1

Short Selling and Readability in Financial Disclosures: A Controlled

Experiment

Minxing Sun

Department of Finance

Clemson University

msun@clemson.edu

Weike Xu*

Department of Finance

Clemson University

weikex@clemson.edu

May, 2018

Abstract

We examine the causal effect of short-selling on a firm’s annual report readability using

Regulation SHO, which relaxes short-sale constraints for a random sample of pilot stocks. Pilot

firms’ annual report readability decreases during the experiment period. This short-selling effect

on 10-K readability is more pronounced for firms that receive less investor attention and for

firms with worse news. Pilot firms also increase the use of uncertainty words in 10-Ks during the

experiment period. Our results suggest that firms produce less transparent 10-Ks that are more

costly for investors to comprehend when short-sale constraints are less rigorous.

Keywords: Regulation SHO; Short-selling; Annual report readability; Limited attention

JEL Codes: G14, G18, M41

* Minxing Sun: College of Business, Clemson University, msun@clemson.edu. Weike Xu: College of Business,

Clemson University, weikex@clemson.edu. A previous version of this paper was titled “Short Selling and

Readability in Financial Disclosure.” We thank Sris Chatterjee, Laura Field, Jon Garfinkel, Hugh Hoikwang Kim,

Jin-mo Kim, Jun Li, Rose Liao, Ankur Pareek, Lin Peng, Ben Sopranzetti, Yangru Wu, and seminar participants at

the 2017 Eastern Finance Association annual meeting, Central University of Finance and Economics, Clemson

University, and Rutgers University for helpful comments. All remaining errors are our own.

1

I. Introduction

An important channel for corporate managers to communicate a firm’s financial

disclosures to investors and analysts is the annual report filed under the Securities Exchange Act

of 1934, namely, Form 10-K. Market participants and regulators care about the quantity of

information available to the public, as well as the quality of the information provided in financial

reports. Thus, the readability and other aspects of text analysis in the context of financial

disclosures are crucial to measure the effective communication of valuation-relevant information

between the firm and capital market participants (Loughran and McDonald (2014), hereafter

LM).

Many researchers in accounting and finance have examined the effects of annual report

readability on earnings persistence (Li (2008)), the trading activities of small and retail investors

(Miller (2010) and Lawrence (2013)), firms’ borrowing costs (Ertugrul, Lei, Qiu, and Wan

(2017)), and the valuation of closed-end funds (Hwang and Kim (2017)).1 Recently, LM (2014)

demonstrate that 10-K file size (in megabytes) is a good and robust proxy of the readability of

financial reports. They find that a less readable 10-K (larger file size) is associated with a higher

valuation ambiguity, as demonstrated by higher return volatility, as well as greater earnings

forecast errors and dispersion. These findings of these studies indicate that annual report

readability has significant impacts on corporate decisions and financial markets. How the

1 Others have linked annual report readability to capital investment efficiency (Biddle, Hilary, and Verdi (2009)),

analyst coverage and analyst dispersion (Lehavy, Li, and Merkley (2011)), and long-term return volatility (Belo et al.

(2016)). For a detailed review, see LM (2016).

2

activities of short sellers impact annual report readability is, however, unclear. In this paper, we

address this insufficiency by examining the causal effect of short selling on annual report

readability using a regulatory experiment (Regulation SHO Pilot Program), which relaxes short-

sale constraints for a random sample of pilot firms.

We focus on short selling for two reasons. First, an increase in short-selling activity is

associated with a future decrease in stock returns (e.g., Desai, Thiagarajan, and Balachandran

(2002), and Diether, Lee, and Werner (2009)). Thus, managers care about the potential amount

of short selling in their firms and they take a variety of actions to impede short selling. For

example, Lamont (2012) finds that firms use legal threats, investigations, lawsuits, and various

technical actions to prevent short selling. Second, recent evidence suggests that short selling can

affect managers’ reporting behaviors (e.g., Karpoff and Lou (2010), and Fang, Huang, and

Karpoff (2016)). Recent research shows that annual report readability can be strategically used

by managers to obfuscate earning-relevant information (e.g., Li (2008); LM (2014)). Yet, when

companies encounter short selling pressure, how corporate managers present valuation-relevant

information to investors is unknown. Investigating how short selling affects annual report

readability can provide insight into managers’ decision on information disclosure.

Testing the causal effects of short selling on annual report readability is empirically

challenging due to its endogenous nature. To overcome this issue, we employ an identification

strategy based on the Security and Exchange Commission’s (SEC’s) approval of Regulation

SHO Pilot Program (hereafter, Reg SHO), which removes short-sale price tests for pilot firms

3

that are randomly chosen from the Russell 3000 Index. From May 2, 2005 to August 6, 2007,

986 pilot firms were exempted from short-sale price tests and significantly reduced short-sale

constraints as opposed to non-pilot firms. Prior studies document that short-selling activities

increase significantly for pilot stocks compared to non-pilot stocks (e.g., SEC (2007), Diether,

Lee, and Werner (2009), and Grullon, Michenaud, and Weston (2015)). As Reg SHO is an

exogenous shock to short-sale constraints and with both beginning and ending dates, we can

examine the causal effect of the variation in short-sale constraints on annual report readability

using a difference-in-differences (hereafter, DiD) method.

We begin by confirming that pilot stocks are randomly selected by comparing the firm

characteristics of the pilot and non-pilot firms one year before the announcement of the pilot

program. Following LM (2014), we use 10-K document file size as a proxy for annual report

readability. An annual report with a larger 10-K file size is considered less readable. We find that

the pilot firms have similar firm characteristics to the non-pilot stocks before the pilot program.

We then run DiD regression analysis to investigate how the relaxation of short-sale constraints

affects annual report readability. We demonstrate that readability for the pilot firms is 12.4%

lower than that for the non-pilot firms during the Reg SHO compared to the pre-Reg SHO period.

In addition, the SEC eliminated short-sale price tests for all exchange-listed stocks on July 6,

2007. This setting provides an alternative approach for us to further confirm the causal relation

between changes in short-sale constraints and annual report readability. According to our DiD

analysis, non-pilot firms, whose short-sale constraints are significantly relaxed after the pilot

program period, increase 10-K file sizes (decrease 10-K readability) by 7.9% as opposed to pilot

4

firms. Our results suggest that firms produce less readable 10-Ks when short-sale constraints are

less rigorous.

A plausible explanation for our results is that, when faced with short-selling pressure,

pilot firms bury mandated earnings-relevant information in less readable financial disclosures

that are more costly for investors to comprehend. This reporting behavior may help reduce the

potential amount of short selling for three reasons.2 First, both naïve and sophisticated investors

are subject to limited attention and information processing power (e.g., Fang, Peress and Zheng

(2014)). Second, investors pay less attention to, place less weight on and even neglect complex

and hard-to-process information (e.g., Hirshleifer and Teoh (2003); Hirshleifer, Lim and Teoh

(2011); Cohen and Lou (2012); and Hirshleifer, Hsu and Li (2017)). Third, firms exploit the

limited attention of investors (see Daniel, Hirshleifer, and Teoh (2002) for a review). For

example, Hirshleifer and Teoh (2003) argue that owing to limited attention, firms manage

accounting disclosure and reporting choices to manipulate investors’ perceptions and thereby

create mispricing. Our study indicates that managers may use annual report readability to affect

investor perception when faced with short-selling pressure.

To support the above explanation, we conduct several cross-sectional tests. If the above

argument is true, firms that receive less investor attention are more likely to manipulate annual

report readability when faced with short-selling pressure. Using firm size, institutional ownership

and analyst coverage as proxies for investor attention, we demonstrate that the effect of short

2 We refer “the potential amount of short selling” to the potential new or increased short positions taken by investors.

5

selling on annual report readability is more pronounced for firms that receive less investor

attention. Additionally, firms with good earnings have no incentives to obscure valuation-

relevant information. We also show that firms significantly reduce annual report readability only

when there is bad news.

In addition to readability, ambiguous text in 10-Ks can obstruct investors’ ability to

comprehend reports. LM (2011) find that firms with a greater percentage of uncertainty words

(e.g., approximate, contingency, depend, and uncertain) in annual reports are positively

correlated with subsequent stock return volatility after the 10-K filing. LM (2013) show that

IPOs with high frequencies of uncertainty words are associated with higher first-day returns,

absolute offer price revisions, and subsequent return volatilities. We examine whether the

relaxation of short-sale constraints affects tone ambiguity in 10-Ks. Using the proportion of

uncertainty words as a proxy for the tone ambiguity of 10-Ks, we demonstrate that pilot firms

use greater uncertainty text in annual reports during the Reg SHO experiment period.

Our study makes several contributions to the literature. First, we add to the knowledge of

the effects of short selling on corporate decisions. Grullon, Michenaud, and Weston (2015)

investigate the impact of short-sale constraints on investment and financing policies. Moreover,

Fang, Huang, and Karpoff (2016), and De Angelis, Grullon, and Michenaud (2015) find the

effects of short selling on earnings management, and the design of executive incentive contracts,

respectively. In this paper, we focus on the causal effect of short selling on managers’ decisions

on the wording of financial disclosures, namely, readability and tone ambiguity. Second, we

6

identify a new determinant of readability and tone ambiguity in 10-Ks, namely, short-sale

constraints (for a review, see LM (2016)). Third, our study contributes to the debates on the costs

and benefits of short selling. On the one hand, advocates argue that short sellers can curb

financial misconduct, smooth price discovery, and improve market efficiency (e.g., Boehmer,

Jones, and Zhang (2008), Diether, Lee, and Werner (2009), and Fang, Huang and Karpoff

(2016)). On the other hand, critics claim that short selling can adversely affect stock prices and

increase market volatility because of overselling (e.g., Haruvy and Noussair (2006); Goldstein

and Guembel (2008); and Henry and Koiski (2010)).3 We provide evidence that an exogenous

relaxation of short-sale constraints leads firms to produce less transparent 10-Ks, which are

presumably more costly for investors to comprehend. Fourth, our paper is also related to the

literature on how investor attention affects stock market (See Daniel, Hirshleifer, and Teoh (2002)

for a review). Our cross-sectional analyzes indicate that when faced with short-selling pressure,

firms exploit the limited attention and processing power of investors by producing less readable

10-K reports.

The rest of the paper is organized as follows. In Section II, we discuss the related

literature. We describe the data and sample selection in Section III and report the summary

statistics. The main findings and robustness tests are discussed in Section IV. We provide

concluding remarks in Section V.

3 Additionally, the exchanges and listed firms expressed support for short-sale restrictions in public comments. In a

2008 NYSE survey, 85% of CEOs, CFOs, and investor relation officers surveyed were in favor of re-instituting the

short-sale price tests (Opinion Research Corporation (2008)).

