Causes and Consequences of Corporate Asset Exchanges by Listed Companies in China

Transcription

1 Causes and Consequences of Corporate Asset Exchanges by Listed Companies in China Fang Lou School of Economics Shanghai University of Finance and Economics Tel: Jiwei Wang* School of Accountancy Singapore Management University Tel: Hongqi Yuan School of Management Fudan University Tel: February 2014 * Corresponding author

2 Causes and Consequences of Corporate Asset Exchanges by Listed Companies in China ABSTRACT China s listed companies often exchange corporate assets with their parent companies. We find that listed companies that have been incompletely restructured from former state-owned enterprises and in sound financial condition tend to exchange higher quality assets for lower quality assets (i.e., tunneling). However, when there is a need to avoid reporting a loss and to raise additional capital, listed companies tend to exchange lower quality assets for higher quality assets (i.e., propping). We also find that the market reacts indifferently to asset exchange announcements. Finally, we find asset exchanges motivated by a tunneling (propping) incentive to be associated with poorer (improved) post-exchange stock performance and financial performance. In summary, this study contributes to the corporate asset literature by providing two new incentives: tunneling and propping. JEL classification: G14, G15, G34 Keywords: Asset exchange; Tunneling; Propping.

3 Causes and Consequences of Corporate Asset Exchanges by Listed Companies in China 1. Introduction The literature on corporate assets focuses on transactions involving payment in the form of cash, equity, and/or future considerations (Slovin, Sushka, and Polonchek, 2005) rather than barter-type asset exchanges. 1 In China, however, many listed companies barter by exchanging corporate assets with such related parties as parent companies and sister companies under common control. This paper addresses the reasons for and consequences of the exchange of corporate assets by listed companies in China. We identify two possible nonexclusive incentives for asset exchanges. The first is the incentive for related parties to reclaim higher quality assets (i.e., assets characterized by better investment opportunities and greater profitability) and inject lower quality assets, thereby resulting in the expropriation of minority shareholders. We label this incentive the tunneling incentive in the spirit of Johnson et al. (2000). The second is the incentive for related parties to exchange higher quality assets for lower quality assets to help the listed firm to boost operating performance. We label this the propping incentive in the spirit of Friedman et al. (2003). We argue that both the tunneling and propping incentives exist in China s particular institutional setting. Most of China s listed companies have been restructured from state-owned enterprises (SOE), typically through one of the three following restructuring processes. First, an existing SOE may peel off part of its operating assets to form a listed company, and it remains the parent of that listed company. Second, an SOE may be fully integrated with a listed company, and a government agency or equivalent 1 Although we could not obtain any statistics on barter-type asset exchanges in the United States, we conjecture that this type of asset transaction is quite rare in that country. 1

4 acts as the latter s parent. Finally, several SOEs may be bundled to form a listed company, and a government agency or equivalent acts as its parent. We classify the first type as incomplete restructuring because only part of the existing SOE has been transformed into a listed company. We treat the other two types as complete restructuring because the existing SOE(s) has been integrated into a listed company in its entirety. In an incomplete restructuring process, to help the to-be-listed firm to go public, the unlisted parent company tends to carve out higher quality assets to boost pre-ipo performance (Aharony et al., 2010). The remaining lower quality assets and other non-operating assets such as schools and hospitals become a financial burden, and hence the parent has a strong incentive to exchange lower quality assets for higher quality assets from the listed company in the post-ipo period. In addition, Peng et al. (2011) find that when listed companies in China are financially healthy, their controlling shareholders are more likely to conduct connected transactions to expropriate the minority shareholders of the listed companies. Following the description of tunneling put forward by Johnson et al. (2000), i.e., the transfer of assets and profits out of firms for the benefit of those who control them, we label a parent company s incentive to exchange lower quality assets for higher quality assets from a listed company under its control the tunneling incentive. Thus, we expect listed firms that have undergone an incomplete restructuring process and in sound financial condition to exchange higher quality assets for lower quality assets from their related parties. Unlike the securities regulators in more developed markets, those in China have established two bright-line earnings targets to regulate firm listings: a firm must report at least 0% return on equity (ROE) to maintain its listing status and 10% (6% after 2001) ROE to issue new shares. Although these bright-line rules have some 2

5 benefits, such as reducing adverse selection problems (Chen and Wang, 2007), the ROE targets provide listed firms managers with opportunistic earnings manipulation incentives (Chen and Yuan, 2004) and parent companies with incentives to assist listed firms in boosting their ROE. We thus expect that listed companies that need to avoid reporting a loss and to raise additional capital through equity offerings are likely to exchange lower quality assets for higher quality assets with their related parties. Following the description of propping put forward by Friedman et al. (2003), i.e., transferring private resources into firms that have minority shareholders, we label a parent company s incentive to exchange higher quality assets for lower quality assets from a listed company under its control the propping incentive. Accordingly, we expect listed firms with the intention to avoid reporting a loss and to raise additional capital to exchange lower quality assets for higher quality assets from related parties. Owing to the limited information available on exchanged assets, we were unable to measure asset quality directly. In an arm s length exchange, the valuation of the assets exchanged should be equivalent. If a manager wants to exchange lower quality assets for higher quality assets without extra compensation, he or she is likely to opportunistically manipulate the valuation of the lower quality assets to match that of the higher quality assets, thus indicating that the former should have a higher abnormal valuation. Accordingly, we should be able to infer asset quality through comparison of the abnormal valuation rates of the exchanged assets. When the abnormal valuation of the assets surrendered by a listed company is higher than that of the assets it acquires, we infer that the quality of the surrendered assets is lower than that of the acquired assets and vice versa. 3

