Pricing Efficiency and Market Transparency: Evidence from Corporate Bond Market

Transcription

1 Pricing Efficiency and Market Transparency: Evidence from Corporate Bond Market Jia Chen Guanghua School of Management Peking University Ruichang Lu Guanghua School of Management Peking University February 13, 2017 Abstract This paper investigates how post-trade market transparency affects pricing efficiency. Using the Phases implementation of TRACE in the corporate bond market, we find two primary results. First, when post-trade market transparency increases, the bond return drift after bond analyst report or after credit rating change becomes shorter, the delay with which bond prices incorporate market information reduces, and bond prices more closely approximate random walks. In contrast, when post-trade market transparency increases, bond analysts issue fewer reports. These effects are similar in bond groups sorted by liquidity, trading activity, and maturity and are robust to controlling for time-varying liquidity and trading activity. These results highlight the market transparency s impact on two dimensions of pricing efficiency: the increase in speed of information incorporation and the reduction of market participants incentives to produce information. Keywords: Bond Analyst Pricing Efficiency, Market Transparency, TRACE, Corporate Bond, We thank Bing Han, Jean Helwege, Laura Liu, Qiao Liu, Xiong Wei, Lucy White, Liyan Yang, and Haoxiang Zhu and brown bag seminar participants at Guanghua School of Management, UC Riverside, Renmin University as well as the discussants at the Five-star workshop at Tsinghua PBC School of Finance, CAFM conference, Australian Banking and Finance Conference, and Auckland Finance Meeting. 1

2 1 Introduction Pricing efficiency and market transparency are both fundamental issues regarding the functionality of financial markets, to which regulators and financial economists have paid substantial attention. A vast literature studies pricing efficiency because it is critical for financial markets to provide accurate information for resource allocation (Fama, 1970, 1991). There is also a large literature on how market transparency induced by different market structure impacts welfare, trading activity, and liquidity, theoretically in Biais (1993); Naik, Neuberger, and Viswanathan (1999); Madhavan (1995, 1996); Pagano and Röell (1996); Duffie, Dworczak, and Zhu (2016) and empirically in Boehmer, Saar, and Yu (2005); Gemmill (1996); Bessembinder, Maxwell, and Venkataraman (2006); Madhavan, Porter, and Weaver (2005); Edwards, Harris, and Piwowar (2007); Goldstein, Hotchkiss, and Sirri (2007); Asquith, Covert, and Pathak (2013). However, there is little empirical evidence on how market transparency impacts pricing efficiency, except for evidence based on laboratory experiments (Bloomfield and O Hara, 1999, 2000). We fill this gap by providing empirical evidence in the corporate bond market. There have been disagreements on what effects market transparency has on pricing efficiency among regulators and researchers. On the one hand, a beneficial view is often held by U.S. regulators: market transparency improves the price discovery, fairness, competitiveness, and attractiveness of U.S. markets (SEC, 1994). Several studies support this view using different contexts: market fragmentation versus consolidation (Madhavan, 1995), auction versus dealer markets (Pagano and Röell, 1996), experiments (Bloomfield and O Hara, 1999), publishing a benchmark (Duffie, Dworczak, and Zhu, 2016). The intuition is simple. Market transparency helps pricing efficiency because patterns in trades allow market participants to learn information from trades better, more easily enter the market, match better, search better, and thereby set their prices more efficiently. On the other hand, U.K. regulators have had concerns that increased market transparency may reduce liquidity and market efficiency (Franks and Schaefer, 1995). Several recent papers argue that market transparency may have adverse impacts on pricing efficiency in the contexts of correlated asset values (Asriyan, Fuchs, and Green, 2015), investors with or without immediate liquidity needs (Bhattacharya, 2016), and liquidity traders learning about fundamentals (Banerjee, Davis, and Gondhi, 2016). In these models, market participants incentives for collecting and producing information can be lowered by transparency, leading to lower pricing efficiency. Despite its importance, the empirical research on this topic is relatively little, partially due to the lack of data and the rarity of the regulation change events. In this paper, we offer 2

3 empirical evidence on how a particular form of market transparency, post-trade transparency, causally impacts pricing efficiency. 1 Specifically, we use the implementation Phases of Trade Reporting and Compliance Engine (TRACE) to examine whether disclosing the trading price and volume information after the trade affects pricing efficiency. 2 TRACE implementation provides a unique setting to study the impacts of market transparency on pricing efficiency. From the 1940s to July 2002, there was no public disclosure of transaction details in the corporate bond market, where transactions happened over-thecounter with private negotiations. 3 Starting from July 2002, the price and volume information of transactions became publicly disclosed after the trades were completed. National Association of Securities Dealer (NASD) 4 requires all transactions of U.S. corporate bonds by regulated market participants be reported to TRACE shortly after transaction completion. TRACE then publicly released the prices and volume information, which is known as dissemination. The implementation of TRACE was not applied to all bonds at the same time. Although TRACE began collecting all price and volume information for all corporate bonds in July 2002, it began the dissemination of trading information in Phases (Phase 1, 2, 3A, and 3B), gradually including more bonds into the dissemination. NASD assigned bonds into each Phase according to bond issue size, credit quality, and previous levels of trading activity. 5 Starting from February 2005, all corporate bonds price and volume information were publicly disseminated shortly after trade completion. We take advantage of these implementation Phases to conduct a difference-in-difference analysis that compares the pricing efficiency of bonds subject to a change in transparency to that of the bonds that are not. This research design helps control for the confounding effects of unobserved shocks to the corporate bond markets. We measure pricing efficiency in two dimensions. The first dimension is the speed of information incorporation, and we use three primary proxies for it. The first proxy is bond return drift after bond analyst report or credit rating change. We follow Gleason and Lee 1 Duffie, Dworczak, and Zhu (2016) show that under natural parameter assumptions, publishing a benchmark is socially efficient. Their setting is related to post-trade transparency and is more consistent with an increase in pre-trade transparency. 2 Anderson (2012) uses TRACE implementation Phases to study whether an improvement of bond liquidity can positively impact stock returns and finds that a 2.2% cumulative average equity return for firms appeared in TRACE s first dissemination Phase. Our paper examines how increases in market transparency impact pricing efficiency in the corporate bond market. 3 See Biais and Green (2007) and Piwowar (2011) for detailed account of evolution of the transparency regulations in the bond market. 4 The National Association of Security Dealers (NASD) changed its name to Financial Industry Regulatory Authority (FINRA) in See Table 1 for more details. 3

