The Real Effect of Financial Disclosure: International Evidence

Transcription

1 The Real Effect of Financial Disclosure: International Evidence Presented by Dr Xi Li Associate Professor of Accounting London School of Economics and Political Science #2016/17-11 The views and opinions expressed in this working paper are those of the author(s) and not necessarily those of the School of Accountancy, Singapore Management University.

2 The Real Effect of Financial Disclosure: International Evidence Xi Li London School of Economics 44 (0) Xuan Tian Kelley School of Business Indiana University (812) Fariborz Moshirian Australian School of Business UNSW Australia 61(02) Bohui Zhang Australian School of Business UNSW Australia 61(02) Current Version: August 2016 * We thank Zhengyuan Wang for his early contribution to this paper. We also thank comments from Zachary Kaplan (discussant) and participants at the 2016 FARS mid-year meeting and 2016 AAA annual meeting. We remain responsible for all errors and omissions.

3 The Real Effect of Financial Disclosure: International Evidence Abstract Rajan and Zingales (1998, 2003) argue that good accounting standards and disclosure rules reduce the wedge between the cost of internal and external funds and enhance growth. We test the causal link between financial reporting and growth using a quasi-natural experiment the mandatory adoption of International Financial Reporting Standards (IFRS) across the world - and examine its effect on innovation, a corporate activity that directly drives economic growth. Our Difference-in-Differences (DiD) results suggest that improved financial reporting leads to more innovation in the long run it generates more patents and patents with higher impact. We also find that the positive effect of improved financial reporting on innovation is more pronounced among industries with higher dependence on external financing, consistent with the role of good financial reporting in reducing the cost of external financing. In addition, we find results consistent with the managerial learning hypothesis that managers are able to learn from the stock market after improved financial transparency. Our paper sheds new light on the real effects of financial reporting. JEL Classifications: G15; G19; M41; M42 Keywords: Financial reporting; IFRS; Innovation; Patents; Citations

4 1. Introduction Innovation is vital to a firm s business success and a country s long-term economic growth. From the perspective of the accounting literature, it is intriguing to understand the role of financial disclosure on corporate innovation. Overwhelmingly, the recent literature finds that firm innovation is impeded by several disclosure mechanisms such as financial analysts (He and Tian, 2013), media coverage (Dai, Shen, and Zhang, 2015), accounting conservatism (Chang et al., 2016), and mutual fund holding disclosure (Agarwal, Vashishtha, and Venkatachalam, 2016). This dark side evidence raises a fundamental concern about whether transparency is a recipe for innovation. To answer this important research question, we take advantage of the mandatory adoption of International Financial Reporting Standards (IFRS) as a quasi-natural experiment that provides plausibly exogenous variation in financial reporting, and explore its effect on corporate innovation around the world. The European Commission (EC) Regulation No. 1606/2002 requires that all firms which have traded in major stock exchanges of the European Union (EU) adopt IFRS from 2005 onwards. Australia and South Africa also mandated IFRS adoption in 2005, followed by New Zealand in 2007, and Israel and Turkey in This event represents one of the largest changes in the history of financial reporting, and the IFRS adoption has substantially improved firms accounting practices and transparency. Financial reporting can promote corporate innovation through two channels. First, motivating innovation is a challenge for most firms. According to the World Bank Enterprise Surveys ( ), almost 40% of firms cite insufficient access to finance as the foremost obstacle to their operations and growth. Innovation is a long-term process that tends to exhaust internal capital and entails uncertainty, which hinders effective communication with outside investors (Bhattacharya and Ritter, 1983). Innovative firms thus suffer more severely from limited external financing. Rajan and Zingales (1998, 2003) argue that an accounting and 1

5 disclosure system that promotes transparency is essential to the development of a country s financial system, because it directly affects the ability of firms to raise external funds and thus affects capital-allocation efficiency in the economy. Therefore, we expect enhanced financial reporting to increase a firm s innovation output via this financial constraint channel. Second, financial reporting can affect corporate innovation through facilitating the feedback effects of outsiders. Allen and Gale (1999) state that innovative projects are usually difficult to evaluate, because information about their prospects is either sparse or hard to process. Financial reporting discloses firms investment information to the public and allows outsiders to judge the prospects of firms investment opportunities, which, in turn, affects firm managers real investment decisions. For example, Loureiro and Taboada (2015) document that the improvement in financial reporting increases insiders ability to learn from outsiders measured by investment-to-price sensitivity and the relation between the market reaction to merger and acquisition (M&A) deal announcements and the likelihood of deal completion. In addition, uniformed financial reporting increases comparability across firms and allows managers to learn from their peers and make better investment decisions. Thus, we expect that the feedback role of financial reporting can improve firms investment efficiency in innovation. We term this mechanism the feedback channel. An alternative hypothesis predicts the opposite effect. A high level of information transparency and comparability could create high capital-market pressure and thus lead to managerial myopia aiming to improve short-term profitability, which, in turn, could be detrimental to firms long-term value (e.g. Stein, 1989; Bhojraj and Libby, 2005). For example, Graham, Harvey, and Rajgopal (2005) survey 401 chief financial officers (CFOs) in the U.S. and find that the majority of CFOs are willing to sacrifice long-term value, such as R&D spending, for short-term performance because they are pressured to meet short-term targets. He and Tian (2013) find that financial analysts impose short-term pressures on managers and 2