6 We examine a sample of 305 asset exchanges made by 229 listed companies on the Shanghai and Shenzhen Stock Exchanges from 2000 to We present a model for determining asset valuation, and measure the abnormal asset valuation rate by the model s residual (the model is discussed in Section 5). The difference between the abnormal surrendered asset valuation rate and abnormal acquired asset valuation rate is then used as a proxy for the quality of the assets exchanged. If the abnormal valuation of the surrendered assets is higher than that of the acquired assets, we assume the quality of the surrendered assets to be lower than that of the acquired assets. Thus, a greater difference in the abnormal asset valuation rate indicates that there is a greater likelihood that the firm has exchanged lower quality assets for higher quality assets (i.e., the propping incentive) and vice versa (i.e., the tunneling incentive). We then provide empirical evidence to show that firms characterized by incomplete restructuring during the IPO process and in sound financial condition are associated with a lower abnormal asset valuation difference. The evidence indicates that this type of firms exchanges higher quality assets for lower quality assets, which is consistent with the tunneling incentive. Firms that intend to avoid reporting a loss and to raise additional capital through equity offerings, in contrast, are associated with a higher abnormal asset valuation difference. The evidence suggests that these firms are more likely to exchange lower quality assets for higher quality assets, which is consistent with the propping incentive. These results hold even after we control for other firm-level factors, such as return on assets (ROA), firm size, past stock return, market-to-book ratio of equity, cash holdings, leverage ratio, and growth rates in sales and gross property, plant, and equipment (PPE). Tunneling and propping behavior should have different effects on investors, and we would thus expect investors to 4

7 exhibit different reactions to asset exchange announcements depending on their motivating factor. However, we find the market to react indifferently to such announcements, which may cast doubt on the Chinese capital markets reputation for semi-strong efficiency. We also examine the consequences of barter-type asset exchanges. In the presence of the tunneling incentive, managers in listed companies exchange higher quality assets for lower quality assets. Hence, we expect long-term financial and stock underperformance in these companies. The propping incentive, in contrast, results in managers exchanging lower quality assets for higher quality assets. Hence, we expect long-term improvement in both financial performance and stock performance for these firms. Accordingly, we use stock performance (12- and 24-month post-exchange buy-and-hold abnormal return [BHAR]) and financial performance (1- and 2-year average post-exchange ROA) to test the consequences on post-exchange performance. As predicted, we find a positive association between both performance measures and a difference in abnormal asset valuation, thus indicating that asset exchanges motivated by the propping incentive result in improved post-exchange firm performance and those motivated by the tunneling incentive in poorer post-exchange performance. We also control for other factors that may affect firm performance, such as current ROA, firm size and leverage ratio, and find these results to hold. This paper contributes to the corporate asset literature in a number of ways. First, it identifies a sample of firms that engage in barter-type asset exchanges. The existing literature focuses on asset sales and purchases in monetary terms alone (e.g., Maksimovic and Phillips, 2001; Warusawitharana, 2008). Second, to the best of our knowledge, we contribute new incentives for asset sales and purchases, i.e., the tunneling and propping incentives, to the corporate asset literature. The existing 5

8 literature examines corporate asset transactions from either the investment efficiency incentive (e.g., John and Ofek, 1995; Maksimovic and Phillips, 2001; Warusawitharana, 2008) or financing incentive perspectives (e.g., Lang, Poulsen, and Stulz, 1995; Brown, James, and Mooradian, 1994; Asquith, Gertner and Scharfstein, 1994). Finally, we use asset valuation information to infer the quality of surrendered and acquired assets, which to date has gone unexamined in the literature. This study also extends the body of research on the expropriation and propping of minority shareholders by controlling shareholders. Johnson et al. (2000) conjecture that controlling shareholders have an incentive, legally or illegally, to expropriate (or tunnel ) minority investors when the legal environment and corporate governance system are weak. Friedman et al. (2003) extend the findings of Johnson et al. to show that managers (or controlling shareholders) may also have incentives to transfer their private resources to benefit minority shareholders. Using a sample of connected transactions in China, Peng et al. (2011) provide empirical support for both Johnson et al. (2000) and Friedman et al. (2003) by furnishing clear evidence of propping and tunneling occurring in the same company. Su et al. (2014) find that firms that pay less in cash dividends are associated with more related-party transactions, which represents wealth expropriation from general stockholders. In this paper, in contrast, we present the barter-type asset exchanges that take place in China as direct evidence of tunneling and propping by examining various aspects of the country s corporate restructuring processes and securities regulations that have previously been unexamined in the literature. This paper also contributes to the asset valuation literature in the accounting field. Jarrell (1979) finds that utility companies overvalue their assets to increase the price of their products. Our research complements that study by showing that the 6

