IPO valuation for Australia, China, Japan, and the United Kingdom. The interaction between the LOSS dummy and ABSNI measures the differential valuation of loss firms and profitable firms. For IPOs in the United States, the valuation effect of income when it is negative is lower than when it is positive; the magnitude of this differential is, however, economically small with a coefficient of 0.10 on the interaction. We find that LOSS times ABSNI and IPO value are negatively related for UK IPOs as well. However, Canadian and Japanese IPOs that have incurred a loss in earnings are associated with a higher valuation. We do not observe a statistically significant loss- versus- profit valuation differentials for Australia, Germany, and India. Because all Chinese firms report positive pre- IPO profits, the loss- profit interaction is not relevant for them. The results in Panel A of Table 5.5 indicate that leverage s effect on market capitalization varies across countries. The effect is positive and statistically significant in Canada, India, and the United Kingdom, consistent with the idea that debt provides monitoring benefits to shareholders. However, it is negative and significant in Japan, suggesting that Japanese investors interpret it as a proxy for cost of capital. Leverage is not significantly related to market values in Australia, China, Germany, and the United States. Table 5.5, Panel B, highlights the results for the RRV valuation model augmented with investment, growth, and nonfinancial variables for the IPOs in the eight different countries. The magnitudes and significance of the coefficients of the RRV model for China and Germany are similar to those reported in Panel A. While ABSNI loses statistical significance in Canada and Japan, BV becomes statistically and significantly related to market values in Australia and Japan. Further, BV is no longer significantly related to value for Canadian firms. In terms of the differential valuation of loss firms, the results change for UK firms alone; LOSS*ABSNI remains negatively related to market values, but is no longer statistically significant. Finally, leverage is no longer significantly related to market value for Canadian and Indian firms; interestingly, leverage is valued positively by US investors under the augmented RRV model. Part-1.indd Jul-18 10:11:10 PM

22 Table 5.5 Country- Level Valuation Regressions Panel A: Rhodes- Kropf, Robinson, and Vishwanathan (2005) (RRV) Model Australia Canada China Germany Coef. t- stat SL Coef. t- stat SL Coef. t- stat SL Coef. t- stat SL Intercept * * * * BV * * LOGABSNI * * * * LOSS*LOGABSNI *** DTOA * Adjusted R % 50.9% 67.3% 29.6% No. of obs , India Japan United Kingdom United States Coef. t- stat SL Coef. t- stat SL Coef. t- stat SL Coef. t- stat SL Intercept * * * * BV * * LOGABSNI * * * * LOSS*LOGABSNI * ** * DTOA * *** * Adjusted R % 14.0% 50.2% 31.2% No. of obs ,925 Notes: The dependent variable is the log of market capitalization at the offer date. SL is significance level. * indicates significance at the 1% level, ** at the 5% level, and *** at the 10% level. Standard errors account for clustering across firms within a year. Variable definitions are in Appendix 5.1. (Continued) Part-1.indd Jul-18 10:11:10 PM

23 Table.5.5 (Contd.) Panel B: RRV Model, Augmented with Investment, Growth, Nonfinancial Variables Australia Canada China Germany Coef. t- stat SL Coef. t- stat SL Coef. t- stat SL Coef. t- stat SL Intercept * * * BV * * LOGABSNI * * * LOSS*LOGABSNI *** DTOA LOG_ CAPEX ** * * LOG_ RD *** LOG_ SGRO LOG_ INDPE ** RETENTION * * UNDREP ** * * BIGN_ AUD *** *** AGE *** Adjusted R % 45.7% 73.1% 65.6% No. of obs , Part-1.indd Jul-18 10:11:10 PM

24 India Japan United Kingdom United States Coef. t- stat SL Coef. t- stat SL Coef. t- stat SL Coef. t- stat SL Intercept * * ** * BV * ** * LOGABSNI * * * LOSS*LOGABSNI ** * DTOA ** ** * LOG_ CAPEX * * * LOG_ RD LOG_ SGRO *** LOG_ INDPE ** * RETENTION *** ** * * UNDREP * *** * * BIGN_ AUD *** * ** AGE * Adjusted R % 29.6% 68.4% 63.9% No. of obs ,600 Notes: The dependent variable is the log of market capitalization at the offer date. SL is significance level. * indicates significance at the 1% level, ** at the 5% level, and *** at the 10% level. Standard errors account for heteroscedasticity and clustering across firms within a year. Variable definitions are in Appendix 5.1. Part-1.indd Jul-18 10:11:10 PM