6 reduce market s tolerance for failure. Therefore, financial reporting and transparency can also impede firm innovation. To test these two competing views, we focus on the changes in firms innovation output before and after the mandatory adoption to evaluate the real effect of financial reporting). We employ a difference-in-differences (DiD) approach by using mandatory IFRS adopters as the treatment group and two alternative control groups. The first control group includes non-ifrs adopters from countries that did not mandate IFRS adoption during our sample period, i.e The second control group includes firms that voluntarily adopted IFRS before their countries mandatory adoption date. We collect innovation data from the Orbis database for individual firms, and build two proxies to capture a firm s innovation output patent count and patent citation count following the innovation literature (e.g. Hall, Jaffe, and Trajtenberg, 2001; Hsu, Tian, and Xu, 2014). Our baseline analysis shows that mandatory IFRS adoption is positively correlated with corporate innovation. Mandatory IFRS adopters generate 8.7% more patents and their patents receive 8.5% more citations, compared to non-adopters from non-ifrs adopting countries after the mandatory adoption. We also find that mandatory adopters generate 3.3% more patents and their patents receive 1.5% more citations, compared to voluntary adopters domiciled in the same countries after the mandatory adoption date. To further examine the timing of the change, we use a dynamic DiD approach. The result indicates that there are no significant differences in innovation performance between mandatory IFRS-adopters and non- IFRS adopters or voluntary adopters prior to the mandatory adoption dates. IFRS-adopters start to exhibit better innovation performance only after the adoption and incrementally more innovation two years after the adoption, consistent with the view that it takes time for firms to make strategic adjustments to increase their innovation output. 3

7 Next, we explore the financial constraint channel through which mandatory IFRS adoption may facilitate innovation. Innovative firms are more likely to be subject to internal capital constraints (Brown, Fazzari, and Petersen; 2009) and thus rely more on external finance. Prior literature documents reduced costs of capital and improved cross-border financing following mandatory IFRS adoption (Li, 2010; Chen, Ng, and Tsang, 2014), which could potentially improve a firm s ability to innovate, especially for financially constrained firms. Building on an exogenous measure of dependence on external finance, we find evidence suggesting that country-level transparency score is positively associated with firm innovation among industries with high dependence on external finance in pre-ifrs adoption period. After the mandatory IFRS adoption, firms from external-finance-dependent industries experience a disproportionally higher improvement of innovation output. Lastly, we explore the feedback channel through which mandatory IFRS adoption may encourage more innovation. We first find evidence consistent with the argument that IFRS adoption increases insiders ability to learn from capital market, measured as an average increase in investment-to-price sensitivity in our sample of mandatory adopters after the adoption. We then document evidence suggesting that the positive effect of IFRS adoption on innovation is more pronounced among firms from industries experiencing improvement in managerial learning. This paper contributes to the finance and accounting literature in the following ways. First, it contributes to the literature examining the real effect of financial reporting. There is a large body of literature in accounting examining the role of accounting in capital markets (see Kothari (2001) for a review). However, very little research has been devoted to examining the real effect of accounting. Kedia and Phillipon (2009) examine the real economic consequences of fraudulent accounting practices and find that earnings management distorts resource allocation, as misreporting firms hire and invest excessively and shrink quickly after the fraud 4

8 is detected. Our paper complements theirs by showing the bright side of improved financial reporting its encouragement of corporate innovation, which plays a critical role in sustaining economic growth. Our paper also contributes to the rapidly growing body of literature on motivating technological innovation, especially in a global market. Holmstrom (1989) points out that innovation activities may mix poorly with routine activities in an organization. Manso (2011) theoretically discussed several mechanisms to motivate innovation. There are a range of factors identified by the literature exhibiting positive or negative effects on corporate innovation, including the timing of financial markets (Nanda and Rhodes-Kropf, 2013), laws (Acharya and Subramanian, 2009; Brown, Martinsson, and Petersen, 2013), financial market development (Hsu et al., 2014), firm boundaries (Seru, 2014), stock liquidity (Fang et al., 2014), analyst coverage (He and Tian, 2013), banking competition (Cornaggia et al., 2015), product market competition (Aghion et al., 2005), and institutional investors (Aghion et al., 2013; Chemmanur et al., 2014). To the best of our knowledge, our paper is among the first to take the perspective of financial reporting as a motivation for corporate innovation. This paper also contributes to studies examining institutional features as determinants for innovation. Brown et al. (2013) use accounting standards as a proxy for a firm s ease of access to external financing, as Rajan and Zingales (1998) argue that a country s accounting system facilitates external equity financing. However, as Brown et al. (2013) observe and point out in earlier literature also, a country s accounting system is largely static and endogenously determined by other institutional features, such as legal origin and market demand. Using an exogenous change in a country s accounting system enables us to draw the casual link between transparency and innovation. This paper is related to the literature examining IFRS adoption on investment efficiency. Using reported accounting numbers, such as capital expenditures, R&D expenditures, earnings 5

