key: cord-1049760-njz1bzi3 authors: Kanagaretnam, Kiridaran; Mawani, Amin; Shi, Guifeng; Zhou, Zejiang title: Impact of social capital on tone ambiguity in banks’ 10-K filings date: 2020-10-16 journal: J Behav Exp Finance DOI: 10.1016/j.jbef.2020.100411 sha: 664ab97dbcbb7ea3db606aff39b1005477d22cfb doc_id: 1049760 cord_uid: njz1bzi3 We examine whether the social capital index of the county where the bank is headquartered is associated with the ambiguity of tone measures constructed from the textual analysis of banks’ 10-K filings. We hypothesize and find that banks located in high social capital areas exhibit lower ambiguous tone in their 10-K filings. Furthermore, the impact of social capital on management’s 10-K disclosure for banks located in high social capital areas is not mitigated during recessionary periods when management may have more unfavorable news to report. Unlike other studies that suggest that social norms can be forsaken when motive and opportunity exist, our results suggest that social capital is reasonably entrenched in banks’ reporting. In contrast, we find that banks located in low social capital areas report more ambiguously during recessionary periods when management may have to report unfavorable news. Like other publicly traded firms, banks disclose in their annual reports their audited financial statements and their 10-K Reports, with the latter constituting almost 80% of the annual report in a typical bank. Management's discretion in audited financial statements is limited due to measurement and reporting conventions imposed by Generally Accepted Accounting Principles (GAAP) and Generally Accepted Auditing Standards (GAAS), as well as bank-specific reporting requirements imposed by regulators (e.g., reporting requirements under Federal Deposit Insurance Corporation Improvement Act (FDICIA)). Principles like consistency, conservatism and materiality limit what management can report in their audited financial statements. However, management has significant discretion in what it can report in the 10-K reports without being limited by consistency, conservatism, materiality or even a historical focus. While the subject matter and structure in the 10-K reports are mandated, the clarity (or ambiguity) of the content is left to management's discretion. Stakeholders' understanding of such information and their resulting interpretation and decisionsdepend non-trivially on the ambiguity of management's tone. The U.S. Securities and Exchange Commission (SEC) has appropriately called for narrative clarity and reduced ambiguity in 10-K filings, and has raised concern that firms may be deliberately ambiguous to protect themselves against potential claims (SEC, 2007) . Such ambiguity may be heightened during recessionary periods when management may have to report unfavorable news. Ambiguity includes -but not restricted to -presenting both positive entrenched in the DNA of the bank, thereby motivating them to report with a less ambiguous tone during periods of financial stability as well as during recessionary periods. This contrasts with Liu, Lu and Veenstra (2014, 289) who find that "social norms can be crossed when motive and opportunity exist." In other words, we find that social norms need not be contingent upon managerial incentives as suggested by Hechter (2008) in his analysis of the rise and fall of the Arthur Andersen accounting firm. More recently, Horowitz (2019) emphasizes the importance of embedding the (social) norms and culture deep enough so that management behaves the right way even when no one is looking. When social norms are set sufficiently at the top, then management and employees will apply it consistently in all areas of reporting as well. Social capital can determine and reinforce mutual trust, as well as more pro-social and less opportunistic behavior. Such social capital can reduce informational and transactional risks, thereby reducing the scope for unethical, opportunistic or selfserving actions. It is in this sense that social capital is rightfully considered as a form of capital. Social capital can enhance economic returns to all sides of the transaction even in situations of information asymmetry and risk. Social capital can mitigate "the fear that one's exchange partner will act opportunistically" (Bradach and Eccles 1989) , and allows both sides to an exchange transaction to anticipate the others' actions with more certainty (Larcker and Tayan 2013). The mutually accepted norms that constitute social capital can be beneficial to all parties to a long-run transaction, with mutual expectations that social contracts would be enforced fairly, ethically and unambiguously. In our specific context, readers of banks' 10-K reports could trust management to take less opportunistic or self-serving actions and report outcomes truthfully, ethically and unambiguously if social capital levels are high. While tone ambiguity is not necessarily lying or manipulation of information, we argue that readers of 10-K filings prefer less ambiguity to more ambiguity and can at least qualitatively detect the relative extent of ambiguity in 10-K reports. Such a setting would incentivize managers to provide less ambiguous disclosure despite extant information asymmetry and despite not having a hard and defining boundary between ambiguous and unambiguous disclosures. The influence of social environment on the quality of firms' financial reporting has been documented by Kang, Han, Salter and Yoo (2010) , Kanagaretnam, Lim and Lobo (2011), and McGuire, Omer and Sharp (2012) . Using a social capital index, Jha (2019) finds that firms headquartered in high social-capital counties report less accrual and real earnings management, while Jha and Chen (2015) find that firms headquartered in counties with higher social capital incur statistically and economically lower audit fees. If auditors can rely more on management in high social-capital counties, then arguably so can readers of 10-K filings. Therefore, if management conceals less about its performance in high social-capital counties, it will also be less ambiguous in their We focus on the banking industry given its unique nature of being highly regulated (at the federal, state, FDIC and SEC levels), and perhaps where informal influences or structures such as social capital may not matter as much. The banking sector also presents a setting where risk-taking incentives are high, the reporting of which may or may not be masked by management with ambiguity. Documenting the existence of social capital within the banking sector would allow us to illustrate that it is a strong and resilient form of capital. Examining the banking industry alone also provides an acid test for an association between social capital and tone ambiguity. We construct our social capital index using principal component analysis (PCA) based on four different publicly available measures of civil society and social organizations at the county level following Rupasingha and Goetz (2008) , as well as an alternate measure of social capital following Guiso, Sapienza and Zingales (2004) and Buonanno, Montoli and Vanin (2009) . Both these measures offer sufficient variation in social capital across counties. Our findings are consistent with our hypotheses. After controlling for bank characteristics based on the prior literature, we find that the ambiguity of tone in 10-K filings is significantly lower in banks that are headquartered in high social-capital counties. In additional tests, we find that the impact of social capital on the ambiguity of tone was greater during recessionary periods of 2001 and 2007-2009 . Ceteris paribus, firms headquartered in high social-capital counties reported less ambiguously in their 10-K filings compared to firms headquartered in low social-capital counties during recessionary periods (2001 and 2007-2009) relative to non-recessionary periods (2002-J o u r n a l P r e -p r o o f Journal Pre-proof 2006 and 2010-2014) . In other words, we did not find the impact of social capital on management's 10-K disclosure altered during recessionary periods, suggesting that social capital is reasonably entrenched in banks' reporting, and not forgotten during bad times when management may have more unfavorable news to report. In robustness tests, we show that social capital continues to mitigate the ambiguity in banks' 10-K using alternate measures of social capital; alternate measures of ambiguity; additional control variables; alternate subsamples; and alternate econometric specifications. Our findings contribute to the literature on the impact of social capital on management's disclosures. Contrary to other studies (e.g., Liu et al. 2014), we document that social norms can be entrenched and therefore adhered to even when incentives for ambiguous reporting are high. Additionally, we document that the social environment influences relationships with investors, and not just auditors (as shown by Jha and Chen, 2015) . Our finding of a negative association between social capital and ambiguous disclosure is a link that can explain two strands of emerging literature: why firms headquartered in high social-capital counties incur lower bank loan spreads The rest of the paper proceeds as follows. The next section reviews the literature and develops our hypotheses. We present the research design and describe the data in section 3, discuss the results in section 4, offer robustness checks in section 5 and conclude in section 6. The seminal paper by Li (2008) examined the relationship between the readability of annual reports and financial performance of firms, and found that firms with lower reported earnings were more difficult to understand. Li estimated the complexity of disclosure with the Fog Index, where a higher reading of the index represents disclosure that is more difficult to understand. In their survey of the literature on textual analysis, Loughran and McDonald (2016) summarize the words selected by managers to describe their operations, and show how they are correlated with future stock returns, earnings, and even future fraudulent activities of management. A seminal study by Loughran and McDonald (2011) attempted to measure ambiguity of disclosure instead of the complexity of disclosure. They claimed that firms that used 'uncertain' (in all its form such as approximate, contingent, indefinite and uncertainty) or weak modal words (e.g., possible, might, approximate and contingent) in their 10-K filings (per 1,000 words) arguably had greater ambiguity, and was associated with higher stock return volatility J o u r n a l P r e -p r o o f Journal Pre-proof in the year following the disclosure. 1 The common element of complexity and ambiguity is that higher scores on both measures make the disclosures harder to understand and therefore more difficult to interpret. 2 In the context of corporate culture, Audi, Loughran and McDonald (2016) "hypothesize that a more frequent count of 21 unique "trust" words, like trust, character, and virtue, in the MD&A [Management Discussion and Analysis] section of a firm's 10-K indicates a corporate culture that involves greater trust," especially if such firms also use audit and control-type words. Balvers, Gaski and McDonald (2016) document that "voluntary use by firms of monitoring and measurement of customer satisfaction is a credible signal that is associated with higher subsequent customer satisfaction." Audited financial statements with notes that are an integral part of the financial reports have limited scope for complexity or ambiguity for sophisticated users. In contrast, the firm's management has room to introduce complexity and ambiguity in their 10-K filings where rules dictate only the subject matter and structure, but not the scope and content. The challenge in determining complexity or ambiguity in corporate disclosures is not about discovering hidden facts, but about establishing the relevance of the fact to readers, or equivalently, management's intent of what the disclosed fact may mean. Gladwell (2007) argues that everything that Jonathan Weil -a reported at the Wall Street Journal -uncovered about Enron was reported by Enron itself, but the tone of the disclosure was not unambiguous. In his legal analysis of the Enron case, Macey (2004) blamed the financial intermediaries for not processing and interpreting the financial information disclosed by Enron, even though he blames Enron for the engineered complexity and the tone ambiguity. With an abundance of information available via the internet, we have access to lots of facts but with limited knowledge of their relevance or the intent of parties disclosing the facts. This makes the examination of tone ambiguity just as important as readability. With the increasing availability of online full-text information databases of SEC reports (including 10-K filings), many researchers are using textual analysis to investigate the link between language attributes and economic decisions or outcomes. We build on the premise that social capital can reduce informational and transactional risks, and therefore reduce the scope for opportunistic or self-serving reporting by management. Social capital exerts its influence via norms of social peers surrounding corporate headquarters as well as via associational networks that make norm-consistent (2015) find that firms headquartered in counties with higher social capital incur statistically and economically lower audit fees, consistent with the notion that trustworthiness is priced. If auditors can rely more on management in high socialcapital counties, then arguably so can readers of 10-K filings. Therefore, if management conceals less about its performance and risks in high social-capital counties, it can afford to be less ambiguous in their 10-K disclosures. This study extends Jha (2019) by documenting that more forthcoming disclosure in counties with higher social capital extends into both recessionary and non-recessionary periods. This study also motivates why the examination of tone ambiguity is just as important as (or even more important than) the analysis of readability conducted by Jha (2019). We extend these two strands of emerging literature to examine whether social capital of the county in which the bank is headquartered is associated with the ambiguity of tone by banks in their 10-K filings. More specifically, our hypothesis is as follows: H1: Banks headquartered in counties with higher social capital exhibit lower ambiguity of tone in their 10-K filings. We further investigate whether the impact of social capital on the ambiguity of tone in 10-K filings is strengthened or mitigated during recessionary periods. An argument for the impact of social capital to remain influential during times of crisis when management has to report bad news is that social capital remains fairly constant or sticky across time, and the effect of social capital (like generosity or philanthropy) can be more pronounced during times of acute crisis when it is needed most. An alternate view is that managers accept the influence of their social environment and the social capital within their counties only when it is not too costly to be so influenced. If the cost of virtuous or unambiguous reporting during a financial crisis becomes too high, then perhaps management may transition to more ambiguous reporting in their 10-K filings. Our second hypothesis is therefore two-sided and specified as follows: H2: The impact of social capital on banks' ambiguity of tone in their 10-K filings during recessionary periods is different from the impact of social capital during non-recessionary periods. Social capital may also be manifested in the importance managers attach to honoring one's word that cannot be contractually obligated, and the value of such social capital may be underestimated. For example, Loughran, McDonald and Yun (2009) suggest that firms may incorrectly believe that the benefit of ambiguous reporting about business conduct may outweigh the cost of losing their long-term credibility (or social capital, in our terms). Liu et al. (2014, 305) document that if the relative price of obeying social norms becomes high, some market participants will "forego their J o u r n a l P r e -p r o o f adherence to social norms for financial rewards." In the context of tax compliance, Blanthorne and Kaplan (2008) find that opportunities to evade income taxes influence the formation of taxpayers' ethical beliefs, which in turn, affect intentions and decisions to evade taxes. Prentice and Miller (1996) describe such "compromises" as becoming part of acceptable future social norms. We therefore examine whether banks in high social capital areas report with a more consistent ambiguity tone during both recessionary and non-recessionary periods compared to banks in low social capital areas. This would be the case if behavior exhibited by high social capital banks is entrenched into the DNA of the banks, and therefore the less ambiguous tone of reporting is maintained even during bad times when management may have to report unfavorable news. Our third hypothesis is specified as follows: H3: Ambiguity of tone in 10-K filings of banks located in high social capital areas will remain unchanged from periods of financial stability to periods of recessions, while the ambiguity of tone in 10-K filings of banks in low social capital areas will decline from periods of financial stability to periods of recessions. We test our hypotheses using a multivariate regression model where the dependent An alternate measure of social capital based on Guiso, Sapienza and Zingales (2004) and Buonanno, Montolio and Vanin (2009) is also used for our robustness checks. The formal regression model estimated is as follows: where subscripts i and t refer to bank and year, respectively, making our unit of analysis a bank-year. BankLevelControls is a vector of bank-specific control variables adapted from Li (2008), and include size, market-to-book ratio, age of the firm, the proportion of total income that is made up of non-interest income, special items ( In our robustness checks, we also include the following independent variables: Fog Index based on Li (2008), scaled length of the annual report in words, scaled size of the 10-K report, natural log of the population in the county, the ratio of male-to-female population in the county, the ratio of married-to-total households in the county, the ratio of population over age 25 who have at least an undergraduate degree, the natural log of per capita personal income of the county, the proportion of the population that selfidentifies as being religious, and the Herfindahl index for ethnic heterogeneity in the county during a year. The variables are defined in the appendix. We do not control for country, institutions and legal origin since our entire sample is from the U.S. We cluster the standard errors at the county level since social capital is Table 2 presents the Pearson correlations. As expected, the correlation between ambiguity and social capital is negative and significant (p < 0.01). Our main multivariate results presented in table 3 -are consistent with our first hypothesis. Ceteris paribus, banks located in counties with high social capital exhibit lower ambiguous tone in their 10-K filings under both our measures of ambiguity. Because of the ordinal nature of the dependent variable Financial Uncertainty and Financial Weakness -we also estimate ordered logistic regressions (not reported) and obtain similar results to OLS regressions (in columns 1 and 2 of Table 3 ). Size of the bank is found to be positively associated with ambiguity when ambiguity is measured by Financial Uncertainty, but not statistically associated when ambiguity is measured by Financial Weakness. The positive association may be justified because larger banks have more transactions as well as more complex interactions that naturally lead to more ambiguous reporting. The lack of association with Financial Weakness may be justified on the grounds that larger banks have more stable operations that can withstand variations or shocks, and financially stronger (including perhaps "too big to fail"). Similarly, age is marginally negatively associated with ambiguity for arguably the same reason. On the grounds that norms and values may matter less for more dispersed firms, we decompose our sample into small banks (total assets < $1 billion) and large banks (≥ $1 billion), and find the coefficients of the two sub-groups are not statistically different (results not presented). Growth -as proxied by market-to-book ratio -is found to be marginally negatively associated with ambiguity when ambiguity is measured by Financial Uncertainty, but not significantly associated when ambiguity is measured by Financial Weakness. The proportion of income not related to interest is positively associated with ambiguity as measured by Financial Uncertainty, suggesting that banks may have a harder time communicating about operations outside their natural deposit-taking and loan granting line-of-businesses. Finally, the volatility of monthly stock returns is positively associated with ambiguity, suggesting that management may find it difficult to convey the higher levels of operational risk in their 10-K filings. J o u r n a l P r e -p r o o f Table 4 presents the results of our second hypothesis. We find that the mitigating impact of social capital on the ambiguity of tone in 10-K filings is maintained at a statistically significant level (when ambiguity is measured by Financial Uncertainty it is in fact is enhanced) during the recessionary periods (2001 and 2007-2009) compared to the non-crisis period (2002-2006 and 2010-2015) . This suggests that the impact of social capital is reasonably consistent over time, and continues to be exhibited (like generosity or philanthropy) during times of financial crisis when it may be most necessary to readers of 10-K filings. 5 Growth banks had a marginally negative impact on the ambiguity of 10-K reports. Non-interest income contributed to greater ambiguity when ambiguity is measured by for both measures of ambiguity (although interaction variable is positive for both measures, the total effect is not statistically different to zero). Our results suggest that (similar to philanthropy or generosity) high social capital becomes entrenched in such banks and continues to serve as a mitigating factor as banks report their 10-K filings in a less ambiguous tone during both good times and bad. 6 Table 5 presents results for the potential impact of social capital on some commonly Table 5 further shows that social capital has a mitigating effect on the Fog index of a bank (Column 1) and on the size of the 10-K computer file (Column 3), but not on the length of the 10-K report (Column 2). with lower levels of ambiguity tone in banks' 10-K reports. In summary, our key results continue to hold after all sensitivity tests. While management's discretion on how and what to report is limited to audited financial statements, it has significant discretion in what it can report in the 10-K filings without being too limited by consistency, conservatism, materiality or even a historical focus. While the subject matter and structure in the 10-K reports are mandated, the clarity or ambiguity of the content is left to management's discretion. Stakeholders' understanding of such information and their resulting interpretation and decisionsdepend non-trivially on the ambiguity of management's tone. Ambiguous 10-K Reports contribute to information risk and can reduce stakeholders' ability to understand or assess firms' investment and financing risks, and therefore valuation. Deliberate ambiguity that serves corporate or managerial interests at the expense of other stakeholders' interests can lead to a decline in investor confidence. We examine whether the ambiguity of the tone in the bank's 10-K filings is associated with the social capital in the local county where the bank is headquartered. We argue that textual reporting is consistently less ambiguous in banks headquartered in counties with high social capital since such banks are more forthcoming and therefore have fewer reasons for opportunistic or self-serving reporting. Social capital can enhance economic returns to all sides of the transaction in situations of information asymmetry and risk, and can mitigate the fear that one's exchange partner will act opportunistically. In our context of 10-K reports, investors could trust management to look after their long-run interests and report outcomes truthfully, consistently and unambiguously if social capital levels are high. Our findings are consistent with our hypothesis. After controlling for bank characteristics, we find that banks' ambiguity of tone in their 10-K filings is significantly lower in banks that are headquartered in high social-capital counties. In further tests, we find that the impact of social capital on ambiguity of tone was Fin_Unc t Fin_Unc is the proportion of occurrences for uncertainty words for each year for every 10 words in a 10-K report. Uncertainty word list includes words denoting uncertainty, focusing on the general notion of imprecision rather than exclusively focusing on risk, such as approximate, contingency, depend, fluctuate, uncertain, variability. Fin_Weak is the proportion of occurrences for weak words for each year for every 10 words in a 10-K report. The list of weak words expresses low level of confidence, such as could, might, depending, possibly. The word list is from Loughran and McDonald (2011). Pvote t-1 Respn t-1 Nccs t-1 Percentage of voters who voted in presidential elections Response rate to the Census Bureau's decennial census Number of tax-exempt non-profit organizations per 10,000 of population Assn t-1 Number of social organizations per 100,000 of populations SC1 t-1 Social capital is defined as the first principal component of a principal component analysis (PCA) based on the above four NRCRD variables at the county level. This variable is the measure of social capital at the county level. It is constructed following Rupasingha and Goetz (2008) . Specifically, the variable is constructed by using the first component from a principal component analysis that uses four different measures. For example, we use the following four measures: assn97, nccs97, pvote96, respn00 for 1997, where assn97 is the sum of the religious organizations, civic and social associations, business associations, political organizations, professional organizations, labor organizations, bowling centers, physical fitness facilities, public golf courses, sport clubs, and recreation clubs, and membership organizations not elsewhere classified in 1997. We divide the total by 10 because there are 10 different categories. Further, we also divide it by the population of the county. We then multiply it by 10,000. The measure nccs97 is the total number of nongovernment organizations excluding the ones with an international focus in 1997 divided by the population multiplied by 10,000. The measure pvote96 is the number of votes casted divided by the population above 18 times 100. The measure respn00 is the census response rate. Then we use a principal component analysis and use the first component to construct the social capital index for each county. We use an analogous approach for 2005 and 2009. For each of these years, we use the presidential elections and census response closest to 2005 and 2009, respectively. We then linearly interpolate and fill the social capital data for the in-between years. Source: Northeast Regional Center for Rural Development (NERCRD), Rupasingha and Goetz (2008) SC2 t-1 This variable is the state-level per capita organ donor, which is the total number of organ donors in a state in a given year divided by total state population in that year multiplied by 1,000. Donor is a person from whom at least one organ or tissue is recovered for the purpose of transplantation. Organ donation data can be obtained from the OPTN via the link: http://optn.transplant.hrsa.gov/latestData/stateData.asp?type=state. We follow Guiso, Sapienza, and Zingales (2004) and Buonanno, Montolio, and Vanin (2009) to construct this variable. Natural log of market value of common equity (log(PRCC_F*CSHO)). Source: Compustat Bank Fundamentals. MTB t-1 Market-to-book ratio at year-end (PRCC_F*CSHO/ ceq). Source: Compustat Bank Fundamentals. ROA t-1 Net income divided by book assets (NI/AT) in fiscal year t-1. Source: Compustat Bank Fundamentals. Deposits/Assets t-1 Total deposits divided by book assets (DPTC/AT) in fiscal year t-1. Source: Compustat Bank Fundamentals. FIRM_AGE t-1 Number of years that a firm has been in CRSP monthly stock return files. NIINT_INC t-1 Ratio of non-interest income to the sum of interest and noninterest incomes (TNII/(NIINT+TNII)). Source: Compustat Bank Fundamentals. SI t-1 Special items scaled by book value of assets. RET_VOL t-1 Standard deviation of the monthly stock returns. MA t-1 Indicator variable coded 1 for non-zero spending on acquisitions (AQC). Source: Compustat Bank Fundamentals. DLW t-1 Indicator variable coded 1 if a company is incorporated in Delaware and 0 otherwise. Big5 t-1 An indicator variable coded 1 if audited by a big 5 firm and 0 otherwise; Analysts t-1 Number of analysts for the latest consensus forecast (numest). If this number is not available for a firm, then the number of analysts following is assumed to be zero. Bank Type includes (1) Fog t Fog is the Fog Index of the firm's annual report. Fog index of the firm's annual report, defined as 0.4 multiplied by the sum of the average number of words per sentence and the percentage of complex words. Complex words are those with more than two syllables. Length t Length is the number of words in the firm's annual report. File size t File size refers to the file size of 10-K. The file size of 10-K is in megabytes of the SEC EDGAR "complete submission text file" for the 10-K filing Financial Crisis An indicator variable that equals 1 for fiscal years 2007, 2008 and 2009, and 0 otherwise. The natural logarithm of total resident population in the county. Source: Bureau of Economic Analysis. Income t-1 The natural logarithm of median household income in the county. Source: Bureau of Economic Analysis. Education t-1 The natural logarithm of total public school enrollment in the county. Source: U.S. Census. Local religion t-1 The number of self-described religious adherents in a county (from ARDA) divided by the total population in the county (from U.S. Census). Data on religiosity are available for years 1990, 2000, and 2010. We linearly interpolate the data to obtain the values for the years in between and after 2010. Source: ARDA. Industry adjusted EBTP t-1 Industry adjusted EBTP = EBTP-industry average EBTP, and EBTP is a bank's earnings before tax and provision. Ethnicity homogeneity t-1 A Herfindahl index calculated across four basic Census tract ethnic categories including Hispanic, non-Hispanic black, non-Hispanic white, and Asian in a county during a year. J o u r n a l P r e -p r o o f Less: missing social capital data (747) Less: missing control variables in Equation (1) (1,848) Final sample 5,621 Year Number of firm-years Percent% (15) (16) (17) (1)Fin_Unc t 1.000 (2)Fin_Weak t 0.790 Notes: Fin_Unc t is the proportion of occurrences for uncertainty words for each year for every 10 words in a 10-K report. Fin_Weak is the proportion of occurrences for weak words for each year for every 10 words in a 10-K report. The independent variables are defined in the Appendix. J o u r n a l P r e -p r o o f Notes: Fin_Unc t is the proportion of occurrences for uncertainty words for each year for every 10 words in a 10-K report. Fin_Weak is the proportion of occurrences for weak words for each year for every 10 words in a 10-K report. The independent variables are defined in the Appendix. J o u r n a l P r e -p r o o f (2008). Length is the number of words in the firm's annual report. File size refers to the file size of 10-K. The independent variables are defined in the Appendix. J o u r n a l P r e -p r o o f Notes: Fin_Unc t is the proportion of occurrences for uncertainty words for each year for every 10 words in a 10-K report. Fin_Weak is the proportion of occurrences for weak words for each year for every 10 words in a 10-K report. The independent variables are defined in the Appendix. J o u r n a l P r e -p r o o f Notes: SC1 is the measure of Social Capital at the county level. Fin_Unc t is the proportion of occurrences for uncertainty words for each year for every 10 words in a 10-K report. Fin_Weak is the proportion of occurrences for weak words for each year for every 10 words in a 10-K report. The independent variables are defined in the Appendix. J o u r n a l P r e -p r o o f Trust, but Verify: MD&A Language and the Role of Trust in Corporate Culture Financial Disclosure and Customer Satisfaction: Do Companies Talking the Talk Actually Walk the Walk? It pays to partner with a firm that writes annual reports well How much should we trust differences-in-differences estimates? How does financial reporting quality relate to investment efficiency? An egocentric model of the relations among the opportunity to underreport, social norms, ethical beliefs, and underreporting behavior. Accounting Using 10-K Text to Gauge Financial Constraints The impact of narrative disclosure readability on bond ratings and rating agency disagreement We have received unconditional financial support for this research from Social Sciences & Humanities Research Council of Canada. We have no financial arrangements that might give rise to conflicts of interest with respect to the research reported in this paper.Amin Mawani, York University, Toronto, Canada J o u r n a l P r e -p r o o f