9 manipulation of asset valuation can be used to achieve various goals, such as propping and tunneling in our context. The remainder of the paper is organized as follows. Section 2 reviews the asset sale and purchase literature. Section 3 outlines China s particular institutional background and develops our hypotheses. In Section 4, we describe our sample and data, followed by presentation of our empirical results in Section 5. Finally, our conclusions are presented in Section Review of Asset Exchange Literature The existing literature on corporate assets focuses on transactions involving payment in the form of cash, equity, and/or future considerations (Slovin, Sushka, and Polonchek, 2005) rather than barter-type asset exchanges. The overall market for corporate assets includes mergers, acquisitions, and partial asset sales. Asset exchanges are related to partial asset sales. Alexander, Benson, and Kampmeyer (1984) and Jain (1985) were among the first to show the valuation consequences of asset sell-offs. Using a sample of more than 1,000 voluntary sell-off announcements, Jain (1985) shows that there is a positive effect on the shareholders of both the sellers and buyers. 2 Subsequent studies offer various theories to explain the motives and valuation consequences of partial asset sales by corporations. The efficiency hypothesis is the dominant theory, but there are various views of efficiency. It is generally argued that managers reallocate resources efficiently through asset sales and purchases. Managers may sell assets if they discover that another party can manage them more efficiently. Hite, Owers, and Rogers (1987) investigate the valuation consequences of voluntary proposals to sell part of a 2 Alexander, Benson, and Kampmeyer (1984) find similar results but with a much smaller sample (53 announcements). 7

10 corporation s assets. They find that both successful and unsuccessful sellers reap statistically significant abnormal returns from initial proposal announcements but that unsuccessful sellers lose the initial gain at the offer termination. They interpret these findings as evidence that asset sales are associated with the movement of resources to higher-value uses. The rationale is that asset sales are in the best interests of stockholders if and only if the net sale proceeds exceed the present value of the net future cash flows from continued ownership and operation. Thus, potentially productive gains can be realized only by the transfer of the target assets from their current use to the buyer s control. John and Ofek (1995) offer an alternative view of efficiency. They argue that the divested asset s interference with the seller s other operations provides the motive for selling it. Hence, selling an unrelated asset leads to an increase in focus and more efficient operation of the core business. Using several accounting performance measures, such as operating margin and ROA, they find that the firm s remaining assets are more profitable after the sell-off. Maksimovic and Phillips (2001, 2002) provide both a theoretical model and empirical results to support the efficiency view. Their intuition is that some firms are more productive and can produce more than other firms from a given number of plants. They argue that firms adjust in size until the marginal benefit is equal to the marginal cost of production. As output prices increase, more productive firms realize a larger gain in value from the assets they control. As a result, they find it optimal to acquire plants from less productive firms in the industry. By the same token, a positive shock in an industry increases the opportunity cost of operating as an inefficient producer in that industry. Thus, industry shocks alter the value of assets and create incentives for transfers to more productive uses. Their empirical results 8

11 show that assets are more likely to be sold (1) when the economy is undergoing positive demand shocks, (2) when the assets are less productive than their industry benchmarks, (3) when the selling division is less productive, and (4) when the selling firm has more productive divisions in other industries. Warusawitharana (2008) develops a model that links asset purchases and sales to the fundamental properties of a value-maximizing firm. The key economic idea of this model is that firms engage in asset purchases and sales to move the firm toward its optimal size, which varies with profitability. Their empirical results show ROA to be strongly predictive of when firms will purchase or sell assets. In response to improved profitability, firms have the option of growing externally through asset purchases. Firms with a low degree of profitability can improve their average productivity of capital via asset sales. In summary, the foregoing studies characterize asset sales and purchases as a process that efficiently reallocates corporate resources. The existing literature also suggests alternative explanations for asset sales. The financing hypothesis, for example, argues that management values firm size and control, and is thus reluctant to sell assets for efficiency reasons alone. A more compelling motivation for selling assets is to obtain funds when alternative sources of financing are too expensive. This hypothesis also argues that the completion of an asset sale signals good news about the value of the asset because if its value were low, then the sale would not have taken place. Lang, Poulsen, and Stulz (1995) provide empirical results to support the financing hypothesis. They find that firms selling assets tend to be poor performers and/or to have high degrees of leverage, even when bankrupt firms and those in default are excluded. This result suggests that the typical firm selling assets is motivated to do so by its financial situation rather than a desire to efficiently reallocate corporate resources. Lang, Poulsen, and Stulz (1995) also report 9