9 and cash flows, Schleicher, Tahoun, and Walker (2010), Shroff, Verdi, and Yu (2014), and Chen, Young, and Zhuang (2013) find that mandatory IFRS adoption leads to a higher level of investment efficiency. They attribute their findings to better external information environments and better comparability among peers that leads to better monitoring and lower agency problems. Our paper differs from these in several ways. First, we do not rely on financial statement information to measure the output. The IFRS has drastically changed firms financial reporting, including the measurement and recognition of earnings, investments, and R&D expenditures. The differential results documented by prior studies using reported accounting numbers to measure investment efficiency might be confounded by the changes in reporting rules. Second, Koh and Reeb (2015) find that many innovative US firms strategically avoid reporting R&D expenditures in their financial statements. Studies that treat firms without R&D expenditures reported in financial statements as having zero R&D significantly underestimate firms innovation activities. Interestingly, they find that after forced auditor change, firms that did not report R&D before start to report substantial amounts of R&D expenditure, suggesting that not-reporting R&D is a discretionary reporting choice rather than a lack of innovative activities. Considering the consistent reporting standards on R&D and strong enforcement in the US, our sample of international firms is even more likely to be subject to such reporting discretion. Therefore, results relying on reported R&D expenditures as the dependent variable are confounded by the concern that firms strategically disclose more R&D under the IFRS regime. 1 Third, innovation output is fundamentally different from investment efficiency. Firms could have high investment efficiency but lack innovation, and vice versa. The rest of the paper proceeds as follows. Section 2 introduces the database of innovation, explains the proxies for innovation and describes the sample selection procedure. 1 The reason why a firm wants to record more R&D expenditure under IFRS could be to take extra expenditure during the current period in order to create a buffer for future earnings management. 6

10 Section 3 presents model specifications and reports empirical findings of the baseline and dynamic models. In Sections 4 and 5, we discuss the financial constraint channel and feedback channel through which mandatory IFRS affects corporate innovation, and provide empirical evidence. Finally, Section 6 concludes. 2. Data and Sample Selection In this section, we introduce the global patent database used in this study, and explain the construction of key innovation variables. We also describe the sample selection process and present summary statistics Patent Database We use the Bureau Van Dijk s Orbis patent database to construct our innovation variables. This database is sourced from the European Patent Office s (EPO) Worldwide Patent Statistical Database (PATSTAT). Similar to the USPTO, the EPO is one of the largest and most important patent offices in the world. The Orbis patent database offers a comprehensive coverage of more than 88 million patent applications worldwide since These patents are filed by various types of entities, including publicly-traded and privately-held firms, individuals, governments and universities through 94 regional, national and international patent offices. Because of its worldwide coverage, the Orbis patent database is more suitable for international studies. Many prior innovation studies, such as Hall, Jaffe, and Trajtenberg (2005) and Aghion, Van Reenen, and Zingales (2013) among many others, are based on a single country, in most cases the US, and thus relies on the NBER patent database compiled from the United States Patent and Trademark Office (USPTO). Although the NBER patent database is 3 This number is by October Out of 88 million patents, 37.3 millions are granted patents. 7

11 an excellent source of patents filed in the US, its exclusion of international patents from other patent offices is an obvious limitation for international studies. An exception is Hsu, Tian, and Xu (2014) who use the distribution of US patents filed by foreign firms to estimate innovation activities in corresponding countries. However, observed foreign patenting activities may be very different from domestic innovation. For example, Goto and Motohashi (2007) compare data from the USPTO and the Japan Patent Office (JPO), and find that the distribution of USPTO patents filed by Japanese firms is quite different from the distribution of domestic Japanese patents filed in the JPO. In addition, inconsistent administrative procedures across patent offices may contaminate the data and thus proper adjustments may be necessary (Webb et al., 2005). Using the Orbis patent database can help us overcome these shortcomings of the NBER patent database and enable us to more accurately identify innovation at the firm-level Innovation Measures Following the innovation literature, we construct two measures for innovation: total number of granted patents and the total number of citations. The first measure is the total number of patents granted to each firm in every year. This variable captures the output side of innovation instead of the traditional input side of innovation such as R&D expense used in prior literature (e.g. Brown, Martinsson, and Petersen 2013). The availability of patent count is better than R&D expenditure, because the latter is generally not reported consistently during our sample period by non-us firms. We use a patent s application year to match other financial data because it usually takes years before a patent is eventually granted. We only aggregate patents for those without priority numbers to prevent overestimating patent count. The priority number is commonly used in the international patent system as the number of the application in respect to which priority is claimed, i.e., it is the same as the application number of the claimed priority document (Orbis manual). A simple example illustrates the functionality of priority number. For example, the Japanese car manufacturer 8

12 Toyota generates an invention and applies for a patent to the JPO. Several months later, Toyota applies for a patent to the USPTO for the same invention expecting to have protection in the US. This subsequent USPTO patent is associated with a priority number, which is the application number of the prior JPO patent, indicating that the same invention has been applied for a patent before. So this USPTO patent is not considered as a novel new invention, and only one patent is counted for Toyota. The raw patent count is subject to a truncation problem as shown by Hall, Jaffe, and Trajtenberg (2001, 2005). Due to the application-grant lag, many patents may not have been granted if they were applied for in the last several years of database coverage. Our download of the Orbis database is up to July 2014, so we follow Hall, Jaffe, and Trajtenberg s (2001, 2005) method to adjust the raw patent count in the last few years of database coverage ( ) using the application-grant lag distribution of Specifically, we define the application-grant lag distribution (WW ss ), as the percentage of patents applied for in a given year that are granted in s years. For the truncation-adjusted patent count (PP aaaaaa ), we compute PP aaaaaa = PP rrrrrr 2014 tt ss=0 WW ss, where PP rrrrrr is the raw patent count at year t and 2009 t After the adjustment of truncation, we need to transform the value. As the patent count is a discrete variable and highly right-skewed with a large number of zero patent observations in the sample, we use the logarithm of one plus the discrete patent count as the dependent variable in the regression analysis following the innovation literature such as Atanassov (2013) and Hsu, Tian, and Xu (2014). The second innovation measure we build is the number of citations received by patents. This measure captures the quality of innovation (Hall, Jaffe, and Trajtenberg, 2001) because a patent is very likely to be of great technological importance if it receives a high volume of citations from future patents. The raw citation count is subject to the truncation problem as well (Hall, Jaffe, and Trajtenberg, 2001) because a patent may keep receiving future citations after 9