7

II. Related Literature

We discuss short-sale price tests in the U.S. equity markets, and literature on how short

selling can affect financial markets in Section A. The studies of readability and tone ambiguity

in financial disclosures are described in Section B.

A. Short-sale Price Tests and Regulation SHO

Short-sale price tests were initially introduced in the equity markets in the United States

in the 1930s to avoid bear raids by short sellers in declining markets. The NYSE adopted an

uptick rule in 1935, which was replaced in 1938 by a stricter SEC rule, Rule 10a-1, also known

as the “tick test.” According to this rule, a short sale can only occur at a price above the most

recently traded price (plus tick) or at the last traded price if it exceeds the last different price

(zero-plus tick). In 1994, the National Association of Securities Dealers (NASD) adopted its own

price test (the “bid test”) under Rule 3350. According to Rule 3350, a short sale occurs at a price

one penny above the bid price if the bid is a downtick from the previous bid.

In July 2004, the SEC announced Reg SHO to provide a new regulatory framework for

short-selling in the U.S. stock markets. Reg SHO removed the tick test for a group of randomly

selected stocks from the Russell 3000 Index in order to evaluate the effectiveness and necessity

of short-selling restrictions. On July 28, 2004, 986 firms were selected as the pilot firms. Their

stocks were exempt from the tick test from May 2, 2005 to August 6, 2007. The SEC

permanently suspended the tick test for all the publicly-traded U.S. stocks on July 6, 2007. Short-

8

selling activities have been shown to significantly increase for pilot firms (e.g., SEC (2007),

Alexander and Peterson (2008), Diether, Lee, and Werner (2009), and Grullon, Michenaud, and

Weston (2015)). The suspension of the tick test drew criticisms from firms and former regulators,

including former SEC chairman Christopher Cox. The criticism intensified during the 2007-2009

financial crisis due to the concern that financial stocks may have been subject to market

manipulation via short-selling. On February 24, 2010, the SEC reinstated the uptick rule for

situations when a security’s price drops by 10% or more from the last day’s closing price.

There is a rich literature on how short selling impacts asset prices (e.g., Miller (1977),

Jones and Lamont (2002), Boehmer, Danielsen, and Rodrigo (2005), Battalio and Schultz (2006),

Doukas, Kim and Pantzails (2006), Diether, Lee, and Werner (2009), Beber and Pagano (2013),

Boehmer, Jones, and Zhang (2008, 2013) and Chu, Hirshleifer and Ma (2016)). Empirical studies

on the effect of short selling on corporate decisions are limited but growing. Gilchrist,

Himmelberg, and Huberman (2005) show that short-sale constraints distort investment and new

equity issues. Using Regulation SHO, Grullon, Michenaud, and Weston (2015) find that the

relaxation of short-sale constraints reduces investment and stock issues. Fang, Huang, and

Karpoff (2016) document that short-selling activities reduce earnings management. Using

Regulation SHO, others link short selling with corporate innovation (He and Tian (2014)), the

design of executive incentive contracts (De Angelis, Grullon, and Michenaud (2015)), corporate

social responsibility (Gao, He, and Wu (2015)), and management forecasts (Li and Zhang

(2015)).

9

B. Readability and Tone Ambiguity in Financial Disclosures

There is extensive discussion in the literature as to the impact the readability of financial

disclosures has on equity market participants. Li (2008) examines the relation between annual

report readability and firm performance using the Fog Index and the number of words contained

in the annual report. He finds that firms with lower reported earnings tend to have annual reports

that are harder to read (i.e., high Fog Index values or high word counts). Li also shows that

companies with more readable annual reports have higher earnings persistence.

Biddle, Hilary, and Verdi (2009) demonstrate that firms with high reporting quality are

associated with greater capital investment efficiency. Guay, Samuels, and Taylor (2016) show

that firms with less readable annual reports tend to mitigate this negative readability effect by

issuing more managerial forecasts of earnings per share, sales, and cash flows. Miller (2010)

documents that small investors trade significantly fewer shares of firms with high Fog Index

values and word counts around the 10-K filing date. In addition, Lehavy, Li, and Merkley (2011)

find that more readable annual reports have lower analyst dispersion and greater earnings

forecast accuracy.

LM (2014) demonstrate that the Fog Index is a poorly specified readability measure when

applied to business documents. The second component of the Fog Index, percentage of complex

words (those with three or more syllables), adds measurement errors to measure readability of

10-Ks. In particular, LM show that 52 complex words like corporation, company, management,

and operations are widely and frequently in annual reports and are account for more than 25% of

10

the complex word count in the 10-K sample. These common financial terms would likely be easy

for investors or analysts to comprehend. LM find that the Fog Index does not provide significant

explanatory power for analyst dispersion or earnings surprises. They propose that the natural log

of gross 10-K file size is a relevant and robust readability measure. They document that firms

with larger 10-K file sizes are significantly linked with larger subsequent stock return volatility,

analyst dispersion, and absolute earnings surprises 4 . Using 10-K file size as a proxy for

readability, Ertugrul, Lei, Qiu, and Wan (2016) provide evidence that firms with lower annual

report readability are associated with higher cost of borrowing. Moreover, Hwang and Kim

(2016) document equity closed-end funds whose annual reports have lower readability trade at

significant discounts.

LM (2011) develop a word list of ambiguity tones in annual reports. They find that the

firms using fewer uncertainty words are associated with a more positive market reaction and

higher return volatility after the 10-K filing period. LM (2013) document a positive relation

between the uncertainty tone in Form S-1 IPO filings and IPO performance. Specifically, they

find that IPOs with higher frequencies of uncertainty words are associated with higher first-day

returns, higher absolute offer price revisions, and higher subsequent volatilities. Furthermore,

Ertugrul, Lei, Qiu, and Wan (2016) demonstrate that firms with more ambiguous tone in annual

reports experience a higher cost of borrowing.

4 One may argue that 10-K document file size may be a proxy for disclosure. A firm with a larger file size is

considered as more disclosure and not less readable. However, this positive relation between file size and volatility,

earnings surprises, and analyst dispersion is inconsistent with this interpretation. If the argument is true, we would

expect this relation to be positive (LM (2014)).

11

III. Data Description

A. Sample Selection

Our sample is constructed based on the Russell 3000 Index in June 2004. We exclude

stocks that were not previously subject to price tests (i.e., not listed on NYSE, AMEX, or

NASDAQ-NM) and stocks that went public or had spin-offs after April 30, 2004. Then we sort

the stocks based on their daily dollar volume computed over the June 2003 to May 2004 period.

Our initial sample includes 2,952 stocks (986 pilot and 1,966 non-pilot stocks)5.

We obtain the SEC annual filing data from the WRDS SEC Readability and Sentiment

database. This database contains detailed information about firms’ SEC filings since 1994,

including filing date, file size, the proportion of uncertainty words, etc. Following LM (2014),

we include all 10-K filings (i.e., 10-K 405, 10-KSB, and 10-KSB40 filings)6. We require that

firms have a Compustat Permanent Company Identifier match, be ordinary common stock, have

at least 2,000 words in the 10-K, and have a gap of at least 180 days between two filings. Our

control variables are from several sources. First, we collect accounting information from the

CRSP/Compustat Merged Database, stock returns from CRSP, and institutional holdings from

Thomson Reuters Institutional (13-f) Holdings. Second, we gather analyst coverage data from

5 Our sample of pilot and non-pilot stocks is identical to that of Fang, Huang, and Karpoff (2016). We thank Vivan

Fang for sharing the Russell 3000 index merged with CRSP PERMNO numbers and FTSE Russell for sharing the

Russell 3000 membership list with us. 6 We focus on 10-Ks and not 10-Qs for two reasons: (1) 10-Ks are more informative to investors; (2) 10-Qs are

shorter in length and report unaudited financial statements (LM (2014).

12

IBES and corporate events information from Thomson Reuters SDC Platinum M&A and Global

New Issues databases.

Our sample period is 78 months. Our sample includes firms whose fiscal year ending

dates are between May 1, 2002 and June 30, 2004 for pre-event period, between May 1, 2005

and June 30, 2007 for the during-event period, and between May 1, 2008 and June 30, 2010 for

post-event period. We classify May 1, 2005 to June 30, 2007 as the during-event period because

the Reg SHO program effectively ran from May 2, 2005 to July 6, 2007. In our sample, we

exclude financial firms (SIC 6000-6999) and regulated utilities (SIC 4900-4949) because

disclosure requirements are significantly different for these highly regulated industries. We also

require that firms have non-missing data for all key variables. Our unbalanced sample includes

1,899 stocks (630 pilot and 1,269 non-pilot firms). We also construct a balanced sample by

requiring firms to be in the sample over the pre-event and the during-event periods. The balanced

sample contains 1,056 firms (382 pilot and 674 non-pilot firms). We use the balanced sample for

most of our tests, but also verify the robustness of our analysis using the unbalanced sample.

B. Key Variables

Following LM (2014), we measure the annual report readability of our sample firms

using the natural logarithm of 10-K report size. Following Li (2008), we control for a set of firm

characteristics that determine annual report readability. Our control variables include size (the

natural logarithm of the market value of equity at the end of the fiscal year), firm age (the natural

logarithm of firm age since its first appearance in the CRSP monthly return file), special items

13

(special items to asset ratio), stock return volatility, earnings volatility, business complexity (the

natural logarithm of the number of business and geographic segments), financial complexity (the

natural logarithm of the number of non-missing items in Compustat), and corporate events (SEO

and M&A dummy variables). We also include profitability (ROA) because firms with lower

profitability are more likely to obscure valuation-relevant information in annual reports.

Following LM (2011), we use the proportion of uncertainty words to capture tone ambiguity in

10-Ks. Descriptions of all the variables are in the Appendix.

C. Summary Statistics

Table 1 reports the summary statistics of all key variables from the unbalanced sample.

All variables are winsorized at the 1% and 99% levels to minimize the influence of outliers. On

average, a firm’s annual report has a file size of 1.86 megabytes and contains 1.48% uncertainty

words. The average firm also has a market value of $4.96 billion, a book-to-market ratio of 0.6,

has been in business for 22.92 years, has 2.26 business segments and 2.76 geographic segments,

a return volatility of 0.13, and an earnings volatility of 0.06. Additionally, on average, a firm has

a 0.02 ROA, 358.85 non-missing items, and a special item ratio of -0.02.

IV. Results

A. Firm Characteristics before Regulation SHO

Reg SHO is a natural experiment to study the causal effects of short selling on annual

report readability because the selection of pilot and non-pilot firms is random and the costs of

14

short selling are significantly reduced for pilot firms. Therefore, a DiD method is appropriate to

study the effects of short selling on annual report readability.

To verify the selection of pilot firms is random, we compare the firm characteristics of

the pilot and non-pilot firms one year before the announcement of the program (July 2004).

Table 2 presents the summary statistics and mean differences of firm characteristics between the

pilot (treatment) and non-pilot (control) groups for the balanced sample. We report the t-statistics

of the two-sample t-tests and z-statistics of the Wilcoxon signed rank sum tests. We find that the

groups have similar firm characteristics despite pilot firms having a lower proportion of

uncertainty words in their annual reports and exhibiting a lower earnings volatility. The results in

Table 2 show that Reg SHO is a well-controlled experiment that is appropriate for testing the

effects of the relaxation of short-sale constraints on readability in 10-Ks.

B. Multivariate Difference-in-Differences Results

In this subsection, we examine the effect of short selling on annual report readability

using a DiD methodology for multivariate regressions. We estimate the following specification

for the balanced sample:

𝐿𝑜𝑔(𝑓𝑖𝑙𝑒 𝑠𝑖𝑧𝑒𝑖,𝑡) = 𝛼 + 𝛽1 ∗ 𝑃𝑖𝑙𝑜𝑡𝑖 + 𝛽2 ∗ 𝑃𝑖𝑙𝑜𝑡𝑖 ∗ 𝐷𝑢𝑟𝑖𝑛𝑔𝑡 + 𝑌𝑒𝑎𝑟𝑡 + 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑗 + 𝜀𝑖,𝑡, (1)

where 𝐿𝑜𝑔(𝑓𝑖𝑙𝑒 𝑠𝑖𝑧𝑒𝑖,𝑡) is the natural logarithm of 10-K document file size for firm i in year t.

𝑃𝑖𝑙𝑜𝑡𝑖 is a dummy variable that equals one if a stock is selected as a pilot stock in Regulation

SHO’s pilot program and zero otherwise. 𝐷𝑢𝑟𝑖𝑛𝑔𝑡 is a dummy variable that equals one if the end

15

of a firm’s fiscal year t falls between May 1, 2005 and June 30, 2007 and zero otherwise.

Industry and Year are the industry fixed effects (2-digit SIC codes) and fiscal year fixed effects

dummies, respectively. The variable 𝐷𝑢𝑟𝑖𝑛𝑔𝑡 is omitted because it is perfectly correlated with

the fiscal year fixed effects. All standard errors are clustered by firm.

The regression results of equation (1) are reported in column (1) in Table 3. The

coefficient of interest is 𝛽2, which captures the causal effect of short selling on annual report

readability. The coefficient of 𝑃𝑖𝑙𝑜𝑡𝑖 ∗ 𝐷𝑢𝑟𝑖𝑛𝑔𝑡, 𝛽2 is 0.116 and is significant at the 1% level,

implying that the 10-K file sizes of pilot firms are 11.6% higher than those of non-pilot firms

during the Reg SHO period as opposed to the pre-Reg SHO period. The coefficient of 𝑃𝑖𝑙𝑜𝑡𝑖 is

insignificant, suggesting that all firms exhibit similar 10-K file sizes before the pilot program.

We augment equation (1) by including control variables previously shown to determine

the annual report readability: size, book-to-market ratio, firm age, special items to asset ratio,

stock return and earnings volatility, business complexity, financial complexity, ROA, and

corporate events (SEO and M&A dummy variables). We also add industry fixed effects and year

fixed effects. The results in column (2) in Table 3 show that the coefficient on 𝑃𝑖𝑙𝑜𝑡𝑖 ∗ 𝐷𝑢𝑟𝑖𝑛𝑔𝑡

is 0.098 and significant at the 5% level. To further alleviate potential omitted variable bias

arising from unobserved firm characteristic persistent over time, we employ firm fixed effect in

column (3). The variables 𝑃𝑖𝑙𝑜𝑡𝑖 and 𝐷𝑒𝑙𝑎𝑤𝑎𝑟𝑒𝑖 are omitted due to collinearity. The slope of

𝑃𝑖𝑙𝑜𝑡𝑖 ∗ 𝐷𝑢𝑟𝑖𝑛𝑔𝑡 is 0.094 and significant. This indicates that pilot firms produce a 9.4% lower

annual report readability than non-pilot firms during the pilot program period compared to pre-

16

event period. In sum, the evidence in Table 3 shows that pilot firms produce significantly less

readable annual reports during the Reg SHO period compared to pre-Reg SHO period.7

The SEC eliminated short-sale price tests for all exchange-listed stocks on July 6, 2007

(Securities Exchange Act of 1934 Release No. 34-55970, July 3, 2007). This setting provides us

an alternative approach to testing the relation between short selling and annual report readability.

We next examine whether non-pilot stocks significantly reduce annual report readability during

the post-event period. We run DiD tests using the same group of pilot and non-pilot firms and

retain the sample from May 2005 to June 2010. The regression is as follows:

𝐿𝑜𝑔(𝑓𝑖𝑙𝑒 𝑠𝑖𝑧𝑒𝑖,𝑡) = 𝛼 + 𝛽1 ∗ 𝑁𝑜𝑛𝑝𝑖𝑙𝑜𝑡𝑖 + 𝛽2 ∗ 𝑁𝑜𝑛𝑝𝑖𝑙𝑜𝑡𝑖 ∗ 𝑃𝑜𝑠𝑡𝑡 + 𝑌𝑒𝑎𝑟𝑡 + 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑗 +

𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠 + 𝜀𝑖,𝑡, (2)

where 𝑁𝑜𝑛𝑝𝑖𝑙𝑜𝑡𝑖 is a dummy variable that equals one if a stock is not selected as a pilot stock in

Regulation SHO’s pilot program and zero otherwise. 𝑃𝑜𝑠𝑡𝑡 is a dummy variable that equals one

if the end of a firm’s fiscal year t falls between May 1, 2008 and June 30, 2010 and zero

otherwise. Industry and Year are the industry fixed effects (2-digit SIC codes) and fiscal year

fixed effects dummies, respectively. We also augment equation (2) by replacing the industry

fixed effects with the firm fixed effects. The results are presented in Table 4.

In column (1) of Table 4, the coefficient for 𝑁𝑜𝑛𝑝𝑖𝑙𝑜𝑡𝑖 ∗ 𝑃𝑜𝑠𝑡𝑡 is positive and significant,

indicating that the 10-K files of non-pilot firms are larger than for pilot firms during the post-

7 The results are similar using the unbalanced sample.

17

event period compared to the during-event period. After adding firm characteristics and industry

fixed effects, the coefficient of 𝑁𝑜𝑛𝑝𝑖𝑙𝑜𝑡𝑖 ∗ 𝑃𝑜𝑠𝑡𝑡 is positive and significant at the 5% level in

column (2). The coefficients on 𝑁𝑜𝑛𝑝𝑖𝑙𝑜𝑡𝑖 ∗ 𝑃𝑜𝑠𝑡𝑡 remain positive and significant at 5% with

firm fixed effects in column (3). In terms of economic significance, the evidence in column (3)

indicates that annual report readability is 7.1% lower for the pilot stocks than for the non-pilot

stocks during the post-event period compared to the during-event period. The results in Table 4

further confirm the causal relation between short selling and annual report readability.

So far, our results indicate that the relaxation of short-sale constraints leads to a

significant decrease in pilot firms’ annual report readability. A possible explanation for this

observation is that when faced with short-selling pressure, pilot firms produce less transparent

10-Ks that are more costly for investors to comprehend. Corporate managers’ compensation and

job security are positively related to stock prices, thus they pay considerable attention to the

impact of suspending short-sale price tests on the potential amount of short selling in their firms

(see, e.g., Opinon Research Corporation (2008) and Fang, Huang, and Karpoff (2016)). When

the short-sale constraints are less rigorous, managers can bury earnings-relevant information in

less readable documents that are more costly for investors to comprehend. This reporting

behavior may help reduce the potential amount of short selling for three reasons. First, both naïve

and sophisticated investors are subject to limited attention and information processing power

(e.g., Fang, Peress and Zheng (2014)). Second, investors pay less attention to, place less weight

upon and ignore complicated and hard-to-process information (e.g., Hirshleifer and Teoh (2003);

18

Hirshleifer, Lim and Teoh (2011); Cohen and Lou (2012); and Hirshleifer, Hsu and Li (2017)).

Third, firms exploit the limited attention of investors in various ways (see Daniel, Hirshleifer,

and Teoh (2002) for a review). For example, Hirshleifer and Teoh (2003) argue that owing to

limited attention, firms manage accounting disclosure and reporting choices to manipulate

investors’ perceptions in order to create mispricing. We argue that managers may use annual

report readability to affect investor perception when faced with short-selling pressure.

C. Cross-sectional Analyzes based on Investor Attention and Bad News

In this section, we provide empirical tests to verify the above argument that managers

exploit the limited attention of investors by managing 10-K readability. If this argument is true,

firms that receive less investor attention should be more likely to manipulate annual report

readability when faced with short-selling pressure.

We use institutional ownership, firm size and analyst coverage as proxies for investor

attention. Small stocks, stocks with low analyst coverage and institutional ownership are

regarded as stocks that receive less investor attention. For each measure, we partition the samples

into high and low investor attention subsamples based on their median values each year. We then

repeat the analysis above for Table 3 and report the results in Table 5. Panels A, B and C present

the results for institutional ownership, firm size and analyst coverage, respectively. In each panel,

the results with industry and fiscal year fixed effects are presented in columns (1) and (3) and the

results with firm and fiscal year fixed effects are provided in columns (2) and (4).

19

For the low institutional ownership subsample, the slopes of 𝑃𝑖𝑙𝑜𝑡𝑖 ∗ 𝐷𝑢𝑟𝑖𝑛𝑔𝑡 are

positive and significant in columns (3) and (4) of Panel A. For the high institutional ownership

subsample, the coefficients of 𝑃𝑖𝑙𝑜𝑡𝑖 ∗ 𝐷𝑢𝑟𝑖𝑛𝑔𝑡 are insignificant in columns (1) and (2).

Additionally, in Panel B, the coefficients of 𝑃𝑖𝑙𝑜𝑡𝑖 ∗ 𝐷𝑢𝑟𝑖𝑛𝑔𝑡 are insignificant in columns (1)

and (2) for the large stocks. However, in columns (3) and (4), the coefficients of 𝑃𝑖𝑙𝑜𝑡𝑖 ∗

𝐷𝑢𝑟𝑖𝑛𝑔𝑡 are positive and significant at the 5% level among the small stocks. In Panel C, we find

that the effect of short selling on annual report readability is significant only for the low analyst

coverage group. For the low analyst coverage subsample, in column (3), 𝛽2 is 0.244 and

significant, suggesting that the difference in readability between pilot and non-pilot firms is

24.4%. For the high analyst coverage subsample, 𝛽2 is insignificant in column (1). We find

similar results by adding firm fixed effect in our regression analysis. The coefficient of 𝑃𝑖𝑙𝑜𝑡𝑖 ∗

𝐷𝑢𝑟𝑖𝑛𝑔𝑡is 0.249 and significant in column (4), whereas 𝛽2 is insignificant in column (2). The

evidence in Table 5 shows that the effect of short-selling on annual report readability is

significant only for stocks that receive low investor attention. In sum, these results are in line

with the above investor-attention explanation for the effect of short-selling on annual report

readability.

Firms with good earnings have no incentives to obscure valuation-relevant information.

Therefore, the relation between a reduction in the short-sale constraints and annual report

readability should be more pronounced in firms with bad news to report. We run the DiD

analysis across firms with good and bad news. We define that a firm has bad (good) news to

20

report if its ROA is below (above or equal) the industry median. We show in Table 6 that the

positive relation between short-sale constraints and 10-K file size is significant only in the bad

news subsample. The coefficients on 𝑃𝑖𝑙𝑜𝑡𝑖 ∗ 𝐷𝑢𝑟𝑖𝑛𝑔𝑡 are insignificant in the good news

subsample, whereas these coefficients are positive and significant in the bad news subsample.

For example, 𝛽2 is 0.220 and significant at the 5 % level in column (4) for the bad news

subsample.

In sum, the negative relation between the variation in short-sale constraints and annual

report readability is not uniform in the cross-section. Our analysis shows that the effect of the

relaxation of short-sale constraints on annual report readability is significant only for firms that

receive low investor attention and for firms with bad news to report.

D. Short Selling and Tone Ambiguity

In addition to readability, ambiguous text in 10-Ks can obstruct investors’ ability to

comprehend reports (e.g, LM (2011), LM (2013)). In this subsection, we examine whether pilot

firms increase the use of uncertainty words in 10-Ks to obstruct short sellers’ ability to

comprehend documents. The determinants of the use of the uncertainty tone in 10-Ks are

discussed in Section 1. The DiD analysis of the causal effect of short selling and uncertainty tone

in 10-Ks is conducted in Section 2.

1. Determinants of Uncertainty Tone

21

The following variables can influence the cross-sectional variations in the frequencies of

uncertainty words in 10-Ks.

Firm size. Large firms typically have more complex financial disclosures than small firms. We

hypothesize that large firms have a greater proportion of uncertainty words in 10-Ks than small

firms.

Profitability. Firms with high profitability are less likely to use uncertainty tone in 10-Ks

because of good financial performance. We expect that firms with high ROA are associated with

low frequencies of uncertainty words in annual reports.

Firm age. Mature firms are generally less uncertain than young firms. Therefore, we expect that

mature firms use low proportions of uncertainty words in 10-Ks.

Firm risk. Firms with high risk are more likely to be cautious in disclosure of financial

information in 10-Ks due to uncertainty about future performance. We use stock return volatility

and earnings volatility as proxies for firm risk. We posit a positive relation between risk and

ambiguity tones in annual reports.

Complexity of operations. Complex firms are associated with complex financial disclosures.

Using the numbers of business and geographic segments as proxies for firm complexity, we

expect that more complex firms are associated with higher frequencies of uncertainty words in

10-Ks.

Corporate events. Unusual corporate events may lead to complex disclosures due to high

22

uncertainty. Firms that have unusual events are more likely to use ambiguous text in annual

reports. We use two corporate events: seasoned equity offering and merger and acquisition

activities.

Incorporation state: Firms that are incorporated in Delaware have more investor protection, a

higher corporate valuation, and are more likely to receive takeover bids and be acquired (Daines

(2001)). Thus, firms that are incorporated in Delaware have more complex 10-Ks.

To examine whether above variables impact the uncertainty tone of the text in annual

reports, we regress the proportion of uncertainty words on these variables. The sample period

spans from 1994 to 2015. We present the regression results in Table 7. We find that these

variables have significant explanatory power for the use of uncertainty words in 10-Ks. In

column (1), we find that large firms are associated with a high percentage of uncertainty words.

This indicates that large firms use high frequencies of uncertainty words in annual reports. We

also find that ROA is significantly negatively related to the use of ambiguity tone, suggesting

that firms that are financially strong are less likely to use uncertainty words in their annual

reports. Furthermore, mature firms are less likely to use uncertainty words, as shown by a

statistically significant negative coefficient on Log(age). We also find that riskier firms use a

higher percentage of uncertainty words in their annual reports, as indicated by the statistically

significantly positive coefficients on Ret_vol and Earn_vol. In addition, when we add industry

(firm) and fiscal year fixed effects, the coefficients on the number of geographic segments is

positive and significant in columns (2) and (3), suggesting complex firms are associated complex

23

financial disclosures. However, the slope of the number of business segments is insignificant in

column (3). We also find that firms that are incorporated in Delaware are related to a higher

percentage of uncertainty words in their 10-Ks in column (2). Finally, we find that the SEO and

M&A dummy variables are significantly positively related to the use of uncertainty words when

we add firm and fiscal year fixed effects in our regression.

2. Regulation SHO and Tone Ambiguity in 10-Ks

We next examine how changes in short-sale constraints affect the tone ambiguity of 10-

Ks using DiD regression analysis. The regression is as follows:

𝑈𝑛𝑐𝑒𝑟𝑡𝑎𝑖𝑛𝑖,𝑡 = 𝛼 + 𝛽1 ∗ 𝑃𝑖𝑙𝑜𝑡𝑖 + 𝛽2 ∗ 𝑃𝑖𝑙𝑜𝑡𝑖 ∗ 𝐷𝑢𝑟𝑖𝑛𝑔𝑡 + 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑗 + 𝑌𝑒𝑎𝑟𝑡 + 𝜀𝑖,𝑡, (3)

where 𝑈𝑛𝑐𝑒𝑟𝑡𝑎𝑖𝑛𝑖,𝑡 is the proportion of uncertain words in 10-Ks based on LM (2011) for firm i

at year t. 𝑃𝑖𝑙𝑜𝑡𝑖 is a dummy variable that equals one if a firm is selected as a pilot firm in

Regulation SHO’s pilot program and zero otherwise. 𝐷𝑢𝑟𝑖𝑛𝑔𝑡 is a dummy variable that equals

one if the end of a firm’s fiscal year t falls between May 2005 and June 2007 and zero otherwise.

Industry and Year are the industry fixed effects (2-digit SIC codes) and fiscal year fixed effects

dummies, respectively. We report the regression results in Table 8. We also augment the

equation (3) by adding control variables in column (2) and by adding firm and fiscal year fixed

effects in column (3).

In Table 8, we show that pilot firms significantly increase the proportion of uncertainty

words in 10-Ks during the Reg SHO experiment period. The coefficients on 𝑃𝑖𝑙𝑜𝑡𝑖 ∗ 𝐷𝑢𝑟𝑖𝑛𝑔𝑡

24

are positive and significant in all columns. The DiD estimator is 0.047 and significant at the 1%

level in column (1). This corresponds to approximately 3.5% of the mean of 𝑈𝑛𝑐𝑒𝑟𝑡𝑎𝑖𝑛 in the

pre-Reg SHO period. After controlling firm characteristics, the coefficient 𝑃𝑖𝑙𝑜𝑡𝑖 ∗ 𝐷𝑢𝑟𝑖𝑛𝑔𝑡 is

0.042 and significant at the 1% level in column (2). 𝑃𝑖𝑙𝑜𝑡𝑖 ∗ 𝐷𝑢𝑟𝑖𝑛𝑔𝑡 remains positive and

significant when we add firm and year fixed effects. Our results suggest that pilot firms, whose

short-sale constraints are significantly relaxed due to Reg SHO, not only reduce annual report

readability but also increase tone ambiguity in 10-Ks.

E. Robustness Check

In this subsection, we conduct two robustness tests. We re-run our DiD analyses using

alternative pre- and during-event periods. We also perform two placebo tests to enhance our

causal argument.

1. Alternative Test Periods

In our tests, we define test periods using the actual start and end dates of the Reg SHO

program. To confirm our DiD analysis is robust to alternative pre- and during-event periods, we

run the DiD tests in equations (1) and (3) using the balanced sample. Following Fang, Huang,

and Karpoff (2016), the pre-event period sample includes firms that have data to calculate all key

variables from 2001 to 2003. The during-event period sample contains firms that have data to

calculate all key variables between 2005 and 2007. We exclude 2004 because the SEC

announced the pilot and non-pilot firms for Reg SHO in July 2004. 𝐷𝑢𝑟𝑖𝑛𝑔𝑡 equals one if a

25

firm’s fiscal year end is between January, 2005 and December, 2007. The regression results for

annual report readability and tone ambiguity are presented in Panels A and B of Table 9,

respectively. We find that the coefficients on 𝑃𝑖𝑙𝑜𝑡𝑖 ∗ 𝐷𝑢𝑟𝑖𝑛𝑔𝑡 are positive and statistically

significant in Panels A and B. Our conclusions on the causal effect of short selling on the

readability and tone ambiguity of annual reports are unchanged after using alternative test

periods.

2. Placebo Tests

We next perform two placebo tests for our DiD analysis to strengthen our causal

argument using the balanced sample. We address the concern that our identification tests mainly

rely on the SEC’s approval of Reg SHO that took place in 2004. Unobservable shocks that

occurred prior to 2004 but are unrelated to Reg SHO could have driven results. We use the same

pilot and non-pilot firms identified by Reg SHO but artificially pick a “pseudo-event” year when

we assume a regulatory shock reduced short selling costs. We assume that Reg SHO is effective

from May 2001 to June 2003. We conduct the DiD tests using a balanced sample in Table 10.

The results for annual report readability and tone ambiguity are presented in Panels A and B,

respectively. As can be seen, the coefficients on 𝑃𝑖𝑙𝑜𝑡𝑖 ∗ 𝐷𝑢𝑟𝑖𝑛𝑔𝑡 are all insignificant. This

indicates that the identified impacts of short selling on readability and tone ambiguity in 10-Ks

are unlikely to be driven by unobserved shocks.

26

V. Conclusion

We investigate the causal effects of changes in short-sale constraints on readability and

the tone ambiguity in the context of annual reports. We employ Reg SHO, which relaxes short-

sale constraints for a random sample of pilot stocks during 2005 and 2007, to establish causality.

Using the DiD technique, we find that the relaxation of short-sale constraints leads to a reduction

in annual report readability for pilot firms. Furthermore, this negative relation between the

variation in short-sale constraints and annual report readability is heterogeneous in the cross-

section. The results are more pronounced for firms that receive less investor attention (small

firms, firms with low institutional ownership and analyst coverage) and for firms with worse

news. Additionally, we document that pilot firms use higher frequencies of uncertainty words in

10-Ks during the Reg SHO experiment period.

Our findings indicate that the relaxation of short-sale constraints affects corporate

managers’ reporting behavior by producing less readable and more ambiguous 10-Ks. This

reporting behavior may decrease the potential amount of short selling when investors have

limited attention and processing power. Overall, our study provides important implications to

readers of financial disclosures, such as financial analysts and investors.

27

REFERENCES

Alexander, G.J. and M. A. Peterson. “The effect of price tests on trader behavior and market

quality: An analysis of Reg SHO.” Journal of Financial Markets, 11 (2008), 84-111.

Battalio, R. and P. Schultz. “Options and the bubble.” Journal of Finance, 61 (2006), 2071-2012.

Beber, A. and M. Pagano. “Short-selling bans around the world: Evidence from the 2007-2009

crisis.” Journal of Finance, 68 (2013), 343-381.

Belo, F.; J. Li; X. Lin; and X. Zhao. “Complexity and Information Content of Financial

Disclosures: Evidence from Evolution of Uncertainty Following 10-K Filings.” Working Paper (2016).

Biddle, G. C.; H. Gilles; and S. V. Rodrigo. “How does financial reporting quality relate to

investment efficiency?” Journal of Accounting and Economics,48 (2009), 112-131.

Bloomfield, R. J. “The ‘Incomplete Revelation Hypothesis’ and Financial Reporting.” Accounting

Horizons 16 (2002), 233–243.

Boehmer, R.; B. Danielsen; and S. Sorescu. “Short sale constraints, difference of opinion and

overvaluation.” Journal of Financial and Quantitative Analysis, 41 (2005), 455-487.

Boehmer, E.; C. Jones; and X. Zhang. “Which shorts are informed?” Journal of Finance, 63

(2008), 491-527.

Boehmer, E.; C. Jones; and X. Zhang. “Shackling short sellers: The 2008 shorting ban.” Review

of Financial Studies, 26 (2013), 1363-1400.

Boehmer, E. and J.J. Wu. “Short selling and the price discovery process.” Review of Financial

Studies, 26 (2013), 287-322.

Chu, Y.Q, Hirshleifer, D, and Ma, L., “The causal effect of limits to arbitrage on asset pricing

anomalies”, working paper (2016).

Cohen, L., and D. Lou. 2012. “Complicated firms.” Journal of Financial Economics 104 (2012),

383-400.

Daines, R., Does Delaware law improve firm value? Journal of Financial Economics 62 (2001),

525–558.

Daniel, K.D., Hirshleifer, D., Teoh, S.H. “Investor psychology in capital markets: evidence and

policy implications.” Journal of Monetary Economics 49 (2002), 139–209.

De Angelis, D.; G. Grullon; and S. Michenaud. “The effects of short-selling threats on incentive

contracts: Evidence from a natural experiment.” Working Paper (2014).

Desai, H.; K. Ramesh; S.R. Thiagarajan; and B.V. Balachandran. “An investigation of the

informational role of short interest in the Nasdaq market.” Journal of Finance, 57 (2002): 2263-2287.

28

Doukas, J.A., C.F. Kim, and C. Pantzalis. "Divergence of opinion and equity returns." Journal of

Financial and Quantitative Analysis, 41 (2006): 573-606.

Diether, K. B., K. Lee, and I. M. Werner. "Short-sale strategies and return predictability." Review

of Financial Studies, 22 (2009): 575-607.

Diether, K. B., K. Lee, and I. M. Werner. " It’s SHO Time! Short-Sale Price Tests and Market

Quality." Journal of Finance, 64 (2009): 37-73.

Ertugrul, M.; J. Lei, J. Qiu, and C. Wan. "Annual report readability, tone ambiguity, and the cost

of borrowing." Journal of Financial and Quantitative Analysis, 52 (2017): 811-836.

Fang, L.H., J. Peress, and L. Zheng. “Does Media Coverage of Stocks Affect Mutual Funds'

Trading and Performance?” Review of Financial Studies, 27 (2014): 3441-3466.

Fang, V. W.; A. H. Huang; and J. M. Karpoff. “Short selling and earnings management: A

controlled experiment.” Journal of Finance, 71 (2016), 1251-1294.

Gao, L.; J. He; J. Wu. “Standing out from the Crowd via Corporate Goodness: Evidence from a

Natural Experiment.” Working paper (2016).

Gilchrist, S.; C. Himmelberg; and G. Huberman. “Do stock price bubbles influence corporate

investment?” Journal of Monetary Economics, 52 (2005), 805-827.

Goldstein, I. and A. Guembel. 2008. Manipulation and the allocational role of prices. Review of

Economic Studies 75, 133–164.

Gompers, P.; J. Ishii; and A. Metrick. “Corporate Governance and Equity Prices.” Quarterly

Journal of Economics, 118 (2003), 107-155

Grullon, G.; S. Michenaud; and J. Weston. “The real effects of short-selling constraints.” Review

of Financial Studies, 28 (2015), 1737-1767.

Guay, W.; D. Samuels; and D. Taylor. “Guiding through the fog: Financial statement complexity

and voluntary disclosure.” Journal of Accounting and Economics, 62 (2016), 234-269.

Haruvy, E. and C. N. Noussair. “The effect of short selling on bubbles and crashes in

experimental spot asset markets.” Journal of Finance, 61 (2006), 1119-1157.

Jones, C. M. and O. Lamont. Short-sale constraints and stock returns, Journal of Financial

Economics, 66 (2002), 207-239.

He, J. and X. Tian. “SHO time for innovation: The real effects of short sellers.” Working Paper,

Kelley School of Business Research Paper (2015).

Henry, T. and J. Koski. “Short selling around seasoned equity offerings.” Review of Financial

Studies, 23 (2010), 4389–4418.

29

Hirshleifer, D. and S. H. Teoh. “Limited attention, information disclosure, and financial

reporting”, Journal of Accounting and Economics 36 (2003), 337-386.

Hirshleifer, D., S. Lim, and S. H. Teoh. 2011. Limited investor attention and stock market

misreactions to accounting information. Review of Asset Pricing Studies 1:35–73.

Hirshleifer, D., P. Hsu and D. Li. “Innovative Originality, Profitability, and Stock Returns”,

Review of Financial Studies, 2017, forthcoming.

Hwang, B. and H. H. Kim. "It pays to write well." Journal of Financial Economics, 124 (2017):

373-394.

Karpoff, J. M. and X. Lou. “Short sellers and financial misconduct.” Journal of Finance, 65

(2010), 1879-1913.

Lamont, O. A. “Go down fighting: Short sellers vs. firms”. Review of Asset Pricing Studies, 2

(2012), 1-30.

Lawrence, A. “Individual investors and financial disclosure.” Journal of Accounting and

Economics, 56 (2013), 130-147.

Lehavy, R.; F. Li; and K. Merkley. “The effect of annual report readability on analyst following

and the properties of their earnings forecasts.” The Accounting Review, 8 (2011), 1087-1115.

Li, F. “Annual report readability, current earnings and earnings persistence.” Journal of

Accounting and Economics, 45 (2008), 221-247.

Li, Y. and L. Zhang. “Short selling pressure, stock price behavior and management forecast

precision: Evidence from a natural experiment.” Journal of Accounting Research, 53 (2015), 79–117.

Loughran, T. and B. McDonald. “When Is a Liability Not a Liability? Textual Analysis,

Dictionaries and 10-Ks.” Journal of Finance, 66 (2011), 35–65.

Loughran, T. and B. McDonald. “IPO First-day Returns, Offer Price Revisions, Volatility and

Form S-1 Language.” Journal of Financial Economics, 109 (2013), 307–326.

Loughran, T. and B. McDonald. “Measuring readability in financial disclosures.” Journal of

Finance, 69 (2014), 1643-1671.

Loughran, T. and B. McDonald. “Textual analysis in accounting and finance: A survey.” Journal

of Accounting Research, 54 (2016), 1187-1230.

Massa, M.; B. Zhang; and H. Zhang. “Invisible hand of short selling: Does short-selling

discipline earnings management?” Review of Financial Studies, 28 (2015), 1701-1736.

Miller, E. “Risk, uncertainty and divergence of opinion.” Journal of Finance, 32 (1977), 1151-

1168.

30

Miller, B. P. “The effects of reporting complexity on small and large investor trading.” The

Accounting Review, 85 (2010), 2107-2143.

Opinion Research Corporation, 2008, Short selling study: The views of corporate issuers,

prepared on behalf of NYSE Euronext.

Securities and Exchange Commission. “Economic analysis of the short sale price restrictions

under the regulation SHO pilot.” Office of Economic Analysis (2007).

Senchack, A. J., and L. T. Starks. "Short-sale restrictions and market reaction to short-interest

announcements." Journal of Financial and quantitative analysis, 28 (1993): 177-194.

31

Appendix: Definition of Variables

Experiment-related variables:

Pilot: A dummy variable that equals one if a stock is selected as a pilot stock in Regulation SHO’s pilot

program and zero otherwise.

During: A dummy variable that equals one if the end of a firm’s fiscal year t falls between May 1, 2005

and June 30, 2007 and zero otherwise.

Post: A dummy variable that equals one if the end of a firm’s fiscal year t falls between May 1, 2008 and

June 30, 2010 and zero otherwise.

A firm’s annual report readability and ambiguity tone

10-K file size: Loughran and McDonald (2014) argue that file size of a 10-K is a good proxy for

readability. Larger 10-K file size of a firm is less readable. The readability measure is defined as the

natural logarithm of 10-K document file size in fiscal year t.

The proportion of uncertainty words: Loughran and McDonald (2011) develop a list of uncertainty words

(e.g. approximate, contingency, depend, and uncertain) in financial contexts.

Control variables

Firm size: Larger firms have more complex 10-K reports. The size is defined as the natural logarithm of

the market equity of stocks at the end of fiscal year t.

Firm age: Older companies have more readable annual reports because there is less information

asymmetry and less information uncertainty for these companies. The firm age is the number of years

32

since a company’s first appearance in the CRSP monthly stock return file. We use the natural logarithm of

the firm age in the regressions.

Special items (SI): Firms with a significant amount of special items are more likely to experience some

unusual events. Thus, companies with lower special items have more complex 10-Ks. SI is defined as the

amount of special items scaled by book value of assets.

Volatility of business: Stocks with higher volatility have more complex 10-Ks. To capture the volatility of

business, we use two measures: stock return volatility (Ret_vol, measured as the standard deviation of the

monthly stock returns in the prior year) and earnings volatility (Earn_vol, measured as the standard

deviation of the operating earnings during the prior five fiscal years).

Profitability (ROA): Firms that earn higher profits have more readable 10-Ks. ROA is defined as the

income before extraordinary items divided by lagged total assets.

Complexity of operations: Firms with more complex operations are more likely to have complex 10-Ks.

We use the number of business segments (NBSEG) and the number of geographic segments (NGSEG) to

capture the operation complexity of firms. Log(NBSEG) is the logarithm of 1 plus the number of business

segments and Log(NGSEG) is the logarithm of 1 plus the number of geographic segments.

Financial complexity: Companies with more complex financial situations are more likely to have

complicated 10-Ks. We use the logarithm of the number of non-missing items in Compustat as a proxy

for financial complexity (NITEMS). Firms are more financially complex if they need to report more items

in annual reports.

33

Corporate events: Unusual corporate events may require extra and more detailed disclosures, so firms

with corporate events have more complex 10-Ks. We consider two events: seasoned equity offerings

(SEOs) and merger and acquisitions (MA). The dummy variable SEO is equal to 1 if for a year in which a

company has a common equity offering in the secondary market according to the SDC Global New Issues

database and 0 otherwise. The dummy variable MA is set to 1 for a year in which a company is an

acquirer based on the SDC Platinum M&A database and 0 otherwise.

Delaware: Firms that are incorporated in Delaware have more complex and less readable annual report.

The Delaware dummy variable is equal to 1 if a firm is incorporated in Delware and 0 otherwise.

Investor attention proxies

Institutional ownership (IO): Higher institutional ownership firms receive greater investor attention.

Institutional ownership is defined as the number of shares owned by institutions scaled by the total

number of common shares outstanding. We captured institutional holding data from the Thomson Reuters

13-F database.

Firms Size: Investors pay more attention to large companies.

Analyst coverage: Firms that are covered by more analysts receive more investor attention. The analyst

coverage is defined as the logarithm of the number of analysts following a stock from IBES database.

Bad News: Firms with bad news are more likely to obscure valuation-relevant information. Bad news is

defined as one if ROA is below industry median value and zero otherwise.

34

Table 1: Summary Statistics

This table reports the summary statistics of firm characteristics of the pilot (treatment) and non-pilot (control)

groups measured based on the 2004 Russell 3000 index firms. The sample includes firms whose fiscal year ending

dates are between May 1, 2002 and June 30, 2004 for the pre-event period, between May 1, 2005 and June 30, 2007

for the during-event period, and between May 1, 2008 and June 30, 2010 for the post-event period. We require firms

have data available to calculate firm characteristics and 10-K filing size over time. Definitions of the variables are

provided in the Appendix.

N Mean Median SD

File size (in megabytes) 9,588 1.86 1.45 1.58

Log (file size) 9,588 0.34 0.37 0.76

Uncertainty (%) 9,588 1.48 1.48 0.29

Size (in millions) 9,588 4956.91 915.17 17128.51

BM 9,588 0.60 0.48 0.59

Age 9,588 22.92 17.00 17.06

NBSEG 9,588 2.26 1.00 1.59

NGSEG 9,588 2.76 2.00 2.18

RET_VOL 9,588 0.13 0.11 0.08

EARN_VOL 9,588 0.06 0.03 0.08

ROA 9,588 0.02 0.05 0.17

Non-missing items 9,588 358.85 362.00 27.52

SI 9,588 -0.02 0.00 0.07

SEO 9,588 0.06 0.00 0.23

MA 9,588 0.38 0.00 0.49

Delaware 9,588 0.65 1.00 0.48

35

Table 2: Firm Characteristics before Announcement of Regulation SHO

This table provides the firm characteristics of the pilot (treatment) and non-pilot (control) groups one year before the

announcement of the Regulation SHO (July 2004). The sample comes from the 2004 Russell 3000 index and

contains firms that have data available to calculate readability and control variables. Definitions of variables are in

the Appendix. We report the t-statistics of the two-sample t-test and z-statistics of Wilcoxon rank sum test for the

difference between the pilot and non-pilot groups. ***, **, and * indicate statistical significance at the 1%, 5%, and

10% levels, respectively.

Treatment Control Difference

Mean Median SD Mean Median SD T-stat Wilcoxon

Log(file size) 0.01 0.10 0.76 0.01 0.10 0.74 -0.36 -0.47

Uncertainty 1.34 1.33 0.31 1.39 1.37 0.30 -2.67 -2.55

Log(size) 7.16 6.85 1.48 7.11 6.80 1.47 0.59 0.80

Log(BM) 0.94 0.87 0.66 0.96 0.89 0.65 0.64 0.78

Log(age) 2.98 2.94 0.64 2.89 2.77 0.65 2.11 2.18

Log(numbseg) 1.11 1.10 0.43 1.11 0.69 0.46 -0.06 0.19

Log(numgseg) 1.17 1.10 0.45 1.17 1.10 0.46 0.02 0.20

RET_VOL 0.12 0.10 0.07 0.12 0.10 0.07 -0.09 -0.09

EARN_VOL 0.06 0.03 0.06 0.07 0.03 0.10 -1.97 -0.44

ROA 0.04 0.05 0.12 0.03 0.05 0.15 0.95 -0.06

Log(non-missing

items) 5.80 5.81 0.04 5.80 5.80 0.04 1.24 1.32

SI 0.01 0.00 0.03 0.01 0.00 0.05 -0.02 -0.43

SEO 0.07 0.00 0.26 0.07 0.00 0.26 -0.03 0.04

MA 0.39 0.00 0.49 0.38 0.00 0.49 0.16 0.16

Delaware 0.62 1.00 0.49 0.61 1.00 0.49 0.25 0.25

36

Table 3: Multivariate Difference-in-Differences Tests: Annual Report Readability and Regulation SHO

This table presents the results of DiD tests examining how the relaxation of short-sale constraints affects annual

report readability using a balanced panel. The sample comes from the 2004 Russell 3000 index and contains firms

that have data available to calculate readability and control variables over the pre-event (fiscal year ending date is

between May 2002 and June 2004) and during-event (fiscal year ending date is between May 2005 and June 2007)

periods. Column (1) reports the results of the following regression:

𝐿𝑜𝑔(𝑓𝑖𝑙𝑒 𝑠𝑖𝑧𝑒𝑖,𝑡) = 𝛼 + 𝛽1 ∗ 𝑝𝑖𝑙𝑜𝑡𝑖 + 𝛽2 ∗ 𝑝𝑖𝑙𝑜𝑡𝑖 ∗ 𝑑𝑢𝑟𝑖𝑛𝑔𝑡 + 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑗 + 𝑌𝑒𝑎𝑟𝑡 + 𝜀𝑖,𝑡,

where 𝐿𝑜𝑔(𝑓𝑖𝑙𝑒 𝑠𝑖𝑧𝑒𝑖,𝑡) is the natural logarithm of 10-K document file size for firm i at year t. 𝑃𝑖𝑙𝑜𝑡𝑖 is a dummy

variable that equals one if a stock is selected as a pilot stock in Regulation SHO’s pilot program and zero otherwise.

𝐷𝑢𝑟𝑖𝑛𝑔𝑡 is a dummy variable that equals one if the end of a firm’s fiscal year t falls between May 2005 and June

2007 and zero otherwise. Industry and Year are the industry fixed effects (2-digit SIC codes) and fiscal year fixed

effects, respectively. We omit 𝑑𝑢𝑟𝑖𝑛𝑔𝑡 to avoid multicollinearity. We add control variables to the regression and

provide the results with industry and year fixed effects in column (2), and with firm and year fixed effects in column

(3). Variable definitions are provided in the Appendix. Standard errors clustered by firms are displayed in

parentheses. ***, ** and * indicate significance at the 1, 5 and 10 percent levels, respectively.

37

Variables (1) (2) (3)

Pilot -0.048 -0.049

(0.043) (0.040)

Pilot*During 0.116*** 0.098** 0.094**

(0.042) (0.042) (0.047)

Log(size)

0.144*** 0.073*

(0.013) (0.038)

Log(numbseg)

0.106*** 0.148**

(0.034) (0.068)

Log(numgseg)

0.005 0.011

(0.035) (0.067)

Log(BM)

0.111*** 0.043

(0.024) (0.037)

Earn_vol

-0.053 0.357

(0.197) (0.294)

SI

0.136 0.320

(0.195) (0.253)

Ret_vol

1.333*** 0.801***

(0.219) (0.250)

ROA

-0.423*** -0.438***

(0.122) (0.158)

Log (age)

-0.058** -0.313

(0.028) (0.206)

Log (non-missing items)

1.890*** 1.255***

(0.426) (0.478)

SEO

0.018 0.004

(0.043) (0.043)

MA

0.043* 0.004

(0.024) (0.023)

Delaware

0.046

(0.033)

Observations 4,853 4,853 4,853

R-squared 0.192 0.275 0.676

Industry FE YES YES NO

Firm FE NO NO YES

Year FE YES YES YES

38

Table 4: Multivariate Difference-in-Differences Test: During and Post Regulation SHO

This table presents the results examining the effect of short-selling on annual report readability after the Regulation

SHO period using a balanced panel. The sample comes from the 2004 Russell 3000 index and contains firms that

have data available to obtain readability and controls over the during-event (fiscal year ending date is between May

2005 and June 2007) and post-event (fiscal year ending date is between May 2008 and June 2010) periods. The

regression in column (1) is as follows:

𝐿𝑜𝑔(𝑓𝑖𝑙𝑒 𝑠𝑖𝑧𝑒𝑖,𝑡) = 𝛼 + 𝛽1 ∗ 𝑛𝑜𝑛𝑝𝑖𝑙𝑜𝑡𝑖 + 𝛽2 ∗ 𝑛𝑜𝑛𝑝𝑖𝑙𝑜𝑡𝑖 ∗ 𝑝𝑜𝑠𝑡𝑡 + 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑗 + 𝑌𝑒𝑎𝑟𝑡 + 𝜀𝑖,𝑡

where 𝐿𝑜𝑔(𝑓𝑖𝑙𝑒 𝑠𝑖𝑧𝑒𝑖,𝑡) is the natural logarithm of 10-K document file size for firm i at year t. 𝑛𝑜𝑛𝑝𝑖𝑙𝑜𝑡𝑖 is a

dummy variable that equals one if a stock is not selected as a pilot stock in Regulation SHO’s pilot program and

zero otherwise. 𝑃𝑜𝑠𝑡𝑡 is a dummy variable that equals one if the end of a firm’s fiscal year t falls between May 2008

and June 2010 and zero otherwise. We omit post to avoid multicollinearity. Industry and Year are the industry fixed

effects (2-digit SIC codes) and fiscal year fixed effects, respectively. Column (2) presents the results adding control

variables to the regression with industry and year fixed effects. Column (3) reports the results adding firm and year

fixed effects. Variable definitions are provided in the Appendix. Standard errors clustered by firms are displayed in

parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

39

(1) (2) (3)