12 that the stock-price reaction to asset sales is significantly positive for firms that are expected to use the proceeds to pay down debt, but negative and insignificant for those that are expected to keep the proceeds within the firm, which is also inconsistent with the efficiency hypothesis. Asset sales may also be an important way of resolving financial distress. Asquith, Gertner, and Scharfstein (1994) find that asset sales are a way of avoiding Chapter 11 but are limited by industry factors, with firms in distressed and highly leveraged industries less prone to engage in such sales. Brown, James, and Mooradian (1994) find significantly lower returns to shareholders when asset sale proceeds are used to repay debt than when they are retained by the firm. 3. Institutional Background and Hypotheses The peculiar institutional background of Chinese firms renders the motives for asset exchanges different from those discussed in the foregoing section The Restructuring Process of China s Listed Companies In the transition from a centrally planned economy to a market economy, the Chinese government has adopted a gradual approach by introducing private ownership to wholly state-owned enterprises without selling any state-owned assets (Wei, Xie, and Zhang, 2005; Chan, Fung and Thapa, 2007). Existing SOEs are first restructured into corporations, which then go public to raise private capital. There are three types of restructuring: peel-off, integration, and bundling. An existing SOE may peel off part of its operating assets to form a new independent corporation, a process that can be termed incomplete restructuring. The existing SOE becomes the parent of the new independent corporation and retains ownership of all of the peeled-off assets. Peel-off restructuring differs from the typical carve-out or spin-off. It differs from the typical carve-out in that the parent company 10

13 sells no existing assets to other investors, and hence there is no cash flow effect on the parent. It differs from a typical spin-off in that the newly independent corporation gains new investors through issuing new shares in the IPO process. As Aharony, Lee, and Wong (2000) point out, to render the new corporation more marketable and to attract public investors, parent companies have strong incentives to peel off only their most profitable business units for public offering and retain the nonproductive and unprofitable units. Another important incentive arises from the strict IPO quota system established by the Chinese government (Aharony, Wang, and Yuan, 2010). Prior to 1999, the total annual number of IPOs was subject to a quota system, meaning that the central government set a quota for the entire capital value of shares to be issued each year. This total amount was then allocated among local governments, which in turn were directed to identify key industries and nominate worthy companies for listing on the local stock exchanges. Thus, parent companies also had the incentive to strengthen the to-be-listed companies by peeling off the higher quality assets. Although this quota system was eliminated in 1999, the first aforementioned incentive (Aharony et al., 2000) still exists. Such incomplete peel-off restructuring leaves most of the financial and social burden with the remaining parent company, which may reduce that burden by improving the operating efficiency of its remaining assets. Another option is to reclaim the superior assets injected into the listed subsidiary during the restructuring process. One feasible way of doing so is to exchange lower quality assets for higher quality assets. 3 In addition, a sound financial condition in the listed subsidiary serves to exaggerate the tunneling incentives of its controlling shareholder (Peng et al., 3 Aharony et al. (2010) also find that Chinese parent companies expropriate their listed affiliates through the non-repayment of corporate loans. 11

14 2011). This discussion leads us to the following hypothesis, which is stated in alternative form. HYPOTHESIS 1: Listed companies produced by incomplete restructuring and in sound financial condition are more likely to exchange higher quality assets for lower quality assets (i.e., the tunneling hypothesis) Influence of Bright-line Regulations on Firm Listings We consider two situations in which parent companies have the incentive to prop up their listed affiliates by injecting higher quality assets to replace lower quality assets. The former assets may come from the parent companies themselves or from other companies. Although Chen and Yuan (2004) show that regulation based on accounting numbers, such as ROE, triggers opportunistic earnings management, the China Securities Regulatory Commission (CSRC) is notable for its use of bright-line regulations to monitor firm listings. 4 The first situation involves the incentive to avoid reporting a loss by the listed company. According to guidelines introduced by the CSRC in 1998, a listed firm will be designated a special treatment (ST) firm if it reports a net loss for two consecutive years. An ST firm s semi-annual report must be audited. If it reports a net loss for three consecutive years, it is suspended from normal trading, and investors can trade only under a particular transfer (PT) arrangement. If a PT firm has not become profitable by the following year, it is completely delisted. Although the literature considers reporting loss avoidance to be an important incentive for earnings management (Degeorge, Patel, and Zeckhauser, 1999), China s institutional setting 4 Chen and Wang (2007) show that, in China, bright-line rules may serve to reduce adverse selection problems. 12

15 gives managers even stronger incentive to engage in such management to avoid government scrutiny and delisting. The other situation in which a parent company has a strong incentive to engage in asset exchange to prop up its listed affiliate is during rights offerings (ROs) and seasoned equity offerings (SEOs). In the 1990s, listed companies were able to issue additional shares only through preemptive rights offered to their existing shareholders. Because of the lack of any other means for listed companies to raise capital and the insatiable demand for stocks from the investing public in China in the early 1990s, ROs were excessively abused by listed companies (Chen and Yuan, 2004). To curb this excessive activity, the CSRC adopted a minimum ROE of 10% (6% after 2001). 5 Since 2002, a similar threshold (10% of ROE) has regulated SEOs. As ROs and SEOs are the primary channels by which Chinese listed companies raise capital, qualification for such offerings is an important objective for parent companies. Li and Zhou (2005) also argue that listed companies are better able than unlisted companies to relieve unemployment problems and enhance the infrastructure development of the ministries to which they belong or the regions in which they operate. Thus, both the central and local governments that act as the ultimate controlling owners of these firms have strong incentives to help them to maintain their listing status and qualify to raise more funds. We thus predict parent companies to have strong incentives to replace listed companies lower quality assets with higher quality assets in the two aforementioned situations, as summarized in the following hypothesis. 5 Table 1 in Chen and Wang (2007) summarizes the regulations on ROs and SEOs in China. 13