13 the end of the database coverage period. Another issue artificially distorting the raw citation count is the inconsistency of administration procedures implemented by different patent offices. Webb et al. (2005) and Goto and Motohashi (2007) document that citations received by USPTO patents are significantly higher than citations received by EPO or JPO patents. They find the reason for this discrepancy is due to different administrative procedures among these patent offices. As the USPTO imposes a legal requirement on applicants to supply a complete list of citations at the time of application, applicants are likely to provide more than the necessary citations in order to avoid any punishments. However, USPTO patent examiners may not have enough time to verify every citation (Webb et al., 2005). In contrast, the EPO does not impose any similar requirements. In the EPO, it is the patent examiners duty to determine appropriate citations. The EPO follows a parsimonious philosophy by including only the most relevant and important citations. The JPO s policy has changed several times, and its current policy is a mixture of the USPTO and the EPO. Additionally, the JPO s citation system contains some other unique features, described by Goto and Motohashi (2007). To reduce noise in the raw data, we choose a fixed effect adjustment approach suggested in Hall, Jaffe, and Trajtenberg (2001). Specifically for the raw citation count of each patent, we divide the raw value by the average raw number of citations across all patents from the same patent offices and in the same year. This approach helps reduce year and patent office effects. After making this adjustment, we then aggregate the total number of citations received by patents for each firm in every year. As with patent counts, citation counts are a discrete variable, also. As a result, we use the logarithm of one plus patent counts (citation counts) in our regressions. Although using patenting activities to measure innovation has been widely used, it is important to note that this type of measure has its own limitations. For example, not all kinds of innovation are well captured by patents. As filing a patent requires public disclosure of technological details, some firms may choose to keep their inventions secret instead of seeking 10

14 protection by filing patents. Other conceptual innovation or operational optimization type innovations are not eligible to file patents under current regulations. There are other formats of intellectual property protection such as trademarks or copyright. Firms in different industries may choose different ways to materialise their innovation. Furthermore, patents only reflect successful innovative activities, leaving unsuccessful innovative attempts unobserved. Nevertheless, there are no other widely accepted innovation measures yet (Acharya and Subramanian, 2009). Despite these imperfections, patenting activities still reflect very important technological innovations that are available for public access, and can be quantitatively measured. We carefully design additional controls which help strengthen the credibility of the findings Sample Selection We employ firms from countries with mandatory IFRS adoption during our sample period, i.e , as a treatment group. Since the majority of our sample countries adopted IFRS in 2005, in order to analyse firms innovation activities during pre- and post-ifrs mandatory adoption periods, our sample period is a nine-year ( ) window centred around the mandatory adoption year of This length of sample period is appropriate, because a longer period may introduce unnecessary noise due to other events, while a shorter period may not sufficiently reflect the changes of innovation given its nature as a long-term activity. For the treatment sample, we only keep firms that switched to IFRS for the first time at the mandatory adoption date, following the IFRS literature (such as Li, 2010; Christensen, Hail, and Leuz, 2013). These firms are referred to as mandatory IFRS adopters. Other firms which are domiciled in IFRS-adopting countries but voluntarily adopted IFRS before their countries adoption date are excluded from the treatment sample but instead used as a control group. We use non-ifrs adopters from non-ifrs adopting countries during our sample period 11

15 as another control group. Similarly, we remove firms that voluntarily adopted IFRS from this group. We obtain firm-year financial data from the Worldscope and Compustat Global databases and require every firm to have at least two years data (one before and one after the mandatory adoption date for the mandatory adopters and voluntary adopters and two observations during the whole sample period for non-adopters). We impose a restriction on the eligible sample countries to have a minimum of two firms and one patent record during our sample period. We use the two-digit SIC code to classify industries and exclude financial industries (SIC 60-69) and utility industries (SIC code 49) from the sample. Finally, our regression analysis controls for macroeconomic variables obtained from the World Bank. As a result, countries (regions) lacking macroeconomic data are not included. After imposing the above conditions, our final treatment sample consists of 3,217 mandatory adopters from 24 IFRS-adopting countries. Our first control sample consists of 20,740 non-adopters from 14 non-ifrs countries. Our second control sample consists of 916 voluntary adopters from 23 IFRS-adopting countries (Spain has been dropped). Table 1 reports sample distribution. Within the treatment group, Australia and the UK are the two largest countries with 914 and 455 firms in our sample, respectively. Within the non-ifrs country group, the US with 5,757 firms is ranked first, followed by Japan with 4,218 firms. Table 2 reports sample distribution by calendar year of financial information. Both the treatment and control samples are evenly distributed during our sample period. 3. Research Design and Empirical Results 3.1. Baseline Model Specification We employ a DiD approach in the baseline analysis. Specifically, we estimate the following model: 12