4 (2003) to define the return drift measure as the absolute value of eight-week sum of the abnormal return ( 8 ) after a bond analyst report or credit rating change. For the n=1 AR n second proxy, we follow Hou and Moskowitz (2005) to measure the delay with how bond prices respond to past information. Both proxies rely on an intuitive principle: a security price that is slow to incorporate information in credit events or market return is less efficient than a security price that instantaneously incorporates them. The third proxy is how closely bond prices approximate random walks. We follow Lo and MacKinlay (1988) among others to use variance ratio to measure the extent to which bond price movements deviate from random walks. We define this variable as the absolute value of the ratio of the variance of 2-week log returns divided by two times the variance of 1-week log returns minus one. The second dimension is information production. We follow Griffin, Kelly, and Nardari (2010) to consider the number of bond analysts reports as a measure of information production, which is a building block of pricing efficiency. A higher number of bond analysts reports means higher information production intensity. We have two key findings. First, we find that post-trade transparency of price and volume leads to smaller drift, lower delay, and less deviation from random walks. According to the difference-in-difference regression analysis, drift after bond analyst reports decreases by about 40% after Phase implementation in Phases 3A and 3B compared to pre-phase levels. Drift after credit rating change reduces by about 45% after Phase implementation in Phase 3A and 3B. Drift does not experience a significant reduction in Phase 2. Delay decreases by about 17% after Phase implementation for Phase 2 bonds, 19% for Phase 3A bonds, and 18% for Phase 3B bonds. These results are robust to using different ways of measuring drift and delay and logarithm and logistic transformations of the drift and delay measures. Variance ratio decreases by about 22% after Phase implementation for Phase 2 bonds, 13% for Phase 3A bonds, and 22% for Phase 3B bonds. These findings suggest that greater market transparency improve the speed of information incorporation, which is consistent with the conclusions of Madhavan (1995), Pagano and Röell (1996), and Bloomfield and O Hara (1999). Second, we find that post-trade transparency significantly reduces the number of bond analyst reports. The reduction ranges from 28% to 46% in different phases. These findings are consistent with the conjecture that transparency can lead to reduced incentives for analysts to produce bond-specific information (Asriyan et al., 2015; Bhattacharya, 2016; Banerjee et al., 2016). We further investigate whether the improvement of pricing efficiency varies depending on the illiquidity and trading activity of the bonds. First, we conduct difference-in-difference analysis for different bond groups sorted by illiquidity and trading activity. Using Amihud 4

5 as the illiquidity measure and the ratio of volume divided by issue amount as the trading activity measure, we find that the difference-in-difference estimates do not significantly differ between bonds with high and low illiquidity or bonds with high and low trading activity. Second, we control for time-varying illiquidity and trading activity measures in the difference-in-difference regressions. The results show that controlling for time-varying illiquidity and trading activity does not change the magnitude or statistical significance of the original difference-in-difference estimates. Therefore, the impact of market transparency on pricing efficiency is relatively uniform across bonds with different ex-ante liquidity and trading activity characteristics and is robust to controlling for time-varying illiquidity and trading activity. We also study whether higher bond market post-trade transparency has a spill-over effect on the pricing efficiency in the stock market. We repeat the difference-in-difference regression analysis using the stocks corresponding to the treated bonds and the stocks corresponding to the control bonds and use as the dependent variable the delay measure based on the stock prices. We find no significant improvements in pricing efficiency or price informativeness of stocks matched with the treated bonds relative to that of the stocks matched with the control bonds. Our paper is closely related to and contributes to two strands of literature. First, it adds new empirical evidence to the literature that directly studies market transparency on pricing efficiency. This literature uses rule changes in equity market (Boehmer et al., 2005; Gemmill, 1996; Madhavan et al., 2005) and find mixed evidence. A potential reason for the mixed evidence from equity market is that the degree of market transparency is already high before the rule changes. The implementation of TRACE in the corporate bond market is potentially a more substantial rule change. This literature also studies the effect of market transparency on pricing efficiency in a setting of laboratory experiment (Bloomfield and O Hara, 1999, 2000), but there is a lack of empirical analysis using real market data. Our paper fills that void. Second, our paper contributes to the literature that studies how the implementation of TRACE impacts trading costs and liquidity (Bessembinder et al., 2006; Edwards et al., 2007; Goldstein et al., 2007) and price dispersion and trading (Asquith et al., 2013). Although transaction costs is an essential building block of pricing efficiency, these papers do not directly examine pricing efficiency. Not only does our analysis contribute to the above literature, but it is also relevant to the ongoing regulatory changes. During and after the 2008 financial crisis, the issues relating to the trading and the valuation of over-the-counter instruments during the crisis has made many people promote greater transparency in such markets. As a result, regulatory 5