25 132 Sanjai Bhagat, Jun Lu, and Srinivasan Rangan IPOs are usually associated with growth companies. Most of the value of a growth company is in its growth opportunities, and less in assets in place. We consider four proxies for growth opportunities: capital expenditures (CAPEX), R&D expenditures (RD), sales growth rate (SGRO), and industry price to earnings ratio (INDPE). Panel B, Table 5.5 indicates that capital expenditures are significantly and positively related to IPO valuation in Canada, Germany, India, the United Kingdom, and the United States. A 1% increase in capital expenditure is associated with a 0.25% increase in valuation in Canada, but only 0.11% increase in valuation in India (estimates for Germany, UK, and US are between the preceding estimates). There is no significant relation between capital expenditure and IPO valuation in China and Japan; the relation in Australia is positive but only marginally significant. In general, we do not find a statistically significant relation between IPO valuation and either R&D expenditures or sales growth. The exceptions are Germany for R&D and the United States for sales growth. 7 We find a positive and significant relation between industry price- to- earnings ratios and IPO values for Japan, a marginally significant and positive relation for Germany and India, and no significant relationship for Australia, Canada, China, the United Kingdom, and the United States. Overall, capital expenditures appear to be the most important driver of value from our four indicators of growth opportunities Cross- Country Differences in Relation of Nonfinancial Variables to IPO Valuation In addition to being high- growth firms, IPO firms are perceived as facing considerable uncertainty about future cash flows. Given the importance of future cash flows to IPO valuation, knowledgeable insiders may choose to communicate their information about future cash flows via costly and credible signals. We consider three such credible signals: ownership in the IPO firm retained by key insiders during the IPO (RETENTION), choice of underwriter (UNDREP), and choice of public accounting auditor (BIGN_ AUD). Further, because younger firms are likely to have more uncertain cash flows, we include the age of the firm as a fourth measure of uncertainty. We find a positive and statistically significant relation between insider retention and IPO valuation in China, Germany, the United Kingdom, and the United States; positive but marginally significant relationship for India and Japan; and no significant relationship for Australia and Canada. Underwriter reputation has a positive and statistically significant relationship for IPOs in China, Germany, India, the United Kingdom, and the United States; marginally positive relationship for Canada and Japan, and no significant relationship for Australia. Auditor reputation is positively and statistically significantly related to market values in the United Kingdom; marginally positive for Canada, Germany, Japan, and the United States; and is not significant related for Australia, China, and India. Age influences valuation only in India and China, and that too in opposite ways. In India, as expected, older firms garner higher valuations; but in China younger firms are valued relatively higher. Part-1.indd Jul-18 10:11:11 PM

26 IPO Valuation: The International Evidence 133 In Table 5.6, we report the results of a global valuation model that combines data for all eight countries. We include all variables from the augmented RRV model except R&D and INDPE; we exclude these two variables to minimize sample size reduction because of missing data. To formally test for differences across countries in mean valuations as well as in the coefficients of the augmented RRV model, we include country dummies for seven countries and the interactions between these dummy variables and the variables in the augmented RRV model. In terms of estimation method, we report results based on ordinary least squares (OLS) as well as quantile regression. The latter is robust to the effect of outliers in the data; therefore, in our discussion, we emphasize results that are consistent across the two estimation methods. Because we choose the United States as the base country, the coefficients on the main independent variables measure the effect of these variables on market values for US firms. The interaction term coefficients measure the country- level differences in valuation effects relative to US firms. Our first key finding is that, for US firms, the effect of most of the independent variables are significant and in the expected direction. Absolute value of net income and book value of equity are positively related to market capitalization and loss firms net income are valued less. Leverage is valued positively consistent with a monitoring effect. Among the growth variables, capital expenditures alone are valued positively and significantly under both estimation methods. Market valuations are increasing in our three nonfinancial signals insider retention, underwriter reputation, and auditor reputation. Our second finding relates to the country- level dummies. Among the seven countries, the dummy on China alone is negative and significant at the 1% level after controlling for firm characteristics, Chinese firms alone are valued less than US firms. The cause for this difference is an interesting subject for future research. Our third set of findings relate to the relative differences in the effect of various value- drivers between the United States and the other seven countries. We summarize these findings: (1) The absolute value of net income of IPO firms in Australia, China, India, and the United Kingdom is valued more than that of US firms. Negative values of income are valued more for Canadian IPOs. (2) The book value of equity of German and Japanese firms are valued more, whereas the book value of Chinese firms is valued less, relative to that of US firms. The cross- country differences in the valuation could be attributed to the differences in financial reporting standards. For example, the accounting and finance literatures document that earnings opacity and financial reporting standards influence a range of capital market outcomes (Bhattacharya, Daouk, and Welker, 2003; Gelos and Wei, 2005; Jin and Myers, 2006). Difference in investor sophistication across countries could also drive these differences. For example, while US IPOs are largely bought by institutional investors, in India, retail investors have significant participation. More research that disentangles these alternate explanations is needed. Part-1.indd Jul-18 10:11:11 PM