16 HYPOTHESIS 2: Listed companies with the intention to avoid reporting a loss and to raise additional capital are more likely to exchange lower quality assets for higher quality assets (i.e., the propping hypothesis). If the market were efficient, then investors would be able to recognize the tunneling and propping behavior involved in asset exchanges. Cheung, Rau, and Stouraitis (2006) find that listed firms in Hong Kong announcing asset sales that are a priori likely to result in the expropriation of minority shareholders earn significantly negative abnormal returns in the days following the exchange announcement. Hence, we expect investors in Shanghai and Shenzhen to react negatively to asset exchanges arising from the tunneling incentive and positively to those arising from the propping incentive if the Chinese market is as efficient as Hong Kong s. We summarize these predictions in the following hypothesis. HYPOTHESIS 3: Investors react negatively to asset exchanges motivated by the tunneling incentive and positively to those motivated by the propping incentive surrounding the asset exchange announcement date. Given the differing quality of the exchanged assets, we predict differences in post-exchange firm performance, as summarized in the following hypothesis. HYPOTHESIS 4: Listed companies that exchange higher (lower) quality assets for lower (higher) quality assets experience a performance decline (improvement) in the post-exchange period. 14

17 4. Asset Exchange Data Description We hand-collected all 305 public announcements of asset exchanges between parent companies and other parties released by 229 listed companies on the Shanghai and Shenzhen Stock Exchanges during the period. Our sample period begins in 2000 because there were very few asset exchanges (only five in total) prior to that year. We also hand-collected IPO restructuring data from each company s IPO prospectus. Other data such as stock return and financial performance data were obtained from the China Stock Market & Accounting Research (CSMAR) database. 6 Table 1 presents the sample composition for each year from 2000 to 2006, classified by 10 major industry categories (two-digit SIC code). The original industry classifications were first obtained from the CSRC, and we then reclassified the industries into 10 categories based on Campbell (1996). As there are only three firms in the petroleum industries (SIC codes 13 and 29), we combine them with basic industries. Financial services industries (SIC codes 60-69) are included in the services category. As Table 1 shows, four industry groups, the basic (including petroleum), consumer durables, capital goods, and conglomerate categories, had a higher proportion of asset exchanges during the sample period (ranging from 38 to 54 cases, or 12.5% to 17.7%) than the remaining industry categories (ranging from 14 to 25 cases, or 4.6% to 8.2%). The table also shows that, in general, there were fewer asset exchanges in the early period than in the later period (the fewest cases, 10, were in 2000, and the most, 79, in 2003). (Insert Table 1 here) 6 The CSMAR is a leading data vendor that provides financial accounting and stock price data on all listed companies in China. It also offers corporate governance and mergers and acquisitions databases. The CSMAR database can be obtained from Wharton Research Data Services. 15

18 We present the types of exchange parties and types of assets exchanged in Table 2. Exchange parties are generally classified as related parties and non-related parties. Listed companies disclose this classification in their asset exchange announcements, and the definition of related parties should follow Chinese Accounting Standard 36 (CAS 36, issued in 2006), which is the same as International Accounting Standard 24 (IAS 24, revised in 2011). Related parties consist of the parent companies (i.e., the largest corporate shareholders) of listed companies, other large corporate shareholders, sister companies under common control with the listed companies, and others. 7 As shown in Panel A of Table 2, the majority of asset exchanges are between listed companies and their parent companies (219 cases, or 71.8% of the sample). Non-related parties account for just 49 cases (16.1%). However, non-related parties may be de facto related to listed companies because they are under the common control of the government, but they are not treated as related parties by CAS 36 or IAS 24. Because all of the non-related parties in our sample are non-listed companies, we cannot identify whether they are de facto related to listed companies. 8 As shown in Panel B of Table 2, five types of assets were exchanged in our sample: asset groups, equity shares, receivables, PPE, land and other tangible assets, and intangibles. An asset group is a group of assets and liabilities such as a production line and an operating unit. Equity shares refer to a company s equity ownership in a separate entity. 9 Receivables are exchanged assets that consist primarily of receivables. PPE, land, and other tangible assets include PPE, land, or other tangible assets such as inventories or a combination thereof. Intangibles refer to 7 The others category is disclosed as other related parties by the listed company, and we do not know their specific relationship with it. 8 Our research results remain qualitatively the same when we exclude the 49 assets exchanges with non-related parties. 9 The split share structure reform in 2005, which was designed to convert non-tradable shares to tradable shares, should have no effect on equity shares in our sample because they refer to the company s equity ownership in a separate entity, rather than the company s own stocks. 16