16 Log(1+Innovationi,t+2 )= α + β1 Mandatoryi Post_IFRSi,t + β2 Post_IFRSi,t+ Controls (1) where Innovationi,t+2 is one of the two innovation measures constructed in the previous section for firm i in year t+2. As discussed above, we use the patent application year to determine the timing of innovation. Because it takes time to observe innovation output and the effect of mandatory adoption on firm innovation could be lagged, we approximate the gap between IFRS and patent application by a two-year lag in our main analysis. The two-year gap is comparable to most other existing innovation studies such as Tian and Wang (2014), Atanassov (2013), and Hsu, Tian, and Xu (2014). Mandatoryi is a binary variable that equals one for mandatory adopters i from mandatory IFRS adoption countries and zero otherwise. Post_IFRSi,t is a binary variable that equals one for fiscal years ending on or after the country s mandatory adoption date (listed in Table 1) and zero otherwise. It is set as zero for firms from non-ifrs adopting countries. We include a set of firm-, industry-, and country-level time-varying control variables that may affect innovation in the regressions. Following Atanassov (2013), Hsu, Tian, and Xu (2014) and Brown, Martinsson, and Petersen (2013), we use total assets, market-to-book ratio, leverage, return on assets (ROA), tangible assets scaled by total assets, capital expenditures scaled by total assets, percentage of insider holdings, percentage of institutional ownership, and the number of analysts following as firm-level control variables. At the industry-level, we control for industry concentration measured by the sales Herfindahl index for each two-digit SIC industry in every country. We include its squared term to control for any non-linear relationship between industry concentration and innovation (Aghion et al., 2005; Atanassov, 2013). At the country-level, we control for the logarithm of Gross Domestic Production (GDP) per capita and total stock market capitalisation scaled by GDP as proxies for country and capital market development. Our model includes firm fixed effect to alleviate the concern of timeinvariant omitted variables such as unobservable firm characteristics that may drive our results. 13

17 The firm-level variable Mandatoryi is subsumed by firm fixed effects. Year fixed effect is controlled to absorb external annual shocks such as the global financial crisis in Our main variable of interest is the coefficient estimate of the interaction term ββ 1. If ββ 1 is positively significant, it means that there is a significant improvement in innovation for mandatory IFRS adopters during the post-ifrs adoption period compared to the control group. We cluster standard errors at the country level to address the correlations among residuals within the same country Baseline Result Table 3 reports the summary statistics for variables used in our regressions separately for the treatment group, i.e. mandatory adopters from IFRS-adopting countries, and two control groups, i.e. non-adopters from non-ifrs adopting countries and voluntary adopters from IFRSadopting countries. Table 4 reports the result of the baseline analysis. In the first column, the coefficient estimate of Mandatory Post_IFRS is and is significant at the 1% level. The magnitude of the coefficient estimate suggests that during the post-ifrs adoption period, mandatory IFRS adopters generate about 8.7% more patents, on average, than non-ifrs adopters. This observation is not only statistically significant, but also economically sizeable. In the second column when citation count is the dependent variable, the coefficient estimate of the interaction term is and is significant at the 1% level. It means that mandatory IFRS adopters receive about 8.5% more citations, on average, than non-ifrs adopters during the post-ifrs period. In columns (3) and (4), we exclude the US from the control sample and get similar results. In columns (5) and (6), we use voluntary adopters from IFRS-adopting countries as the control group. These firms are from the sample countries as the mandatory adopters, mitigating the concern that the results are driven by different country institutions across IFRS-adopting and non-adopting countries. We continue to find positive and significant coefficients on Mandatory 14

18 Post_IFRS, suggesting that, relative to voluntary adopters who adopted IFRS earlier, mandatory adopters generate more patents and citations after the mandatory adoption date. The coefficient estimate on Post_IFRS is negative and largely insignificant, suggesting that innovation for voluntary adopters do not change after their countries mandatory adoption date. Across all columns, the adjusted R-squared values are reasonably high indicating a good fit for our specification. Overall, our baseline result empirically shows that mandatory IFRS adoption facilitates innovation Dynamics A critical assumption of the DiD approach is the parallel trend assumption in the treatment and control groups before the event. Therefore, we introduce a dynamic model to verify this assumption by investigating whether there was an existing difference in innovation between mandatory IFRS adopters and non-ifrs adopters before In the dynamic model, we repeat Equation (1) by replacing the Post_IFRSi,t indicator with three separate event window indicators, including the two years leading up to the adoption (Pre_IFRSi,t-2,t-1), the first two years after the adoption (Post_IFRSi,t,t+1), and the remaining years (Post_IFRSi,>=t+2). Pre_IFRSi,t-2,t-1 is thus defined as being one for observations from the IFRS countries and with fiscal years ending on or after Month -24 (relative to the IFRS adoption date) and before Month 0, and zero otherwise. Post_IFRSi,t,t+1 is defined as being one for observations from the IFRS countries and with fiscal years ending on or after Month 0 and before Month +24, and zero otherwise. Post_IFRSi,>=t+2 is defined as being one for observations from the IFRS countries and with fiscal years ending on or after Month +24 and zero otherwise. Other control variables and fixed effects are the same as those in the baseline regressions. Table 5 reports the results of the dynamic model. In the first column, we find that the coefficient estimate of Mandatoryi Pre_IFRSi,t-2,t-1 is not significant. Therefore, there is no evidence showing that there is an existing difference in innovation between the treatment and 15