6 agenda has considered implementing for these instruments trade reporting systems similar to TRACE and Transaction Reporting System (TRS). Such a system became effective for OTC swap trades in January 2013, and TRACE has expanded its coverage to including agency debentures, mortgage-backed securities, 144-A private placement, and asset-backed securities in recent years. Our paper can help understand the potential impacts of these regulatory changes. The rest of this paper is organized as follows. Section 2 reviews the related literature. Section 3 discusses key empirical measures and the research design. Section 4 describes TRACE s history, data sources, and key variable construction. Section 5 presents descriptive statistics of key variables. Section 6 shows primary empirical results. Section 7 concludes. 2 Related literature Market transparency and its relation to how markets functions have received significant attention from researchers. A number of papers point out the positive effects of market transparency on the functioning of markets. Madhavan (1995) argues that fragmented markets will not become consolidated by themselves unless transparency of trade price information is required. Madhavan (1996) demonstrates that an increase in transparency of order flow information can lead to higher price informativeness but can potentially lead to higher price volatility and lower liquidity. Pagano and Röell (1996) define transparency as the extent to which market makers can observe the size and direction of the current order flow and find that greater transparency generates lower trading costs for uninformed traders on average, although not necessarily for every size of the trade. Using laboratory experiments, Bloomfield and O Hara (1999) conclude that trade disclosure significantly improves the informational efficiency of the markets but can cause spread to widen. They also find that quote disclosure has little effects. A few recent theoretical papers discuss how market transparency might negatively impact efficiency. Asriyan, Fuchs, and Green (2015) argue that when asset values are correlated, higher market transparency does not necessarily lead to higher welfare. In a multi-period auction model, Bhattacharya (2016) shows that potential counter-parties may delay their trades when there is transparency because they can monitor transaction prices and learn more before participating. As a result, investors with immediate liquidity needs suffer in terms of revenue. Banerjee, Davis, and Gondhi (2016) find that an increase in transparency can lower liquidity traders incentives for learning about fundamentals and hence lead to lower informativeness. Goldstein and Yang (2016) and Han, Tang, and Yang (2016) both discuss the potential adverse effects of public information transparency on real efficiency. 6

7 There is empirical evidence on the effect of market transparency in stock markets. The effect of pre-trade transparency is mixed. Boehmer, Saar, and Yu (2005) study the introduction of Open Book on the NYSE in They find that greater transparency made prices more informationally efficient. But, Madhavan, Porter, and Weaver (2005) obtain opposite results for the Toronto Stock Exchange. They show that execution costs increased after a rule change on the Toronto Stock Exchange in Focusing on post-trade transparency, Gemmill (1996) finds that there is no significant gain or loss in liquidity from delayed publication of block trades, as the spreads and the speed of price adjustment are not affected by the disclosure regime. Four empirical studies have examined the liquidity or trading costs due to TRACE Phase implementation. Bessembinder, Maxwell, and Venkataraman (2006) study the bonds in Phase 1 using transaction data from the National Association of Insurance Commissioners. They find that the transaction costs reduce after the implementation of Phase 1 of TRACE. Edwards, Harris, and Piwowar (2007) examine the imputed transaction cost for Phase 2 bonds and find that transparent bonds have lower transaction costs. Goldstein, Hotchkiss, and Sirri (2007) use a controlled experiment of 120 BBB Phase 2 bonds. Using a matching sample, they find that the transaction costs for all but the group with the smallest trade size experience a reduction. Asquith, Covert, and Pathak (2013) use the Phase 1, 2, 3A, 3B of TRACE setting and find that the transparency causes a significant decrease in price dispersion for all bonds and a considerable reduction in trading activity for some categories of bonds. They conclude that the mandated transparency may help some investors and dealers through a decline in price dispersion while harming others through a reduction in trading activity. 3 Research design 3.1 Measures of pricing efficiency in corporate bond market In this section, we discuss our measures of pricing efficiency, including return drift after bond analyst reports, return drift after credit rating change, delay, variance ratio, and the number of bond analyst reports Speed of information incorporation Return drift Following Gleason and Lee (2003), we consider return drift after credit events as the first proxy of speed of information incorporation. We use an eight-week window after bond analyst report or credit rating change to measure return drift. Specifically, we define the drift variable 7