19 exchanged assets that primarily comprise intangible assets. Panel B shows that, of the assets surrendered, equity shares and receivables are the most popular types (accounting for 113 [37.0%] and 110 [36.1%], respectively). In contrast, more than half of all acquired assets are equity shares (176, or 57.7%), with about one quarter PPE, land, and other tangible assets (73, or 23.9%). (Insert Table 2 here) Table 3 reports the end-of-fiscal-year summary statistics for firms that exchanged assets in the subsequent year (thus, the period for the statistics reported is , i.e., one year ahead of the sample period). For comparison, we also present the statistics for all listed firms with sufficient data during the same period. As Table 3 shows, the sample firms tend to have a lower ROA. The mean ROA for the sample firms and all firms is 0.8% and 3.4%, respectively. The t-statistic for the test of mean differences is 5.80, with a statistical significance level of 1%. We observe the same pattern for the other profitability and performance measures, such as stock returns (SRET), cash holdings (CASH), and sales growth (SALESG), which suggests that the managers of firms with a low ROA, low stock returns, low cash holdings, and low sales growth are more likely to engage in asset exchanges with their parent companies. This finding is consistent with Warusawitharana s (2008) observation concerning corporate asset sales in the United States. Table 3 also shows that the other statistics of our sample firms, such as firm size (SIZE), market-to-book ratio (MTB), LEVERAGE, and PPE growth (PPEG), are similar to those of average listed firms. The deal value reported in the third to last row reveals the average and median 17

20 values of assets exchanged to be about RMB270 million and RMB98 million, respectively. (Insert Table 3 here) 5. Empirical Results Our empirical implementation tested the four hypotheses derived in Section 3. In this section, we discuss our research design for testing these hypotheses and present our empirical results Abnormal Asset Valuation During the asset exchange process, firms need to hire professional valuers to value both the surrendered and acquired assets. In an arm s length transaction, the valuation of the two should be equivalent and approximate the fair value of the assets. However, in China, professional valuers are not properly regulated and are typically not independent. If a manager wants to exchange lower quality assets for higher quality assets, he or she can collude with professional valuers to opportunistically manipulate the valuation of the two to match. We use the following numeric example to illustrate the concept of abnormal asset valuation. Suppose asset A s book value (BV) is $100 and fair value (FV) is $150, its market to book ratio (FV/BV) is 1.5. Asset B s book value is $100 and fair value is $200, its market to book ratio is 2.0. Since market to book ratio is a good proxy of investment opportunities and profitability, we infer that asset A has lower quality than asset B. If a listed company wants to exchange A for B without paying extra compensation, the company may value asset A at $200 (the transaction price which is denoted as P hereafter). The difference between transaction price ($200) and fair value ($150) is the abnormal valuation ($50). For asset B, assume the company fairly 18

21 values it at $200 and then its abnormal valuation is $0. Thus the company can exchange A for B without any additional compensation. The difference in abnormal valuation of A and B is positive ($50). The example illustrates that a higher abnormal valuation difference between surrendered asset (A) and acquired asset (B) indicates that the quality of surrendered asset (A) is lower than the quality of acquired asset (B). Intuitively, the abnormal valuation difference may indicate whether the book value of the exchanged assets is high or low relative to the transaction price, which indicates whether the assets are of high or low quality and whether the company is "stretching" the valuation in order to accomplish its tunneling or propping aims. The challenge is that the fair value of exchanged assets is unobservable. Based on the information disclosed in the asset exchange announcements, we employ the following model to determine the fair value of the assets. The abnormal valuation of the asset is the estimated residual of the model. OUTVAL (or INVAL) = a 1 OUTBOOK (or INBOOK) + a 2 OUTFIX (or INFIX) + a 3 OUTINT (or ININT) + a 4 RPT + a 5 AUDIT + a 6 INDDIR + a 7 FINCON + Industry Dummies + e (1) We run separate ordinary least square (OLS) regressions based on this model (without intercept) for surrendered and acquired assets. OUTVAL (INVAL) is the valuation of the surrendered (acquired) assets disclosed in the asset exchange announcements of listed companies. Book value is an important determinant of asset valuation, and hence the model includes OUTBOOK (INBOOK), which is the book value of the surrendered (acquired) assets. OUTFIX (INFIX) takes a value of one if the surrendered (acquired) assets include fixed assets, and zero otherwise. Fixed assets 19

22 are carried at their historical cost under the current CAS, and hence fixed assets are subject to a higher valuation. OUTINT (ININT) takes a value of one if the surrendered (acquired) assets include intangible assets, and zero otherwise. We include this variable because intangible assets are more difficult to revalue. RPT takes a value of one if the exchange party is a related party of the sample firm, and zero otherwise. We believe that it is easier for managers to collude with a related party in manipulating asset valuation. AUDIT takes a value of one if the exchange transaction is audited, and zero otherwise. INDDIR takes a value of one if the exchange transaction is supported by independent directors, and zero otherwise. FINCON takes a value of one if the exchange announcement is accompanied by an independent financial consulting report. The three aforementioned variables are introduced to control for the monitoring effects of auditors, independent directors, and professional consultants on asset valuation. Finally, we introduce nine industry dummies in the model. (Insert Table 4 here) Panel A of Table 4 reports the OLS regression results. As expected, the book value of assets (OUTBOOK and INBOOK) is significantly positively associated with the valuation of both surrendered and acquired assets. In general, the other independent variables have no significant effects on asset valuation. The adjusted R- squares for both regressions are above 90%, indicating that the model has very high predictive power. The abnormal asset valuation rate is the residual obtained from the regressions in Panel A scaled by the book value of exchanged assets. We then take the difference between the abnormal surrendered asset valuation rate and the abnormal 20

23 acquired asset valuation rate to infer the quality of the exchanged assets. As previously noted, a greater abnormal valuation difference indicates that the surrendered assets are lower in quality than the acquired assets, which is indicative of possible propping behavior by the exchange parties. Panel B of Table 4 presents the descriptive statistics of the abnormal valuation rates. We find that of the surrendered assets to be significantly higher than that of the acquired assets. The mean and median abnormal valuation rate differences are 13.5% and 10.8%, respectively, both significant at the 1% level. We now present the results of further empirical tests to explain this asymmetry in the valuation rates Tunneling and propping incentives for asset exchanges Hypotheses 1 and 2 posit that firms with tunneling and propping incentives behave differently when exchanging assets. We employ the following model to investigate the two types of incentive. ABVALDIF = b 0 + b 1 TUNNELING + b 2 PROPPING + b 3 ROA + b 4 SIZE + b 5 SRET + b 6 MTB + b 7 CASH + b 8 LEVERAGE + b 9 SALESG + b 10 PPEG + Industry Dummies + Year Dummies + e (2) The dependent variable, ABVALDIF, is the abnormal valuation difference between surrendered and acquired assets obtained from the regressions in model (1) and reported in Table 4. The independent variables are as follows. (1) TUNNELING: This variable takes a value of one if the listed firm has met both of the following two characteristics, and zero otherwise: (i) the listed firm 21

24 was peeled-off from an existing SOE during its IPO process, and (ii) its previous ROE was over 10%. As we discussed, firms with incomplete restructuring and healthy financial condition are more likely to have tunneling incentives. 10 (2) PROPPING: This variable takes a value of one if the listed firm has at least one of the following characteristics, and zero otherwise: (i) it had a previous net loss; (ii) it has a current net loss; (iii) its current ROE is lower than 1.5%; (iv) it is raising additional capital in the current year; and (v) it intends to raise additional capital in the next two years but does not currently fulfill the ROE threshold (either 10% or 6%). These characteristics are indications of propping incentives. (3) ROE, ROA, SIZE, SRET, MTB, CASH, LEVERAGE, SALESG, PPEG: These variables are as defined in Table 3. (4) Nine industry dummies and six year dummies. (Insert Table 5 here) Table 5 presents the OLS regression results of the foregoing model. Regression 1 includes only the tunneling incentive indicator (TUNNELING). The estimated coefficient on TUNNELING is , which is statistically significant at the 5% level. As discussed in our hypothesis development, a lower valuation difference implies that the quality of the surrendered assets is higher than that of the acquired assets. Thus, these results show that firms that have undergone incomplete restructuring and are in sound financial condition tend to exchange higher quality assets for lower quality assets (i.e., display a lower valuation difference), which is 10 Alternatively we define TUNNELING as ROE>10% only or ROE>10% and firm does not raise capital and the results remain qualitatively the same. This is consistent with the findings in Peng et al. (2011). 22

25 consistent with the tunneling hypothesis (Hypothesis 1). Regression 2 includes only the propping indicator (PROPPING). The estimated coefficient is 0.583, which is statistically significant at the 5% level. Thus, the results indicate that firms with an intention to avoid a loss or raise additional capital tend to exchange lower quality assets for higher quality assets, which is consistent with the propping hypothesis (Hypothesis 2). Regression 3 includes both TUNNELING and PROPPING. We find the estimated coefficients of both TUNNELING and PROPPING to remain significant at the 5% level. Of the control variables, only ROA and CASH have a significant effect on the valuation difference. In all three regressions, the estimated coefficients of ROA are negative and significant at the 5% level. Hence, highly profitable firms tend to exchange profitable assets for poorer quality assets, which is consistent with tunneling behavior. When listed companies are more profitable, their parent companies have more excuses to tunnel assets back. The same logic holds for CASH: the more cash the company holds, the more profits the parent company is able to tunnel. None of the other control variables has a significant influence on asset exchange valuation. The adjusted R-squares of the regressions range from 8.3% to 11.0%. As a robustness test, we use the observable transaction price (P) and book value (BV) to proxy for the abnormal valuation. In Regression 4 in Table 5, ABVALDIF is defined as the difference in book value of acquired and surrendered assets, i.e., (BV (acquired) BV (surrendered))/bv (surrendered). In Regression 5, we define ABVALDIF as the difference in price-to-book ratio, i.e., P/BV (surrendered) P/BV (acquired). Unlike the measure of abnormal valuation difference derived from regression Model (1) in Table 4, these two measures do not factor in any other factors in Model (1) which may affect the valuation of exchange 23

26 assets. As expected, the results reported in Regression 4 and Regression 5 are weaker but still significant at 10% level. It indicates that regression Model (1) for estimating abnormal valuation has incremental explanation power Market reaction to asset exchanges Hypothesis 3 states that if listed firms exchange higher quality assets for lower quality assets, then investors should react negatively, and vice versa. To test the market reaction to such exchanges, we employ the event-study methodology summarized by Campbell et al. (1997). The event date (day zero) is defined as the date the firm announces an asset exchange. For each company, we use an event period of 300 days (starting with day -279 and ending with day +20 relative to day zero). The first 259 days of this period (-279 through -21) are designated the estimation period, and the following 41 (-20 through +20) the event period. We run OLS regressions using a security s daily return as the dependent variable and the market daily return as the independent variable. The abnormal daily return is obtained from the regression model s residual. Table 6 reports the 3-, 5-, and 11-day cumulative abnormal returns (CAR) surrounding asset exchange announcements. We also divide sample firms by the tunneling and propping incentives. As Table 6 shows, all of the average CARs are positive. For example, the average 3-day window CAR is 0.706%, and the average 11-day window CAR is 0.947%. However, there are no significant differences between the subgroups for any of the window periods. For example, for the 3-day window, the average CAR for firms with tunneling incentives is 0.856% and that for firms without is 0.731%, but the difference is insignificant. Other windows, such as 1-, 2-, 5-, and 10-day windows, produce similar results to those in Table 6. 24

27 We also regress the various CARs on the abnormal valuation difference (ABVALDIF) and other factors such as return on assets, firm size, market-to-book ratio, cash holding, leverage ratio and industry effect. However, we do not find significant association between CAR and ABVALDIF. We thus conclude that investors in China are unable to recognize the various incentives driving asset exchanges, which casts doubt on the efficiency of China s capital markets. These findings also suggest that the mainland Chinese market is less efficient than Hong Kong s, where Cheung et al. (2006) found investors able to see through tunneling behavior. (Insert Table 6 here) 5.4. Post-exchange firm performance Although investors cannot see the profitability of exchanged assets in the short term, we expect that such assets will affect firm performance in the long term. We measure post-exchange performance by both stock returns and financial performance. We employ the following model to test our last hypothesis. BHAR (or AROA) = c 0 + c 1 ABVALDIF + c 2 ROA + c 3 SIZE + c 4 MTB + c 5 CASH + c 6 LEVERAGE + Industry Dummies + e (3) BHAR is the firm s 12-month (or 24-month) post-exchange buy-and-hold abnormal return starting one month after the asset exchange announcement month. AROA is 1-25

28 year (or 2-year) average ROA in the two years after the announcement year. The independent variables are as follows. (1) ABVALDIF: the abnormal valuation difference between surrendered and acquired assets obtained from regression model (1). (2) ROA, SIZE, MTB, CASH, and LEVERAGE: defined in the same way as in Table 3, but calculated in the year of the asset exchange in this regression. These variables are included to control for factors that may affect firm performance. (Insert Table 7 here) Table 7 reports the results of the foregoing regression model. In Regressions 1 and 2, when 12- and 24-month BHAR are used as the dependent variables, the estimated coefficients on ABVALDIF are and 0.051, with a significance level of 5%. The positive association between a valuation difference and long-term stock return performance indicates that firms outperform the market if they exchange lower quality assets for higher quality assets with their parent companies, which is consistent with Hypothesis 4. This result is intuitive because when higher quality assets are acquired, the firm earns more income and thus achieves better stock performance in the future. We find the same results in Regressions 3 and 4 when 1- and 2-year average ROA, respectively, are used as the dependent variable. These consistent results show that firms that exchange more (less) profitable assets for less (more) profitable assets will see poorer (improved) stock and financial performance in the long term. 26

29 In both regressions, ROA in the exchange year has significantly positive effects on firms future performance. Firm size (SIZE) has a positive effect on future financial performance but not on stock performance. Consistent with the existing literature, MTB is negatively associated with future stock performance but not with financial performance. Cash holdings (CASH) have a positive association with future financial performance but not with stock performance. Finally, we find firm leverage (LEVERAGE) to have no effect on either stock performance or financial performance. The adjusted R-squares are about 11% for the stock performance regressions and 39% for the financial performance regressions. 6. Summary and Conclusion This paper examines the causes and consequences of a sample of asset exchanges carried out by listed companies in China. Our dataset is unique because it is very rare for U.S. firms to engage in barter-type asset exchanges. Contrary to the efficiency and financing hypotheses examined in our review of the asset sale and purchase literature, we identify two other reasons for asset exchanges in China: tunneling and propping incentives. When firms are restructured incompletely from existing enterprises and are in sound financial condition, they tend to exchange higher quality assets for lower quality assets to help their unlisted parent companies. When the intention is to avoid a loss and raise additional capital, in contrast, the listed companies tend to exchange lower quality assets for higher quality assets with their parent companies. We find empirical evidence consistent with our hypotheses. We further examine whether investors can recognize the different incentives for asset exchanges in the short term, and the results suggest that they cannot, possibly 27