19 control group before the IFRS adoption. The assumption of DiD is unlikely to be violated. We also observe that throughout the sub-periods after the IFRS adoption, the coefficient of Mandatoryi Post_IFRSi,>=t-1 is more positive and significant than the coefficient of Mandatoryi Post_IFRSi,t,t+1. This observation suggests that mandatory IFRS adopters become incrementally more innovative compared to control firms about two years after the adoption. This finding is consistent with our conjecture that it takes time for firms to adjust their operations in response to policy changes. It is also consistent with the long-term nature of innovation discussed before. 4. Financial Constraint Channel As Rajan and Zingales (1998, 2003) argued, the accounting system affects a firm s ability to raise external funds, which could affect their innovative activities. According to the existing accounting literature, IFRS improves a firm s external financing ability for at least two reasons. Firstly, IFRS enhances financial disclosure because IFRS generally requires more disclosure than local accounting rules (Ashbaugh and Pincus, 2001). Enhanced disclosure helps lower the cost of capital by mitigating information asymmetry (Diamond and Verrecchia, 1991) or by lowering systematic risk (Barry and Brown, 1985; Lambert, Leuz, and Verrecchia, 2007). Secondly, adoption of IFRS increases cross-border comparability of financial information (DeFond et al., 2011; Yip and Young; 2012). Chan, Covrig, and Ng (2005) and Covrig, DeFond, and Hung (2007) argue that the cost of acquiring and processing information, particularly faced by foreign investors, is reduced as a consequence of IFRS. De Franco, Kothari, and Verdi (2011) find that analyst following and the accuracy of forecasts are positively correlated with accounting information comparability. Therefore, it is reasonable to expect a drop in the cost of capital when the information cost is lowered. Indeed, Daske et al. (2008) and Li (2010) provide evidence that the cost of capital is reduced following mandatory IFRS adoption. 16

20 Based on above findings, we hypothesize that improved ability to obtain external financing is one underlying channel through which mandatory IFRS adoption facilitates innovation. Brown, Fazzari, and Petersen (2009) argue that innovative firms are more likely to be subject to capital constraints thus relying on external financing. Brown, Martinsson, and Petersen (2012), Brown, Martinsson, and Petersen (2013) and Hsu, Tian, and Xu (2014) all empirically document that increased availability of external finance promotes corporate innovation. We conduct two analyses to test this hypothesis. First, we examine whether innovation is positively associated with the transparency of a country s accounting system among industries with high dependence on external finance during the pre-ifrs adoption period. To measure a country s transparency in the pre-ifrs adoption period, we use country-level opacity scores constructed by Leuz, Nanda, and Wysocki (2003) and subsequently updated by Leuz (2010) using data from To match this sample, we re-construct our sample period to We exclude the year 2005 to make sure our sample firms are not affected by the IFRS adoption in We follow Rajan and Zingales (1998) to measure the dependence on external finance as the portion of capital expenditures not internally generated, i.e., capital expenditures minus cash flow from operations divided by capital expenditures. 4 We first obtain information on all US firms with non-missing annual data on capital expenditures and cash flows from operations as covered by Compustat North America during the same sample period, i.e To smooth temporal fluctuations, we then sum each firm s capital expenditure minus cash flow from operations over the sample period and then divide it by the sum of capital expenditures 4 Rajan and Zingales (2003) s approach has been widely used in the literature (such as Gupta and Yuan, 2009; Brown, Martinsson, and Petersen, 2013; Hsu, Tian, and Xu, 2014). Brown, Martinsson, and Petersen (2013) propose a modified version of dependence on external finance by dividing the amount of capital expenditure not funded by operational cash flow over the total amount of capital expenditure plus R&D costs. However, due to the concerns about R&D disclosure as pointed out by Koh and Reeb (2015), we do not use the Brown et al. measure. 17

21 over the sample period. This procedure creates the firm-level external finance dependence measure. Lastly, we create the industry-level external finance dependence measure by taking the industry medians at the two-digit SIC level. Although this measure is calculated using US data, we apply it to the same industry across all countries, as Rajan and Zingales (1998) argued that dependence on external finance is an inherent firm characteristic mainly driven by technological reasons and thus is unlikely to be affected by a country s accounting system or to vary across countries. We follow Rajan and Zingales (1998) and conduct the analysis at the country-industry-year level using the model below: Log(1+Innovationj,k,t+2 )= α + β1 Transparencyk External Dependencej + Controls (2) where Innovationj,k,t+2 is the total number of patents or citations in industry j of country k in year t+2. Transparencyk is a time-invariant country-level opacity score for country k obtained from Leuz (2010). We multiply the score by -1, so that a higher score indicates a higher level of transparency. External Dependencej the time-invariant external dependence ratio for industry j as described above. We include various industry- and country-level controls and additionally industry, country, and year fixed in this regression. Therefore, the main effects of Transparencyk and External Dependencej are subsumed. We expect the coefficient estimate ββ 1 to be positive, as industries with higher dependence on external finance are more innovative in countries with more transparent accounting systems. We follow a similar sample selection approach as described in Section 2.3. We exclude US firms from this analysis as US firms are used to calculate the external financing dependence variable. The results are reported in Table 6, Panel A. Consistent with our prediction that transparency leads to higher innovation among industries with high dependence on external finance, we find the coefficient on the interaction term Transparencyk External Dependencej is positive and significant in all models. However, a potential concern with the above approach is that a country s financial reporting system rarely changes over time and a cross-sectional association may be subject to 18