8 as the absolute value of the eight-week sum of the abnormal return ( 8 n=1 AR n ) after the bond analyst report, excluding the week of the report. We choose the eight-week window for two reasons. First, because corporate bond trading happens relatively infrequently, it can take time for bond prices to incorporate information. Second, a too short time window may primarily capture the initial reaction to the events and omit a large part of the drift. To better capture information diffusion instead of over- or under-reaction, we use a long enough window. Because both directions of the reaction can generate meaningful drift and because we care about the magnitude of the drift rather than its direction, we take absolute value when constructing the drift variable. We use bond analyst reports as the events for following reasons. First, corporate bond investors are almost exclusively institutions. The average level of investor sophistication is higher in the corporate bond market than in the equity market. Institutional investors likely understand well how to utilize the bond analysts reports. Therefore, the drift after bond analyst report can well capture how the average investors in the corporate bond market incorporate information. Second, unlike quarterly earnings announcements, the fixed income reports are issued throughout the quarter. Because of their frequency and timeliness, these reports have become a vital source of information for many users of corporate financial reports. According to Bond Market Association (BMA, 2004), fixed income research analysts play a significant role in informing the marketplace about particular issues or securities. Hence, the bond analyst reports are critical in promoting market efficiency in the corporate bond price discovery process. In an alternative specification of the events, we use return drift after a credit rating change by Moody s to measure the pricing efficiency. Rating agencies have preferential access to both publicly available information and proprietary information. Because rating agencies are alternative information intermediaries with high reputation capital at stake, analysis based on drift after rating changes can serve as an independent check on the analysis based on the drift after bond analyst reports Speed of information incorporation Delay We adopt Hou and Moskowitz s (2005) Delay measure as the second proxy for speed of information incorporation. This measure is an estimate of how quickly prices respond to public information in the market and portfolio return movements. We use weekly bond returns to calculate this measure. Although returns at a higher frequency, such as daily, can potentially provide more dispersion in delay, they may introduce confounding microstructure influences such as bid-ask bounce and non-synchronous trading. We consider bond market return, stock market return, and bond portfolio return as the 8

9 relevant market information to which bonds respond. Using the date of a Phase as the cut-off point, we consider 1 year before the Phase and 1 year after the Phase as the event window. For each of these two periods, we then separately estimate a regression of each bond s weekly returns on contemporaneous and four weeks of lagged returns on the bond market, stock market, and bond portfolio. Ret B i,t = α i +β BM i Ret BM,t + β SM i Ret SM,t + β BP i Ret BP,t + 4 n=1 4 n=1 4 n=1 δ BM,( n) i Ret BM,t n + δ SM,( n) i Ret SM,t n + δ BP,( n) i Ret BP,t n + ɛ i,t, (1) where Ret B i,t is the return on bond i, Ret BM,t is the return on value-weighted bond market return in week t, Ret SM,t is value-weighted stock market return 6 in week t, and Ret BP,t is value-weighted bond portfolio return in week t. If the bond responds immediately to market news, then β i will be significantly different from zero, but δ ( n) i will not differ from zero. Otherwise, δ ( n) i will differ significantly from zero. This regression identifies the delay with which a bond s prices respond to market information if expected returns are relatively constant over weekly horizons. δ ( n) i The Delay measure is one minus the ratio of the R 2 from above regression restricting = 0 for all n, over the R 2 from above regression without the restriction. Delay = 1 R2 δ ( n) =0, n [1,4] R 2. (2) The higher this number, the more return variation is captured by lagged returns, and hence the slower are bond prices to incorporate new market information Deviation from random walk Variance ratio Much work has argued that informationally efficient prices follow random walks and has tested this hypothesis using variance ratio (Lo and MacKinlay (1988) and many others). Under the null hypothesis that bond prices follow a random walk with uncorrelated increments, variance ratio is equal to one. When it is above one, the autocorrelation of increments is positive. When it is below one, the autocorrelation of increments is negative. Because both negative and positive autocorrelation indicate departures from a random walk, we define 6 Stock market return is the value-weighted CRSP market index return. 9

10 variance ratio as the absolute value of the proportion of the variance of 2-week log returns divided by two times the variance of 1-week log returns minus one. If this number is close to zero, price movements approximate a random walk Information production Number of bond analyst reports We follow Griffin, Kelly, and Nardari (2010) to consider the number of bond analysts reports as a measure of information production. According to Bond Market Association (BMA, 2004), fixed income research analysts play an important role in informing the market about particular issues or securities. These bond analyst reports are critical in promoting efficiency in bond price discovery. Therefore, when there are more bond analyst reports about a specific bond, the information production by bond analysts about this bond is more intense. In other words, a higher number of bond analysts reports means higher information production intensity. 3.2 Difference-in-difference method We estimate a difference-in-difference regression to empirically test and quantify the impact of market transparency on pricing efficiency in bond market. y i,t = b 0 + b 1 T reated i + b 2 P ost t + b 3 T reated i P ost t + ɛ i,t, (3) where y i,t is a measure of pricing efficiency for bond i in week t, T reated i is one if bond i s dissemination status changes and P ost t is a dummy variable which equals to one in the period after the dissemination starts and zero otherwise. The coefficient b 3 on interaction term, T reated i P ost t, captures the difference-in-difference effect. This parameter measures how difference in pricing efficiency between treated bonds and control bonds changes before and after bonds public dissemination status changes. Appendix A. Detailed variable description is in Since TRACE Phase implementation is directly related to corporate bonds rating and issue size, we define the control groups as follows to get better match treated and control bonds. We can identify control bonds as bonds that are already in dissemination before a Phase begins or bonds that become disseminated in an upcoming Phase. Specifically, we use Phase 3A and 3B bonds as the control bonds for the Phase 2 and use Phase 1 and 2 bonds as the control bonds for Phase 3A and 3B. 7 We do not use the Phase 1 in the difference-in- 7 Although Phase 3A bonds and 3B bonds are more comparable to each other, we do not include Phase 3A bonds in control bonds for Phase 3B bonds because Phase 3A and 3B are only four months apart and because our test window is one year surrounding a Phase start date. 10