Nonpilot -0.077** -0.060*

(0.034) (0.033)

Nonpilot*Post 0.077** 0.067** 0.071**

(0.031) (0.031) (0.034)

Log(size)

0.142*** 0.082***

(0.011) (0.026)

Log(numbseg)

0.112*** 0.047

(0.030) (0.052)

Log(numgseg)

0.015 -0.040

(0.029) (0.062)

Log(BM)

0.068*** 0.060**

(0.020) (0.027)

Earn_vol

0.290 0.043

(0.185) (0.211)

SI

0.045 0.028

(0.127) (0.129)

Ret_vol

0.507*** -0.049

(0.153) (0.145)

ROA

-0.354*** -0.058

(0.128) (0.130)

Log (age)

-0.004 -1.153***

(0.025) (0.175)

Log (non-missing items)

1.890*** 0.885**

(0.356) (0.442)

SEO

0.050 0.005

(0.042) (0.048)

MA

0.039* 0.018

(0.021) (0.020)

Delaware

0.023

(0.030)

Observations 4,828 4,828 4,828

R-squared 0.132 0.251 0.690

Industry FE YES YES NO

Firm FE NO NO YES

Year FE YES YES YES

40

Table 5: Annual Report Readability and Regulation SHO: Sample Partitioned by Investor Attention

This table describes how investor attention impacts the effect of short-selling on annual report readability. The

sample contains the 2004 Russell 3000 firms that have data available to calculate readability and controls over the

pre-event and during-event periods. Panels A, B and C report the results for institutional ownership, firm size, and

analyst coverage, respectively. Small stocks, stocks with low analyst coverage and institutional ownership are

regarded as stocks that receive less investor attention. For each proxy, we partition the sample into high and low

short-sale constraint subsamples based on its median values each year and then we repeat the DiD tests in Table 3

across each subsample. Variable definitions are provided in the Appendix. Standard errors clustered by firms are

displayed in parentheses. ***, ** and * indicate statistical significance at the 1, 5 and 10 percent levels, respectively.

Panel A: Institutional Ownership

High IO Low IO

(1) (2) (3) (4)

Pilot 0.013 -0.111**

(0.055) (0.055)

Pilot*During 0.047 0.058 0.158** 0.170**

(0.057) (0.072) (0.062) (0.080)

Log(size) 0.131*** 0.054 0.140*** 0.100*

(0.021) (0.069) (0.015) (0.057)

Log(numbseg) 0.080* 0.161* 0.140*** 0.160

(0.044) (0.097) (0.049) (0.117)

Log(numgseg) -0.010 -0.070 0.016 0.111

(0.047) (0.106) (0.046) (0.091)

Log(BM) 0.083** 0.028 0.109*** 0.087

(0.034) (0.063) (0.033) (0.056)

Earn_vol 0.085 0.297 -0.038 0.476

(0.346) (0.588) (0.230) (0.432)

SI 0.367 0.559 0.206 0.246

(0.300) (0.436) (0.261) (0.377)

Ret_vol 1.445*** 1.097** 1.289*** 0.594*

(0.357) (0.445) (0.285) (0.348)

ROA -0.626*** -0.552 -0.399*** -0.463**

(0.229) (0.337) (0.139) (0.197)

Log (age) -0.033 -0.081 -0.056 -0.718**

(0.037) (0.328) (0.041) (0.342)

Log (non-missing items) 1.562** 1.478** 2.232*** 1.270

(0.623) (0.732) (0.537) (0.786)

SEO -0.004 -0.006 0.031 -0.030

(0.057) (0.067) (0.067) (0.069)

MA 0.038 -0.009 0.051 0.032

(0.032) (0.036) (0.034) (0.036)

Delaware 0.049 0.060

(0.045) (0.044)

Observations 2,419 2,419 2,419 2,419

R-squared 0.251 0.703 0.322 0.735

Industry FE YES NO YES NO

Firm FE NO YES NO YES

Year FE YES YES YES YES

41

Panel B: Firm Size

Large Stocks Small Stocks

(1) (2) (3) (4)

Pilot -0.046 -0.054

(0.057) (0.056)

Pilot*During 0.068 0.042 0.147** 0.159**

(0.056) (0.065) (0.063) (0.077)

Log(numbseg) 0.102** 0.081 0.186*** 0.199

(0.045) (0.092) (0.051) (0.128)

Log(numgseg) 0.060 0.031 0.018 0.015

(0.049) (0.104) (0.049) (0.098)

Log(BM) 0.011 0.037 0.073** 0.015

(0.034) (0.051) (0.029) (0.039)

Earn_vol 0.002 0.493 -0.223 0.203

(0.393) (0.419) (0.210) (0.428)

SI 0.356 -0.222 -0.018 0.499

(0.316) (0.387) (0.270) (0.333)

Ret_vol 0.469 0.182 1.402*** 1.066***

(0.383) (0.462) (0.266) (0.317)

ROA -0.796*** 0.117 -0.086 -0.431**

(0.277) (0.306) (0.128) (0.177)

Log (age) 0.043 -0.284 -0.126*** -0.579*

(0.035) (0.287) (0.046) (0.336)

Log (non-missing items) 1.945*** 0.958 1.899*** 0.800

(0.603) (0.703) (0.610) (0.739)

SEO -0.020 -0.016 0.030 0.014

(0.071) (0.067) (0.056) (0.064)

MA 0.080*** -0.022 0.062* 0.018

(0.031) (0.032) (0.036) (0.037)

Delaware 0.037 0.072

(0.048) (0.044)

Observations 2,429 2,429 2,424 2,424

R-squared 0.250 0.704 0.269 0.681

Industry FE YES NO YES NO

Firm FE NO YES NO YES

Year FE YES YES YES YES

42

Panel C: Analyst Coverage

High Analysts Low Analysts

VARIABLES (1) (2) (3) (4)

Pilot -0.017 -0.125**

(0.053) (0.060)

Pilot*During -0.008 -0.005 0.244*** 0.249***

(0.055) (0.067) (0.066) (0.087)

Log(size) 0.111*** 0.038 0.145*** 0.056

(0.017) (0.066) (0.024) (0.065)

Log(numbseg) 0.046 0.037 0.167*** 0.139

(0.043) (0.100) (0.053) (0.128)

Log(numgseg) -0.056 0.029 0.062 -0.031

(0.046) (0.108) (0.050) (0.127)

Log(BM) 0.069** 0.053 0.128*** 0.093

(0.031) (0.066) (0.039) (0.062)

Earn_vol -0.111 0.233 0.102 0.578

(0.263) (0.422) (0.293) (0.622)

SI 0.316 0.638 0.189 0.375

(0.302) (0.477) (0.282) (0.359)

Ret_vol 0.782** 0.910** 1.592*** 1.094***

(0.335) (0.442) (0.307) (0.401)

ROA -0.636*** -0.433 -0.348** -0.549**

(0.188) (0.308) (0.172) (0.230)

Log (age) -0.004 -0.274 -0.068 -0.490

(0.038) (0.353) (0.044) (0.420)

Log (non-missing items) 1.972*** 1.162 1.559*** 0.615

(0.582) (0.771) (0.596) (0.880)

SEO -0.021 -0.011 -0.005 -0.010

(0.064) (0.068) (0.067) (0.075)

MA 0.004 -0.049 0.064* 0.038

(0.032) (0.035) (0.036) (0.043)

Delaware 0.027 0.081*

(0.043) (0.046)

Observations 2,401 2,401 2,109 2,109

R-squared 0.258 0.719 0.300 0.736

Industry FE YES NO YES NO

Firm FE NO YES NO YES

Year FE YES YES YES YES

43

Table 6: Annual Report Readability and Regulation SHO: Sample Partitioned by Bad News

This table tests whether the relation between short-selling and annual report readability is uniform across bad and

good news firms. The sample comes from the 2004 Russell 3000 index and contains firms that have data available to

calculate readability and control variables over the pre-event and during-event periods. If a firm’s ROA is below

(above or equal) the industry median, we define this firm has a bad (good) news. We repeat the DiD tests in Table 3

across different subsamples. Variable definitions are provided in the Appendix. Standard errors clustered by firms

are displayed in parentheses. ***, ** and * indicate statistical significance at the 1, 5 and 10 percent levels,

respectively.

Good News Bad News

VARIABLES (1) (2) (3) (4)

Pilot -0.006 -0.087

(0.050) (0.071)

Pilot*During 0.072 0.058 0.190** 0.220**

(0.055) (0.066) (0.076) (0.103)

Log(size) 0.112*** 0.048 0.120*** 0.014

(0.017) (0.064) (0.021) (0.070)

Log(numbseg) 0.104** 0.047 0.168*** 0.339**

(0.043) (0.102) (0.059) (0.138)

Log(numgseg) 0.038 0.100 -0.001 -0.083

(0.046) (0.094) (0.064) (0.160)

Log(BM) 0.075** 0.023 0.055 0.008

(0.037) (0.067) (0.040) (0.060)

Earn_vol -0.004 0.052 0.018 0.163

(0.286) (0.390) (0.307) (0.659)

SI -0.976* -0.122 -0.345* -0.104

(0.507) (0.571) (0.197) (0.263)

Ret_vol 1.653*** 1.205*** 0.532 -0.116

(0.304) (0.395) (0.366) (0.438)

Log (age) -0.057 -0.247 -0.078 -0.418

(0.038) (0.311) (0.051) (0.438)

Log (non-missing items) 2.168*** 1.680** 1.418** 0.500

(0.557) (0.671) (0.635) (0.995)

SEO -0.009 0.054 0.027 -0.040

(0.057) (0.068) (0.063) (0.072)

MA 0.024 -0.010 0.043 0.060

(0.029) (0.034) (0.037) (0.045)

Delaware 0.031 0.076

(0.046) (0.054)

Observations 3,186 3,186 1,667 1,667

R-squared 0.407 0.733 0.458 0.759

Industry FE YES NO YES NO

Firm FE NO YES NO YES

Year FE YES YES YES YES

44

Table 7: Determinants of Tone Ambiguity in Annual Reports

This table reports the regression results of tone ambiguity in annual reports on potential determinants. The

dependent variable is the proportion of uncertainty words defined by Loughran and McDonald (2011). The

independent variables include firm size, ROA, firm age, return volatility, earnings volatility, the numbers of

business and geographic segments, a seasoned equity offer dummy variable, and a merger and acquisition dummy

variable. The sample period spans from 1994 to 2015. Variable definitions are provided in the Appendix. We use

industry fixed effects (2-digit SIC codes) and fiscal year fixed effects in column (2) and employ firm fixed effect

and fiscal year fixed effect in column (3). Standard errors clustered by firms are displayed in parentheses. ***, **

and * indicate statistical significance at the 1, 5 and 10 percent levels, respectively.

(1) (2) (3)

VARIABLES uncertainty uncertainty uncertainty

Log(size) 0.041*** 0.019*** 0.022***

(0.002) (0.001) (0.002)

ROA -0.140*** -0.062*** -0.035***

(0.015) (0.012) (0.010)

Log(age) -0.049*** -0.066*** -0.096***

(0.005) (0.004) (0.013)

Ret_vol 0.216*** 0.176*** 0.061***

(0.023) (0.021) (0.016)

Earn_vol 0.271*** 0.068** 0.043

(0.030) (0.027) (0.027)

Log(numbseg) -0.018** -0.057*** -0.010

(0.007) (0.006) (0.006)

Log(numgseg) 0.012 0.015** 0.021***

(0.008) (0.007) (0.007)

SEO 0.057*** 0.035*** 0.014***

(0.007) (0.005) (0.004)

MA -0.052*** 0.007** 0.006***

(0.004) (0.003) (0.002)

Delaware 0.040*** 0.022***

(0.007) (0.006)

Observations 73,987 73,987 73,987

R-squared 0.069 0.492 0.793

Industry FE NO YES NO

Firm FE NO NO YES

Year FE NO YES YES

45

Table 8: Multivariate Difference-in-Differences Tests: Tone Ambiguity and Regulation SHO

This table provides the results of multivariate DiD tests on the impact of the relaxation of short-sale constraints on

tone ambiguity in annual reports. We run the following OLS regressions:

𝑈𝑛𝑐𝑒𝑟𝑡𝑎𝑖𝑛𝑖,𝑡 = 𝛼 + 𝛽1 ∗ 𝑝𝑖𝑙𝑜𝑡𝑖 + 𝛽2 ∗ 𝑝𝑖𝑙𝑜𝑡𝑖 ∗ 𝑑𝑢𝑟𝑖𝑛𝑔𝑡 + 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑗 + 𝑌𝑒𝑎𝑟𝑡 + 𝜀𝑖,𝑡

where 𝑈𝑛𝑐𝑒𝑟𝑡𝑎𝑖𝑛𝑖,𝑡is the proportion of uncertain words in 10-Ks based on Loughran and McDonald (2011) for firm

i at year t. 𝑃𝑖𝑙𝑜𝑡𝑖 is a dummy variable that equals one if a stock is selected as a pilot stock in Regulation SHO’s pilot

program and zero otherwise. 𝐷𝑢𝑟𝑖𝑛𝑔𝑡 is a dummy variable that equals one if the end of a firm’s fiscal year t falls

between May 2005 and June 2007 and zero otherwise. Industry and Year are the industry fixed effects (2 digits SIC

codes) and fiscal year fixed effects, respectively. We replace industry fixed effect with firm fixed effect in column

(3). The sample comes from the 2004 Russell 3000 index and contains firms that have data available to calculate

𝑈𝑛𝑐𝑒𝑟𝑡𝑎𝑖𝑛𝑖,𝑡 and controls over the pre-event (fiscal year ending date is between May 2002 and June 2004) and

during-event periods. Variable definitions are provided in the Appendix. Standard errors clustered by firms are

displayed in parentheses. ***, ** and * indicate statistical significance at the 1, 5 and 10 percent levels, respectively.

(1) (2) (3)

VARIABLES uncertainty uncertainty uncertainty

Pilot -0.045*** -0.036**

(0.017) (0.016)

Pilot*During 0.047*** 0.042*** 0.032**

(0.013) (0.013) (0.014)

Log(size)

0.016*** 0.026***

(0.005) (0.009)

ROA

0.062 -0.067*

(0.048) (0.039)

Log(age)

-0.094*** -0.188***

(0.012) (0.060)

Ret_vol

0.521*** 0.228***

(0.096) (0.073)

Earn_vol

0.334*** 0.191*

(0.097) (0.098)

Log(numbseg)

-0.064*** -0.036*

(0.016) (0.021)

Log(numgseg)

-0.008 0.041**

(0.015) (0.020)

SEO

0.060*** 0.031**

(0.020) (0.015)

MA

0.003 0.016**

(0.009) (0.007)

Delaware

0.026*

(0.014)

Observations 4,853 4,853 4,853

R-squared 0.140 0.228 0.775

Industry FE YES YES NO

Firm FE NO NO YES

Year FE YES YES YES

46

Table 9: Multivariate Difference-in-Differences Tests: Alternative Test Periods

This table reports the multivariate DiD test results on how the relaxation of short-sale constraints affects annual

report readability and tone ambiguity using alternative experiment periods. The sample comes from the 2004 Russell

3000 index and contains firms that have data available to calculate firm characteristics over pre-event period (fiscal

year ending date is between January 1, 2001 and December 31, 2003) and during-event period (fiscal year ending

date is between January 1, 2005 and December 31, 2007). Variable definitions are provided in the Appendix.

Standard errors clustered by firms are in parentheses. ***, ** and * indicate statistical significance at the 1, 5 and 10

percent levels, respectively.

Panel A: annual report readability

(1) (2) (3)

Pilot -0.047 -0.041

(0.040) (0.036)

Pilot*During 0.115*** 0.095** 0.099**

(0.040) (0.040) (0.043)

Log(size)

0.129*** 0.065**

(0.012) (0.031)

Log(numbseg)

0.083** 0.071

(0.035) (0.056)

Log(numgseg)

0.052 0.093*

(0.035) (0.055)

Log(BM)

0.111*** 0.046

(0.022) (0.032)

Earn_vol

-0.198 0.059

(0.205) (0.273)

SI

-0.043 -0.100

(0.235) (0.234)

Ret_vol

0.853*** 0.570***

(0.192) (0.204)

ROA

-0.521*** -0.329*

(0.129) (0.169)

Log (age)

-0.036 0.061

(0.028) (0.162)

Log (non-missing items)

1.554*** 0.857**

(0.381) (0.411)

SEO

-0.024 -0.041

(0.043) (0.042)

MA

0.077*** 0.035*

(0.022) (0.021)

Delaware

0.091***

(0.032)

Observations 5,468 5,468 5,468

R-squared 0.385 0.449 0.696

Industry FE YES YES NO

Firm FE NO NO YES

Year FE YES YES YES

47

Panel B: tone ambiguity

VARIABLES (1) (2) (3)

Pilot -0.043** -0.031*

(0.017) (0.016)

Pilot*During 0.050*** 0.041*** 0.042***

(0.015) (0.014) (0.015)

Log(size)

0.017*** 0.026***

(0.006) (0.008)

ROA

0.015 -0.072

(0.053) (0.045)

Log(age)

-0.100*** -0.233***

(0.013) (0.059)

Ret_vol

0.642*** 0.247***

(0.084) (0.063)

Earn_vol

0.254** 0.085

(0.119) (0.123)

Log(numbseg)

-0.053*** -0.033

(0.017) (0.021)

Log(numgseg)

-0.012 0.050**

(0.017) (0.021)

SEO

0.050*** 0.024*

(0.019) (0.014)

MA

-0.002 0.012*

(0.009) (0.007)

Delaware

0.033**

(0.015)

Observations 5,468 5,468 5,468

R-squared 0.203 0.289 0.738

Industry FE YES YES NO

Firm FE NO NO YES

Year FE YES YES YES

48

Table 10 Two Placebo Tests

This table reports the results of two placebo tests. We assume that the Regulation SHO is effective from May, 2002

to June, 2004. We repeat the DiD tests using the same set of pilot and non-pilot stocks in Panels A and B using a

balanced panel. The sample comes from the 2004 Russell 3000 index and contains firms that have data available to

calculate firm characteristics over pre-event period (fiscal year ending date is between May, 1999 and June, 2001)

and during-event period (fiscal year ending date is between May, 2002 and June, 2004). Variable definitions are

provided in the Appendix. Standard errors clustered by firms are displayed in parentheses. ***, ** and * indicate

statistical significance at the 1, 5 and 10 percent levels, respectively.

Panel A: annual report readability

Variables (1) (2) (3)

Pilot -0.005 -0.007

(0.034) (0.031)

Pilot*During -0.048 -0.051 -0.053

(0.038) (0.038) (0.046)

Log(size)

0.133*** 0.004

(0.011) (0.037)

Log(numbseg)

0.091*** 0.079

(0.033) (0.063)

Log(numgseg)

0.062* 0.013

(0.033) (0.078)

Log(BM)

0.132*** 0.021

(0.020) (0.038)

Earn_vol

0.003 -0.274

(0.210) (0.439)

SI

-0.107 0.033

(0.234) (0.273)

Ret_vol

0.914*** 0.034

(0.164) (0.177)

ROA

-0.623*** -0.397**

(0.122) (0.176)

Log (age)

-0.056** 0.032

(0.023) (0.231)

Log (non-missing items)

1.022*** -0.324

(0.323) (0.421)

SEO

-0.001 -0.031

(0.039) (0.044)

MA

0.075*** 0.058**

(0.023) (0.024)

Delaware

0.078***

(0.029)

Observations 4,984 4,984 4,984

R-squared 0.242 0.327 0.691

Industry FE YES YES NO

Firm FE NO NO Yes

Year FE YES YES YES

49

Panel B: tone ambiguity

VARIABLES (1) (2) (3)

Pilot -0.021 -0.017

(0.019) (0.017)

Pilot*During -0.000 0.006 -0.005

(0.014) (0.014) (0.014)

Log(size)

0.023*** 0.014

(0.005) (0.009)

ROA

-0.059 -0.055

(0.055) (0.044)

Log(age)

-0.096*** -0.056

(0.012) (0.074)

Ret_vol

0.616*** 0.111**

(0.070) (0.055)

Earn_vol

0.449*** 0.205

(0.116) (0.153)

Log(numbseg)

-0.057*** 0.004

(0.016) (0.019)

Log(numgseg)

-0.008 0.017

(0.018) (0.027)

SEO

0.053*** 0.009

(0.019) (0.016)

MA

-0.005 0.008

(0.010) (0.007)

Delaware

0.035**

(0.015)

Observations 4,984 4,984 4,984

R-squared 0.160 0.278 0.826

Industry FE YES YES NO

Firm FE NO NO Yes

Year FE YES YES YES