22 an endogeneity concern. To address this concern, our second approach takes advantage of the change in a country s accounting standards, i.e. IFRS adoption, to mitigate this concern. If mandatory IFRS adoption facilitates innovation through a firm s ability to obtain external finance, then firms which are more dependent on external finance should experience a disproportionally higher increase in innovation following mandatory IFRS adoption. Similar to above, we use data from US firms to measure dependence on external finance. Now we use data from to map our sample period of IFRS adoption analysis. We use the median value of external financing for each industry across our sample period. Industries which fall into the top 50% of the sample are classified as having high external financing dependence and the bottom 50% are classified as having low external financing dependence. We then split our sample into industries with high and low external financing dependence. We report the result in Table 6, Panel B. We find that the coefficients on Mandatory Post_IFRS have a larger magnitude in the subsample with high external finance dependence than those in the subsample with low external finance dependence in all specifications. The differences in the coefficients across the high and low subsamples are statistically significant in three out of four specifications. This evidence supports our hypothesis that financial constraint is a channel through which mandatory IFRS adoption motivates corporate innovation. 5. Feedback Channel Another channel through which mandatory IFRS adoption could affect innovation is managerial learning. Chen, Goldstein, and Jiang (2007) find that stock prices provide an informational feedback effect and managers can learn from the information contained in stock prices. Loureiro and Taboada (2015) find that mandatory IFRS adoption has increased managers ability to learn from stock markets. In addition, uniformed financial reporting also increases comparability across firms and allows managers to learn from their peers and make 19

23 better investment decisions. Thus, we expect that the feedback role of mandatory IFRS adoption encourages innovation by stimulating managerial learning. We measure managerial learning using the investment-to-price sensitivity. We first replicate the analysis in Loureiro and Taboada (2015) in our sample of firms and find similar results IFRS adoption leads to a higher level of investment-to-price sensitivity. The results are reported in Table A.1. If the managerial learning channel is able to explain our main findings, we expect the effect of IFRS adoption on innovation to be stronger among industries experiencing a larger improvement in managerial learning. To test this conjecture, we first need to capture the effect of IFRS on managerial learning for each industry. We estimate the following model for each industry (two-digit SIC) within each IFRS-adopting country: Capexi,t/PPEi,t-1 = α + β1 Qi,t-1 Post_IFRS + β2 Qi,t-1 + β3 Log(Assetsi,t-1) + β4 Cashflowi,t/Assetsi,t-1 + Controls (3) where Qi,t-1 is Tobin s Q for firm i at year t-1, measured as market value of equity plus total assets minus book value of equity scaled by total assets. Capexi,t/PPEi,t-1 is capital expenditures scaled by lagged net PP&E for firm i at year t. Cashflowi,t/Assetsi,t-1 is net income plus depreciation and amortization scaled by lagged total assets. Post_IFRS is defined as the same as before. The coefficient estimate β1 thus captures the incremental investment-to-price sensitivity after IFRS adoption, i.e. the extent to which managers learn from stock prices. We require each industry to have at least 20 observations to estimate the coefficient. We identify industries with β1 above the sample median as those experiencing greater improvement in manager learning. We thus modify Equation (1) by adding an interaction term Post_IFRS Mandatory Learn to capture the incremental IFRS adoption effect on innovation among industries with improvement in managerial learning. If our conjecture is supported, we expect to observe a positive and significant coefficient estimate. The results are reported in Table 7. Consistent with the manager learning channel, we observe a positive coefficient on Post_IFRS Mandatory Learn in most of the models. 20

24 6. Conclusion In this paper, we examine the real effect of financial reporting. Using firm-level innovation data for a large number of mandatory IFRS adopters from 24 countries as the treatment and voluntary adopters from the same countries as well as non-adopters from 14 non- IFRS adoption countries as two control groups, we document a positive effect of mandatory IFRS adoption on corporate innovation. Our difference-in-differences results suggest that financial reporting has positive real effects: firms that adopted IFRS become more innovative in the long run they generate more patents and patents with higher impact. We also establish the financial constraint and feedback channels through which improvements in financial reporting facilitate corporate innovation. Our paper complements the findings in Rajan and Zingales (1998, 2003) and sheds new light on the real effects of financial reporting on economic growth. 21