11 difference analysis because complete bond return data are not available for the period before TRACE started. 8 Therefore, we cannot compare pricing efficiency before and after Phase 1. Panel A of Table 1 describes the treatment and control groups of bonds for each Phase. We face a trade-off when choosing measurement windows before and after Phase implementation. On the one hand, to focus on the immediate effect of dissemination, we should use a short time window which is close to the event. On the other hand, to have enough observations for bond analyst reports, credit rating changes, the number of bond analyst reports as well as for estimating delay and variance ratio, we should use a longer time window. In general, we use as the measurement window the period between two Phase implementation. Details could be found in Panel B of Table 1. 4 TRACE, Data, and Variable Construction 4.1 Implementation Phases of TRACE In this section, we describe the background of Trade Reporting and Compliance Engine (TRACE). Price and volume information of completed U.S. corporate bond transactions became publicly disclosed when NASD launched TRACE in All transactions in U.S. corporate bonds by regulated market participants were required to be reported to TRACE on a timely basis. NASD then made this information transparent by publicly releasing it. NASD called this disclosing process disseminating. TRACE now covers any US dollardenominated debt security that is depository-eligible and registered with SEC or issued according to Section 4(2) of Securities Act of 1933 and purchased or sold under Rule 144a. When NASD implemented TRACE on 1 July 2002, it required all transactions on TRACE-eligible securities to be reported within 75 minutes of trading. As described in Table 1, NASD began disseminating price and volume data for trades in selected investment-grade bonds with the original issue of $1 billion or greater. We call these bonds Phase 1 bonds. The dissemination occurred immediately upon reporting for these bonds. Additionally, the 50 high-yield securities previously disseminated under FIPS were transferred to TRACE, whose trades were disseminated. We denote these bonds NASD50. About 520 securities had their information disseminated by the end of After NASD and SEC had approved the expansion of TRACE beyond Phase 1, Phase 2 of TRACE was implemented on 3 March 2003, and it expanded dissemination to include smaller 8 Although the data from National Association of Insurance Companies (NAIC) covers the period before Phase 1, it is not as complete as TRACE because it contains only transaction data for insurance companies. 9 Price and volume information of corporate bonds was publicly available in the 1930s and 1940s when corporate bonds were primarily traded on exchanges. 11

12 investment grade issues. The securities added into dissemination include those with at least $100 million par value or greater and rating of A- or higher. Also, dissemination began on 14 April 2003 for a group of 120 investment-grade securities rated BBB. We denote these BBB bonds NASD120. After Phase 2 had been implemented, the number of disseminated bonds increased to about 4,650 bonds. Meanwhile, the NASD50 subset did not remain constant over the period. On 13 July 2003, the NASD50 list was updated, and the list was then updated quarterly for the next five quarters. In order to have a clean and simple control group, we exclude NASD50 and NASD120 bonds from our analysis. Finally, on 22 April 2004, after TRACE had been in effect for some bonds for almost two years, NASD approved the expansion of TRACE to nearly all bonds. The last Phase came in two parts, which NASD designates as Phase 3A and Phase 3B. The distinction between Phase 3A and 3B is that Phase 3B bonds are eligible for delayed dissemination. Dissemination is delayed if a transaction is over $1 million and occurs for a bond that trades infrequently and is rated BB or below. Besides, dissemination is delayed for trades immediately following the offering of TRACE-eligible securities rated BBB or below. In Phase 3A, effective on 1 October 2004, 9,558 new bonds started having their information about trades disseminated. In Phase 3B, effective on 7 February 2005, an additional 3,016 bonds started dissemination, though sometimes with delay. According to NASD, at that point, there was real-time dissemination of transaction and price data for 99 percent of corporate bond trades. In an effort parallel to increasing the number of bonds with disseminated trade information, NASD reduced the time delay for reporting a transaction from 75 minutes on 1 July 2002, to 45 minutes on 1 October 2003, to 30 minutes on 1 October 2004, and to 15 minutes on 1 July On 9 January 2006, the time delay for dissemination was eliminated. Since most bonds trade infrequently, our analysis uses one week as the primary unit of time. Therefore, we do not focus on changes in time to dissemination, but instead on new dissemination. 4.2 Data sources Bond trading and pricing data are from TRACE Enhanced. To obtain bond characteristics, including maturity date and bond rating, we use the Mergent s FISD (Fixed Income Security Database). Financial and accounting data are from Compustat, and equity return data are from CRSP. The source of sell-side bond analyst report data is Investext, a provider of full-text analyst reports. In our sample, the sell-side bond analyst report data cover the period from 2001 to The intersection of the bond pricing and debt report data is the period from 1 12

13 July 2001 (i.e. one year before the implementation of TRACE) to 7 February 2006 (i.e. one year after the dissemination of the Phase 3B bonds) because this period aligns with the implementation of Phases. We manually collect bond analysts reports and code the names of the analysts and brokerage firms who issue the reports, report dates, names of the companies the reports are about, and the analysts recommendation. We exclude reports about industries, geographic areas, general economics, and reports that are aggregated either by industry or time, which often repeat previously published information. 4.3 Individual bond, bond market, and bond portfolio returns Both return drift and delay measures rely on bond returns, which we calculate at a weekly frequency as follows. For bond i in week t, we take all trades within the week and calculate the clean price for the week as the transaction size-weighted average price of these trades. Returns are then calculated as ( ) Pi,t + AI i,t + C i,t Ret i,t = ln, (4) P i,t 1 + AI i,t 1 where P i,t is the transaction size-weighted clean price 10, AI i,t is accrued interest, and C i,t is coupon paid in week t. Coupon rates and maturities are obtained from FISD. To construct the drift, delay, and variance ratio measures, we need bond market returns and bond portfolio returns. We define corporate bond market return as the amount outstanding weighted average of bond returns for all bonds from TRACE. 11 We follow Bessembinder, Kahle, Maxwell, and Xu (2009) to create portfolios segmented by both bond rating and time-to-maturity and calculate amount-outstanding weighted bond portfolio returns. segment bonds by Moody s six major rating categories (Aaa, Aa, A, Baa, Ba, and below Ba) and three time-to-maturity categories. For investment grade bonds, the time-to-maturity cutoffs are 0 to 5 years, +5 to 10 years, and +10 years. For non-investment grade bonds, the cutoffs are 0 to 6 years, +6 to 9 years, and +9 years. These cutoffs are designed to ensure roughly equal terciles. Because the Aaa sample size is too small to be split into three subsets based on maturity, we follow Bessembinder, Kahle, Maxwell, and Xu (2009) to divide the Aaa sample into two maturity categories, 0 to 7 years and +7 years. The procedure above provides a total of seventeen matching portfolios. The portfolio return is the amount outstanding weighted return of bonds in the portfolio. We match each 10 Bessembinder, Kahle, Maxwell, and Xu (2009) recommend calculating prices as the transaction sizeweighted averages of prices because it minimizes the effects of bid-ask spread in prices. Bao and Pan (2013) and Bao, Pan, and Wang (2011) also use transaction size-weighted averages of prices. 11 According to NASD (2005), TRACE covers 99% of corporate bonds. We 13

14 bond with the portfolio of same rating category and the same time-to-maturity category. We then follow Bessembinder, Kahle, Maxwell, and Xu (2009) to construct a bond s abnormal return (AR) as the difference between an individual bond s return and the matching portfolio s return. 5 Summary statistics Table 2 reports the summary statistics of efficiency measures for each Phase implementation. Panel A of Table 2 is about drift after bond analyst report. Three patterns emerge from this panel. First, bonds with smaller issue sizes and lower ratings have larger drift in all Phases. For example, drift is for the Phase 2 treated bonds, whereas it is for the Phase 2 control bonds, which are bonds in Phase 3A and Phase 3B. Phase 2 bonds are larger than and have better rating than bonds in Phase 3A and Phase 3B. This pattern is consistent with the intuition that bonds with large issue size and high rating have better information environment and the market more quickly incorporates information into their prices. Second, bonds experience a decrease in drift after the Phase dissemination. For example, drift is for Phase 3A treated bonds before the dissemination and is after the dissemination. Third, the reduction in Drift from pre- to post-dissemination periods is relatively larger for treated bonds than for control bonds for Phase 3A and 3B. The differencein-difference effects capture the net impact of dissemination, and they are negative for Phase 3A and 3B. Panel B provides the summary statistics for post-credit rating change drift, and we find similar patterns as those in Panel A. These results indicate a reduction in drift after bond analyst report or drift after credit rating change after Phase implementation, which is consistent with prices more quickly incorporating information of credit events when market transparency is higher. The change in drift for the control bonds is close to 0 for Phase 3A and 3B but is the Phase 2 control bonds, which are bonds in Phase 3A and 3B. As a result, the differencein-difference effect is positive for Phase 2 in Panel A and B. One possible explanation is that there is a strong spill-over effect from Phase 2 bonds to Phase 3A and 3B bonds after the implementation of Phase 2. We do not observe a similar spill-over effect in Phase 3A and 3B because their control bonds (Phase 1 and 2 bonds) are already subject to dissemination. In contrast, the Phase 2 control bonds are not yet in dissemination. Panel C of Table 2 reports the summary statistics of Delay. The mean values of treated bonds Delay decrease for all three Phases. Specifically, treated bonds average Delay decreases by 0.16 for Phase 2 (from 0.66 to 0.50), by 0.06 for Phase 3A (from 0.58 to 0.52), 14

15 and by 0.07 for Phase 3B (from 0.62 to 0.55). The change in Delay shows mixed results for control bonds. The mean of control bonds Delay decreases by 0.05 for Phase 2 (from 0.61 to 0.56) and increases by 0.09 for Phase 3A (from 0.36 to 0.45) and increases by 0.07 for Phase 3B (from 0.38 to 0.45). After examining the change in Delay for treated and control bonds separately, we calculate the difference between the treated and control bonds. These differences are for Phase 2 ( ), for Phase 3A ( ), and for Phase 3B ( ). These numbers show that after Phase implementation Delay decreases more for the treated bonds than for the control bonds. Comparing these differences to the mean delay measures before Phase implementation, we find that the economic magnitudes of these effects are large. For example, for Phase 2, is -16.4% (-0.11/0.66) relative to the mean Delay of treated bonds (0.66) and -21.6% (-0.11/0.50) relative to the mean Delay of control bonds (0.50). These results indicate that Delay significantly reduces after Phase implementation. This finding is consistent with bond prices more quickly incorporating market information when market transparency increases. Panel D of Table 2 reports the summary statistics of Variance ratio. The mean values of treated bonds Variance ratio decrease for all three Phases after Phase implementation. Specifically, treated bonds Variance ratio decreases by for Phase 2 (from 0.41 to 0.32), by for Phase 3A (from 0.35 to 0.32), and by for Phase 3B (from 0.34 to 0.29). The mean values of control bonds Variance ratio show increases. Control bonds average Variance ratio increases by for Phase 2 (from 0.34 to 0.35) and increases by for Phase 3A (from 0.28 to 0.31) and decreases by for Phase 3B (from 0.27 to 0.31). The difference-in-difference effects are for Phase 2 ( ), for Phase 3A ( ), and for Phase 3B ( ). The economic significance of these effects is large. For example, for Phase 2, is -22.9% (-0.094/0.41) relative to the mean Variance ratio of treated bonds before Phase implementation. These results indicate that Variance ratio decreases after Phase implementation, indicating that bond prices more closely follow random walks after market transparency increases. Panel E of Table 2 provides summary statistics of the number of the bond analyst reports. In Phase 2, the treated (1.35) and control (1.38) bonds have similar numbers of bond analyst reports before Phase implementation. However, after the Phase implementation, the control bonds have roughly the same number of reports while the treated bonds receive a less number of reports (0.86). The reduction is economically significant, dropping by 0.5 reports with an average number of 1.35 before Phase implementation. In Phase 3A, both treated and control bonds receive more reports in the post-phase period than in the pre-phase period, and the 15

16 control bonds average number of reports increases more than that of the treated bonds. In Phase 3B, while the treated bonds experience a reduction in the average number of reports after Phase implementation, the control bonds experience an increase. Overall, the effect of TRACE on the number of bond analyst reports are negative and economically significant. These results are consistent with the intuition that information production is less intense when market transparency is higher. 6 Difference-in-difference regression analysis 6.1 Drifts Table 3 presents difference-in-difference regression results for Drift after bond analyst report. The dependent variable is the absolute value of the sum of weekly abnormal return after the announcement week of a bond analyst report. Among the independent variables, Treated is a dummy variable equal to 1 if the bond is in dissemination and zero otherwise. Post is a dummy variable equal to 1 if it is after the Phase implementation and zero otherwise. Treated Post is an interaction term between Treated and Post. We cluster the standard errors at the firm level for all regressions. We are interested in the coefficient on Treated Post, which captures the impact of the market transparency on return drift. Three key results emerge from Table 3. First, the difference-in-difference effect on Drift is significantly negative for Phase 3A. The estimated coefficient of the interaction term shows that after phase implementation, Drift on average reduces by 1.2%. Relative to the average Drift for Phase 3A treated bonds before Phase implementation (2.8%, see Panel A of Table 2), this is 42.9% reduction. Therefore, the decrease in Drift of Phase 3A is both statistically and economically significant. Second, the reduction in Drift is strong for Phase 3B treated bonds, which are noninvestment grade bonds and have lower credit ratings than Phase 2 and Phase 3A treated bonds. The treatment effect is -1.9%, and it means a 52.8% reduction relative to the average Drift of Phase 3B treated bonds (3.6%, see Panel A of Table 2). The magnitude of these numbers is larger than those for Phase 3B. The result suggests that those bond with worse information environment may benefit more from an increase in market transparency. Third, the interaction term for Phase 2 is marginally significant at 10% level and positive (0.018 with a t-statistic 1.77). These bonds par values are at least $100 million or greater, and their ratings are of A- or higher. Companies with strong cash flows and good information environment are typical issuers of these bonds. A potential explanation is the strong spillover effect to the control bonds as suggested in Table 2. The control group bonds are not 16

17 subject to dissemination status change before and after Phase 2. However, due to the distinct feature of bond payoff function, bonds already in dissemination that have similar cash flows and credit risks can be substitutes for similar bonds not in dissemination, and thus can serve as pricing benchmarks. Therefore, the interaction term might be insignificant because the control bonds experience an improvement in information environment. 12 This explanation is consistent with the evidence that the Post is significantly negative in Phase 2 regression but insignificant in Phase 3A and 3B regressions. Overall, drift after bond analyst report decreases after TRACE phase implementation. This evidence supports that greater market transparency leads to quicker information incorporation of bond prices. Table 4 presents difference-in-difference regression results for Drift after credit rating change. The dependent variable is the absolute value of the sum of (+1, +8) week abnormal return after the announcement week of a credit rating change. Drift does not significantly change for the Phase 2 bonds but does significantly decrease for Phase 3A and 3B bonds. The difference-in-different effect is 2.2% with a t-stat of 1.63 for Phase 2 and is -3.5% with a t-stat of 2.98 for Phase 3A and -3.5% with a t-stat of 2.15 for Phase 3B. The economic significance of the effects for Phase 3A and 3B is large. They mean a 47.3% reduction in Phase 3A treated bonds average Drift compared with the average Drift before Phase 3A (0.074, see Panel B of Table 2) and a 46.7% reduction in Phase 3B treated bonds average Drift compared with the average Drift before Phase 3B (0.075, see Panel B of Table 2). The evidence in this table confirms the results in Table 3 and supports the view that bond prices more quickly incorporate information when market transparency is greater. 6.2 Delay We report difference-in-difference regression results for Delay in Table 5. The dependent variable is Delay defined in Equation 2. The coefficients on Treated Post are -0.11, -0.15, and -0.15, showing that Delay decreases after Phase implementation for all three Phases. These results confirm those of the summary statistics in Table 2. These differencein-difference estimates are economically significant. The coefficient for Phase 2 means a -16.7% reduction compared with the average Delay of treated bonds before Phase implementation (0.66, see Panel C of Table 2), and the coefficients for Phase 3A and 3B represent a -19.0% and a -17.7% decrease compared with the average Delay of treated bonds for the corresponding Phase (0.58 and 0.62, see Panel C of Table 2). 12 The the spill-over effect would be minimal in Phase 3A and 3B because the respective control groups are bonds already subject to dissemination. 17

18 To alleviate the concern that the distribution of Delay is bounded between 0 and 1 may affect the results, we replace raw Delay with the logistic transformation of Delay and repeat the analysis in Table 5. We find similar results. Overall, the results in Table 5 demonstrate that Delay decreases after Phase implementation. This result is consistent with the intuition that bond prices more quickly incorporate market information when market transparency increases. 6.3 Variance ratio Table 6 shows the difference-in-difference regression results for Variance ratio. The coefficients on the interaction term are negative and statistically significant for all three Phases: with a t-statistic of for Phase 2, with a t-statistic of for Phase 3A, and with a t-statistic of for Phase 3B. The economics significance of these estimates is substantial for all three phases. The coefficient for Phase 2 represents a 22.2% reduction relative to the average Variance ratio of Phase 2 treated bonds before Phase (0.41). The coefficient for Phase 3A means a 13.4% decrease compared with the average Variance ratio of Phase 3A treated bonds before Phase (0.35). The coefficient for Phase 3B is equal to a -21.8% change of the average Variance ratio of Phase 3B treated bonds before Phase (0.34). These results in Table 6 show that the Variance ratio decreases after TRACE Phase, meaning that bond prices more closely approximate a random walk when market transparency increases. 6.4 The number of bond analyst reports As we summarize in Section 2, theories suggest that the improvement of market transparency can decrease market participants incentives for producing information (Asriyan et al., 2015; Bhattacharya, 2016; Banerjee et al., 2016). We consider the number of bond analyst reports for a bond. According to Bond Market Association (BMA, 2004), fixed income research analysts play a critical role in informing the market about particular issues or securities. These bond analyst reports are essential in promoting efficiency in bond price discovery. Therefore, when there are more bond analyst reports about a specific bond, information production by bond analysts about this bond is more intense. Table 7 reports difference-in-difference regression estimates for the number of analyst reports. We use the raw number of analyst reports as the dependent variable in the differencein-difference regressions, and standard errors are clustered at the firm level. The coefficients on T reated P ost are -0.43, -0.46, and for Phase 2, Phase 3A, and Phase 3B, respectively. The coefficients are statistically significant at 10% critical level for Phase 2 and 3A 18

19 and significant at 1% critical level for Phase 3B. Hence, when the market becomes more transparent, the number of bond analyst reports decreases. The economic magnitudes of these estimates are large. For instance, the coefficient on T reated P ost of Phase 2 in Panel A is 31.9% of the mean of the number of analyst report (1.35) for Phase 2 treated bonds before the Phase implementation. The coefficients on T reated P ost of Phase 3A (0.46) and 3B (0.99) are 28.2% and 46.5% of the corresponding average numbers of analyst reports for treated bonds before the Phase implementation. In summary, results in Table 7 suggests that, after Phase implementation, there are fewer bond analyst reports for the treated bonds relative to the control bonds. These findings are consistent with the interpretation that greater market transparency leads to less intense information production. 6.5 Cross-sectional variation by ex-ante bond characteristics We study whether the difference-in-difference estimates differ between different groups of treated bonds grouped by ex-ante bond characteristics, including illiquidity, trading activity, and maturity. We assign treated bonds into groups according to the values of these variables before Phase implementation. Amihud is the measure of illiquidity, and the ratio of volume divided by issue amount is the measure of trading activity. For illiquidity and trading activity groups, we calculate medians of bond characteristics across treated bonds before Phase implementation and consider treated bonds above the median and those below the median. For maturity, we form groups of treated bonds based on whether the bond s maturity is longer than five years. For drift after bond analyst report and drift after credit rating change, we measure illiquidity and trading activity using the average of daily illiquidity and trading activity during the week before the bond analyst report or the credit rating change before the Phase implementation. For Delay and Variance ratio, we measure illiquidity and trading activity using the average of daily illiquidity and trading activity one year before the Phase implementation. We then estimate a regression as follows. y i,t =b 0 + b 1 T reated i + b 2 P ost t + b 3 HighGroup i + b 4 T reated i P ost t + b 3 T reated i P ost t HighGroup i + ɛ i,t, (5) where the dependent variable y i,t is a measure of pricing efficiency for bond i in week t, T reated i is an indicator for whether the bond changes its dissemination status and P ost t is a dummy variable that equals to one in the period after the dissemination starts and zero otherwise. HighGroup i is a dummy variable that is one for treated bonds with high 19