25 References Acharya, V. V., and K. V. Subramanian Bankruptcy Codes and Innovation. Review of Financial Studies 22: Acs, Z. J., and D. B. Audretsch Innovation in Large and Small Firms - an Empirical- Analysis. American Economic Review 78: Agarwal, V., R. Vashishtha, and M. Venkatachalam, Mutual Fund Transparency and Corporate Mypoia, Working Paper, Georgia State University. Aghion, P., N. Bloom, R. Blundell, R. Griffith, and P. Howitt Competition and innovation: An inverted-u relationship. Quarterly Journal of Economics 120: Aghion, P., J. Van Reenen, and L. Zingales Innovation and Institutional Ownership. American Economic Review 103: Ashbaugh, H., and M. Pincus Domestic Accounting Standards, International Accounting Standards, and the Predictability of Earnings. Journal of Accounting Research 39: Atanassov, J Do Hostile Takeovers Stifle Innovation? Evidence from Antitakeover Legislation and Corporate Patenting. The Journal of Finance 68: Ball, R., A. Robin, and J. S. Wu Incentives versus standards: properties of accounting income in four East Asian countries. Journal of Accounting and Economics 36: Barry, C. B., and S. J. Brown Differential Information and Security Market Equilibrium. Journal of Financial and Quantitative Analysis 20: Barth, M. E Research, standard setting, and global financial reporting. Now Publishers Inc. Bhojraj, S., and R. Libby. Captial market pressure, disclosure frequency-induced earnings/cash flow conflict, and managerial myopia. The Accounting Review 80: Bradshaw, M. T., B. J. Bushee, and G. S. Miller Accounting Choice, Home Bias, and U.S. Investment in Non-U.S. Firms. Journal of Accounting Research 42: Brown, J. R., S. M. Fazzari, and B. C. Petersen Financing Innovation and Growth: Cash Flow, External Equity, and the 1990s R&D Boom. Journal of Finance 64: Brown, J. R., G. Martinsson, and B. C. Petersen Do financing constraints matter for R&D? European Economic Review 56: Brown, J. R., G. Martinsson, and B. C. Petersen Law, Stock Markets, and Innovation. Journal of Finance 68: Burgstahler, D. C., L. Hail, and C. Leuz The Importance of Reporting Incentives: Earnings Management in European Private and Public Firms. The Accounting Review 81: Bushee, B. J The Influence of Institutional Investors on Myopic R&D Investment Behavior. The Accounting Review 73: Byard, D., Y. Li, and Y. Yu The Effect of Mandatory IFRS Adoption on Financial Analysts Information Environment. Journal of Accounting Research 49: Chan, K., V. Covrig, and L. Ng What Determines the Domestic Bias and Foreign Bias? Evidence from Mutual Fund Equity Allocations Worldwide. Journal of Finance 60:

26 Chang, X., G. Hilary, J. Kang, and W. Zhang, Innovation, Managerial Myopia and Financial Reporting. Working Paper, Cambridge University. Chen, L., J. Ng, and A. Tsang The effect of mandatory IFRS adoption on international cross-listings. The Accounting Review forthcoming. Chen, C., Young, D., and Zhuang, Z Externalities of Mandatory IFRS Adoption: Evidence from Cross-Border Spillover Effects of Financial Information on Investment Efficiency. The Accounting Review 88: Chemmanur, T. J., E. Loutskina, and X. Tian Corporate Venture Capital, Value Creation, and Innovation. Review of Financial Studies 27: Christensen, H. B., L. Hail, and C. Leuz Mandatory IFRS reporting and changes in enforcement. Journal of Accounting and Economics 56: Cornaggia, J., Y. Mao, X. Tian, and B. Wolfe Does Banking Competition Affect Innovation? Journal of Financial Economics 115: Covrig, V. M., M. L. Defond, and M. Hung Home Bias, Foreign Mutual Fund Holdings, and the Voluntary Adoption of International Accounting Standards. Journal of Accounting Research 45: Dai, L., R. Shen and B. Zhang, Does the Media Spotlight Burn or Spur Innovation? Working Paper, Australian National University. Daske, H., L. Hail, C. Leuz, and R. Verdi Mandatory IFRS Reporting around the World: Early Evidence on the Economic Consequences. Journal of Accounting Research 46: Adopting a Label: Heterogeneity in the Economic Consequences Around IAS/IFRS Adoptions. Journal of Accounting Research 51: De Franco, G. U. S., S. P. Kothari, and R. S. Verdi The Benefits of Financial Statement Comparability. Journal of Accounting Research 49: DeFond, M., X. Hu, M. Hung, and S. Li The impact of mandatory IFRS adoption on foreign mutual fund ownership: The role of comparability. Journal of Accounting and Economics 51: Diamond, D. W., and R. E. Verrecchia Disclosure, Liquidity, and the Cost of Capital. The Journal of Finance 46: Djankov, S., R. La Porta, F. Lopez-de-Silanes, and A. Shleifer The law and economics of self-dealing. Journal of Financial Economics 88: Fang, V., X. Tian, and S. Tice Does Stock Liquidity Enhance or Impede Firm Innovation? Journal of Finance 69: Florou, A., and P. F. Pope Mandatory IFRS Adoption and Institutional Investment Decisions. The Accounting Review 87: Goto, A., and K. Motohashi Construction of a Japanese Patent Database and a first look at Japanese patenting activities. Research Policy 36: Gupta, N., and K. Yuan On the Growth Effect of Stock Market Liberalizations. Review of Financial Studies 22: Hail, L., C. Leuz, and P. Wysocki. 2010a. Global Accounting Convergence and the Potential Adoption of IFRS by the U.S. (Part I): Conceptual Underpinnings and Economic Analysis. Accounting Horizons 24: