Credit Rating Changes and Stock Market Reaction: The Impact of

Investor Sentiment

Nikolaos Karampatsas, Soheila Malekpour, and Andrew Mason

This draft: January 13, 2019

Abstract

Using a unique database this study establishes a relationship between firm-specific investor
sentiment and stock price movements around credit rating changes announcements. We find
that firm-specific investor sentiment is a key determinant of price adjustment in the context
of rating downgrades. We also find that the effect of firm-specific investor sentiment is more
pronounced for speculative-grade firms and direct rating downgrades.

Keywords: Investor sentiment, Credit ratings announcements, Social media, Twitter and

StockTwits.

JEL: G12, G14, G24

Nikolaos Karampatsas is from Surrey Business School, University of Surrey, UK, E-mail:
n.karampatsas@surrey.ac.uk. Soheila Malekpour is from Surrey Business School, University of Surrey,
UK, E-mail:s.malekpourkolbadinejad@surrey.ac.uk. Andrew Mason (corresponding author) is from
Surrey Business School, University of Surrey, UK, E-mail: andrew.mason@surrey.ac.uk. We thank
James Crane-Baker at PsychSignal for providing the social media sentiment data used in this study. All
errors and omissions remain the responsibility of the authors.
Introduction

Credit rating agencies have provided information upon which financial transactions

are based for more than a century and despite periodic bouts of criticism credit ratings are

still the first reference point for many investors, risk managers, and other financial agents.1

The main credit rating agencies, Moody’s, Standard & Poor’s and Fitch provide assessments

of the credit worthiness of firms around the globe; from their long-term credit ratings to their

medium term reviews and short term watch lists.

The literature on the effects of credit rating changes on the stock market fall into the

following broad categories. The assymetric nature of price responses to rating changes, the

credibility and accuracy of ratings asigned to firms and their financial instruments, the

attempts by the credit rating agencies to improve the timeliness of information

dissemination, and the differential impact of rating changes on firms with investment-grade

and speculative-grade debt. Our paper contributes to the literature on the information

content of credit rating agencies’ rating changes and their impact on stock returns by

incorporating the impact of investor sentiment.

We present three main findings. First, our empirical analysis confirms that rating

downgrades are more important than rating upgrades, and also find that the effect of investor

sentiment on stock returns is most pronounced in the presence of rating downgrades.

Second, we confirm the importance of the rating level assigned to a firm’s debt (i.e.

investment-grade or speculative-grade), and find that the effect of firm-specific investor

1
In the wake of the 2007-8 Financial Crisis the integrity of the rating agencies was severely undermined.

1
sentiment on stock returns is most marked for speculative-grade firms. Third, our analysis of

the informational content of credit watch reviews shows that the effect of firm-specific

investor sentiment on stock returns is most marked for firms where the downgrades come

largely as a surprise rather than those that are anticipated. These findings extend and modify

the findings of prior studies.

The remainder of the paper proceeds as follows. The next section reviews the

literature and develops our hypotheses. Then, we describe our sample and methodology.

After the section devoted to the empirical analysis, we provide the conclusion.

Literature Review and Hypotheses Development

Credit Rating Changes and Stock Market Reaction

The link between credit rating changes and stock prices is well established in the

literature and the consensus is that the market response is asymmetric and conditional on

the event type, with rating downgrades having a greater impact than rating upgrades on the

stock prices of firms (Holthausen and Leftwich, 1986; Hand et al., 1992; Followill and Martell,

1998; Bedendo et al., 2018). This result has been explained by He et al. (2011) based on

information asymmetry considerations. Galil and Soffer (2011) offer another explantaion,

that good news is released to the market promptly but the release of bad news is delayed and

is often clustered, by which they mean that bad news tends to be highlighted by several

sources at the same time. This bad news when released via a rating downgrade or appearance

on a watch-list prior to downgrade is treated as new information and is priced in the market.

Additionally, related studies find that the stock market response to rating changes is

2
asymmetric and conditonal on the prior rating level, with much stronger effects for low-rated

relative to high-rated firms (Jorion and Zhang 2007; 2010). This result can be explained based

on uncertainty and information asymmetry considerations too.

Watch-Preceded and Direct Rating Changes

The different impact of watch-preceded rating changes relative to direct rating

changes has long been considered in the literature, from early tentative consideration in

Holthausen and Leftwich (1986) and Hand et al. (1992) up to more recent studies such as

Chan et al. (2011), Chung et al. (2012), and Kiesel and Kolaric (2018).2 These studies find that

direct rating downgrades lead to more significant market responses than watch-preceded

downgrades. Holthausen and Leftwich (1986) and Chung et al. (2012) find significant

abnormal stock returns when a firm is added to a rating watch-list whilst Kiesel and Kolaric

(2018) find a widening of CDS spreads when a firm is added to a watch-list.3 Bannier and

Hirsch (2010) consider the impact of watch-preceded and direct rating changes and find that

the stock market reacts more strongly to direct rating downgrades than watch-preceded

downgrades. Purda (2007) argues that there is no significant difference in the stock market

reaction to watch-preceded relative to direct rating changes, and concludes that although

credit watch additions are signficant events they do not pre-empt the information found in

direct rating changes.

2
A direct rating change is a rating change which is not preceded by addition to a watch list or being placed under
rating review. It is therefore an unanticipated or surprise rating change
3
Firms may have their rating under review for 12-18 months but a firm generally stays on a watch-list for no
more than 90 days.

3
Investor Sentiment and Social Media

The recent literature on investor sentiment and social media has established the

importance of new information dissemination technologies for financial markets (Chen et al.,

2014; Da et al., 2015). Some of these studies have moved beyond general measures of

sentiment such as the widely adopted Baker and Wurgler’s (2006) measure of investor

sentiment (Li and Luo, 2017) or general measures of mood such as Facebook’s Gross National

Happiness Index (Danbolt et al., 2015), and focussed on firm-specific measures of sentiment

(Karampatsas et al., 2018).

Research has generally concluded that investor sentiment can lead to mispricing as

investors may be more optimistic or pessimistic than warranted by fundamental information

(Mian and Sankaraguswamy, 2012). Many studies find the impact of investor sentiment to be

heterogeneous and more pronounced for stocks that are subject to uncertainty, and limits to

arbitrage (Baker and Wurgler, 2006; Joseph et al., 2011; Mian and Sankaraguruswamy, 2012;

Li and Luo, 2017). This is also borne out by studies focusing around a corporate event such as

earnings announcements (Karampatsas et al., 2018). Our study extends research in this area

by analysing the impact of firm-specific sentiment on another corporate event credit rating

changes.

Hypotheses Development

The credit ratings literature reveals that the stock market response to rating changes

is asymmetric, with rating downgrades having a greater impact than rating upgrades on stock

4
returns. The behavioural literature and prospect theory (Kahneman and Tversky, 1979) argue

that investors are loss-averse, and their response to to news is asymmetric, with bad news

having a greater impact than good news. We combine and build on these two strands of

literature and form the following hypothesis:

Hypothesis 1: The impact of investor sentiment on the stock market response to rating

changes should be more pronounced for rating downgrades.

The credit ratings literature concludes that the stock market response to rating

changes is asymmetric, with much stronger effects for low-rated relative to high-rated firms

(Jorion and Zhang 2007; 2010; He et al., 2011) due to uncertainty and information asymmetry

considerations. The behavioural literature argues that the impact of investor sentiment is

more pronounced for firms that are subject to uncertainty, and limits to arbitrage (Baker and

Wurgler, 2006; Joseph et al., 2011; Mian and Sankaraguruswamy, 2012; Li and Luo, 2017).

We build on and combine these two strands of literature and form the following hypothesis:

Hypothesis 2: The impact of investor sentiment on the stock market response to rating

downgrades should be more pronounced for low-rated firms.

The credit ratings literature concludes that the stock market response to rating

changes is asymmetric, with direct rating changes having stronger effects than watch-

5
preceded rating changes and this is due to uncertainty and information asymmetry

considerations (Holthausen and Leftwich, 1986; Chung et al., 2012). The behavioural

literature argues that the impact of investor sentiment is more pronounced for firms that are

subject to uncertainty and information asymmetry. We build on and combine these two

strands of literature and form the following hypothesis:

Hypothesis 3: The impact of investor sentiment on the stock market response to rating

downgrades should be more pronounced for direct downgrades.

Sample and Methodology

Sample Selection

The credit rating changes and watch-lists announcements come from Standard &

Poor’s (S&P). We focus on the period 2011-2016 because this is the period for which we can

obtain firm-specific daily investor sentiment data from PsychSignal for firms listed on NYSE,

NASDAQ and AMEX. This data is based on messages posted on Twitter and Stock Twits. By

cross-referencing the PsychSignal data with the S&P rating changes data we obtain a sample

of 1,214 rating changes, both upgrades and downgrades. Stock price data is obtained from

the Center for Research in Security Prices (CRSP). We obtain firms’ accounting data from

Compustat and the CBOE S&P500 Volatility Index (VIX) from the Chicago Board Options

Exchange (CBOE).

6
 We convert the long-term ratings from an alphabetical scale, AAA to D, where AAA is

the highest credit quality to a numerical scale 22-1, where 22 is AAA and 1 is SD and D. Firms

with a rating BBB- or above are Investment-grade whilst firms with a rating of BB+ or below

are Speculative-grade. Our credit watch data is categorised as positive, negative or

developing. Our main rating change variable DOWN is an indicator variable with a value of 1

for rating downgrades and 0 for rating upgrades. SPECULATIVE is an indicator variable equal

to 1 for speculative-grade firms and 0 for investment-grade firms. Grades are defined based

on pre-event rating levels. Our credit watch data is matched for three periods prior to a rating

change; 90 days, 180 days and 365 days.

We obtain the daily sentiment data required to measure firm-specific investor

sentiment from PsychSignal, a commercial organization that provides public mood data by

analysing the content of messages posted on Twitter and StockTwits using a Natural Language

Processing (NLP) engine. We use PsychSignal’s bullish intensity and bearish intensity

measures which represent the strength of the bullishness and bearishness that is revealed in

the tweets about the firms in our sample. We combine these two measures to generate a net

measure of firm-specific investor sentiment. Following Antweiler and Frank (2004) we

calculate our cumulative firm-specific sentiment index (CSI) as:

1+Bullish Intensity
CSIi,(−2,−1) = ∑−1 i,t
t=−2 Ln (1+Bearish Intensity ) (1)
i,t

Our measure of CSI represents public mood towards firm i over a two-day window,

from two days before until one day before the rating change announcement date. CSI has a

positive value if investors are optimistic about a firm and a negative value if investors are

pessimistic about a firm.

7
 For our measure of Cumulative Abnormal Returns (CAR) we measure market adjusted

model CAR in a three-day event window starting from the day of the announcement as

follows;

CAR i,(0,+2) = ∑+2

t=0(R i,t − R m,t ) (2)

Where: Ri,t is the stock return for firm i on day t and Rm,t is the CRSP value-weighted index
return on day t.

Summary Statistics

Table 1 provides an overview of our sample. Of the 1,214 rating changes, 569 are

downgrades and 645 are upgrades. Changes are generally one notch up or down but 8 firms

experience a downgrade of more than five notches, and 12 firms experience an upgrade of

more than five notches. The greatest number of observations are in 2015 and 2016, 241 and

338 respectively. Firms listed on NYSE account for 74% of the sample, firms listed on NASDAQ

account for 25%, while firms listed on AMEX account for the remaining 1%.

Table 2 presents the summary statistics for the variables used in our study.4 All

variables, except dummy variables and log transformed variables, are winsorized at 1% and

99% of their distributions to mitigate the impact of outliers.5 The average of three-day CAR,

CAR(0,+2), is negative at -0.0043, despite the fact that there are more positive events than

4 Detailed definitions of all variables can be found in Appendix A

5
The variables LEVERAGE, CONVERTIBLEDEBT, COSTDEBT, and CASHHOLDING are winsorized only at the high
end of their distributions.

8
negative events in our sample. The average of the two-day sentiment index CSI(-2,-1), is 0.4590

suggesting that on average investors were optimistic about the firms in our sample.

[Please Insert Tables 1 and 2 about Here]

Table 3 presents the correlations of the variables used in the study. As expected, there

is a positive correlation between CAR(0,+2) and CSI(-2,-1) suggesting that positive (negative)

abnormal returns are associated with bullish (bearish) investor sentiment. CAR(0,+2) is

negatively correlated with DOWN and SPECULATIVE suggesting that CAR(0,+2) is lower for

downgraded firms and speculative-grade firms. The correlation between DOWN and CSI(-2,-1)

is negative suggesting that rating downgrades, are associated with bearish investor

sentiment. In order to check for multicollinearity biases we conduct VIF tests and the results

show that there is no multicollinearity problem.

[Please Insert Table 3 about Here]

Methodology

We employ a standard event study methodology to test the impact of rating changes

and firm-specific investor sentiment on abnormal stock returns. We use OLS regressions that

control for rating, firm, and market characteristics that can affect the stock market response

around rating changes. Our regression model incorporating the full list of variables takes the

following form:

9
CAR i,(0,+2) = α + β1 DOWN ∗ CSIi,(−2,−1) + β2 DOWN + β3 CSIi,(−2,−1) +

β4 SPECULATIVE + β5 RATEDAYS + β6 LOSS + β7 MB + β8 SIZE + β9 PROFITABILITY +

β10 CAR i,(−202,−3) + β11 LEVERAGE + β12 CONVERTIBLEDEBT + β13 COSTDEBT +

β14 CASHHOLDING + β15 VIX + Year F. E. + Sector F. E. +ε (3)

Where: CARi,(0,+2): Cumulative Abnormal Returns, DOWN: Rating Downgrades Dummy Variable, CSIi,(-2,-1): Cumulative Firm-
Specific Investor Sentiment Index, SPECULATIVE: Pre-Event Rating Grade Dummy Variable, RATEDAYS: Rating Days, LOSS:
Loss Dummy Variable, MB: Market to Book Ratio, SIZE: Firm Size, PROFITABILITY: Profitability Ratio, CAR i,(-202,-3): Long Run
Cumulative Abnormal Returns, LEVERAGE: Leverage, CONVARTIBLRDEBT: Convertible Debt, COSTDEBT: Cost of Debt,
CASHHOLDING: Cash Holding Ratio, VIX: Volatility Index.

In order to test Hypothesis 1 we create the interaction variable DOWN*CSI(-2,-1). We

incorporate additional control variables that are known to affect firms’ abnormal returns in

in order to disentangle the effect of firm-specific investor sentiment from other effects. We

add the following control variables: speculative, rate-days, loss, market to book, size,

profitability, stock price momentum, leverage, convertible debt, cost of debt, cash holdings,

and volatility index. We also include year and sector fixed effects, and use standard errors

adjusted for heteroscedasticity and firm clustering. The interaction variable DOWN*CSI(-2,-1)

measures the impact of firm-specific investor sentiment at the announcement of rating

downgrades and we expect to find a statistically significant coefficient. In order to test

Hypothesis 2 we split the sample into speculative-grade and investment-grade firms. The

main variable of interest is the interaction variable DOWN*CSI(-2,-1). We expect to find a

statistically significant coefficient for the interaction variable DOWN*CSI(-2,-1) in the

speculative-grade sample, and insignificant or less significant coefficient in the investment-

grade sample. In order to test Hypothesis 3 we split the sample into direct-rating changes and

watch-preceded rating changes. We control for the timescale between the addition to S&P’s

10
CreditWatch list and the rating change on the basis of a 90-day, 180-day or 365-day period.

The main variable of interest is the interaction variable DOWN*CSI(-2,-1). We expect to find a

statistically significant coefficient for the interaction variable DOWN*CSI(-2,-1) in the direct

rating changes sample, and insignificant or less significant coefficient in the watch-preceded

rating changes sample.

Empirical Analysis

Our Hypothesis 1 states that the impact of investor sentiment on the stock market

response to rating changes should be more pronounced for rating downgrades than

upgrades. As an initial investigation we present graphically, in Figure 1, the CARs for 4

different firm samples over a 21 days window around the rating changes announcements. We

create the 4 samples based on the type of rating changes (downgrades and upgrades) and the

type of investor sentiment (positive and negative). The comparison of the CARs plots suggests

that the stock market response to rating changes is conditional on the type of rating change

and investor sentiment. Firms with rating downgrades and negative sentiment experience a

large decline in CARs while the rest of the firms do not experience any similar large reactions.

These plots suggest that investor sentiment is more pronounced for rating downgrades.

[Please Insert Figure 1 about Here]

Table 4 presents the results of our regression analysis. Model 1 controls only for rating

characteristics. The coefficient of the interaction variable DOWN*CSI(-2,-1) is positive and

11
significant at the 10% level. Positive (negative) investor sentiment leads to higher (lower)

abnormal returns at the announcement of rating downgrades. The variable DOWN is negative

and significant at the 1% level when measured separately, while the variable CSI(-2,-1) which

measures the effect of sentiment for rating upgrades is insignificant when measured

separately. Consistent with prior studies, the price response to rating downgrades is

significant and negative. The variable SPECULATIVE is negative and significant at the 5% level,

while the variable RATEDAYS is insignificant. Model 2 controls for all rating, firm and market

characteristics. The coefficient of the interaction variable DOWN*CSI(-2,-1) is positive and

significant at the 5% level in this model, while the separate DOWN is negative and significant

at the 1% level, and the separate CSI(-2,-1) insignificant. After including the control variables

SPECULATIVE loses its statistical significance. RATEDAYS remains insignificant. Regarding the

other controls, CAR(-202,-3), LEVERAGE, and CASHHOLDING are statistically significant at the 5%

level. Model 2 confirms the results from Model 1 and provides further support to Hypothesis

1. firm-specific investor sentiment has a significant impact on downgraded firms’ abnormal

returns.

[Please Insert Table 4 about Here]

Our Hypothesis 2 states that the impact of investor sentiment on the stock market

response to rating downgrades should be more pronounced for low-rated firms. As an initial

investigation we present graphically, in Figure 2, the CARs for 4 different firm samples over a

21 days window around the rating downgrade announcements. We create the 4 samples

based on the prior rating level (speculative and investment) and the type of investor

12
sentiment (positive and negative). The comparison of the CARs plots suggests that the stock

market response to rating downgrades is conditional on the prior rating level and investor

sentiment. Speculative-grade firms with negative sentiment experience a large decline in

CARs while the rest of the firms do not experience any similar large reactions. These plots

suggest that the impact of investor sentiment is more pronounced for speculative firms.

[Please Insert Figure 2 about Here]

Table 5 presents the results of the regression analysis. We classify our sample into

speculative- and investment-grade firms and run separate regressions. Model 1 presents the

results for the speculative-grade firms. The coefficient of the interaction variable DOWN*CSI(-

2,-1) is positive and significant at the 5% level. Positive (negative) investor sentiment leads to

higher (lower) abnormal returns during rating downgrades for firms with low ratings. The

separate variable DOWN is negative and significant at the 1% level, while the separate

variable CSI(-2,-1) which measures the effect of sentiment for rating upgrades is insignificant.

Model 2 presents the results for the investment-grade firms. The coefficient of the interaction

variable DOWN*CSI(-2,-1) is statistically insignificant. Positive (negative) investor sentiment has

no effect on the abnormal returns at the announcement of rating downgrades for firms with

high ratings. The separate variable DOWN and separate variable CSI(-2,-1) are both insignificant.

This result provides support to Hypothesis 2. Regarding the control variables, LOSS, CAR(-

202,-3), LEVERAGE, and CASHHOLDING are statistically significant.

[Please Insert Table 5 about Here]

13
 Our Hypothesis 3 states that the impact of investor sentiment on the stock market

response to rating downgrades should be more pronounced for direct downgrades. This

focuses on the information content and the element of surprise contained in a direct

(unanticipated) rating downgrade relative to a watch-preceded (anticipated) downgrade and

the role played by investor sentiment in this context. As an initial investigation we present

graphically, in Figure 3, the CARs for 4 different firm samples over a 21 days window around

the rating downgrades announcements. We create the 4 samples based on the type of rating

downgrades (direct and watch-preceded) and the type of investor sentiment (positive and

negative). The comparison of the CARs plots suggests that the stock market response to rating

downgrades is conditional on the type of rating downgrades and investor sentiment. Firms

with direct rating downgrades and negative sentiment experience the sharpest decline in

CARs at the rating downgrade announcements. On comparison, firms with watch-preceded

downgrades and negative sentiment experience a gradual decline in CARs a few days before

the rating downgrade announcements which is consistent with a market anticipation effect.

The rest of the firms do not experience any similar large reactions. These plots suggest that

investor sentiment is more pronounced for direct rating downgrades.

To further investigate the impact of investor sentiment on direct- and watch-preceded

rating downgrades we conduct a regression analysis. The summary statistics in Table 6 show

that the majority of the rating changes in our sample are direct (they were not added to a

watch list prior to a downgrade). Additionally, the majority of CreditWatch placements in our

sample take place 90-days prior to the rating changes announcements. Table 7 presents the

results of the regression analysis. We classify our sample into direct- and watch-preceded

rating changes firms and run separate regressions. Models 1, 3 and 5 present the results for

14
the direct rating changes. In all the models the coefficient of the interaction variable

DOWN*CSI(-2,-1) is positive and significant. Positive (negative) investor sentiment leads to

higher (lower) abnormal returns at the announcement of direct rating downgrades. Models

2, 4 and 6 present the results for the watch-preceded rating changes. In all the models the

coefficient of the interaction variable DOWN*CSI(-2,-1) is statistically insignificant. Positive

(negative) investor sentiment has no effect on the abnormal returns at the announcement of

watch-preceded (anticipated) rating downgrades. These results provide support to

Hypothesis 3 and also confirm the informational effect of CreditWatch lists on stock prices.

CreditWatch lists contain valuable information that is useful for estimating the probability of

actual rating changes. The level of uncertainty during watch-preceded rating changes is low

since, investors have time to assess rationally the risks of the upcoming rating changes and

this makes them less vulnerable to behavioural biases. In contrast, during direct rating

changes investors do not have time to assess rationally the risks of the upcoming rating

changes and this makes them more vulnerable to behavioural biases. Regarding the control

variables, CAR(-202,-3), LEVERAGE, COSTDEBT, CASHHOLDING and VIX are statistically

significant in some models.

[Please Insert Tables 6 and 7 about Here]

Conclusion

This study considers the impact of investor sentiment, based on social media, on the

stock price reaction to the announcement of credit rating changes. Our findings confirm that

the stock price response to a credit rating downgrade is greater than a rating upgrade, the

15
price reaction is greater for a speculative-grade firm than an investment-grade firm and

similarly the stock price impact is greater for a direct (unanticipated) rating change than for a

rating change preceded by addition to a watch list. Our contributions to the literature stem

from the investigation of the role of investor sentiment, based on social media, in the context

of these issues. First, we find that firm-specific investor sentiment has a significant impact on

stock price reaction to rating downgrades; when investor sentiment is positive it moderates

the stock price impact of a ratings downgrade and when it is negative it amplifies the stock

price impact of a downgrade. Our second contribution is the finding that the impact of firm-

specific investor sentiment on stock prices is stronger for speculative-grade firms than

investment-grade firms when credit ratings are downgraded. This is related to information

asymmetry and illiquidity of more risky stocks. Our final contribution is the finding that when

stock prices react negatively to direct (unanticipated) rating downgrades investor sentiment

has a significant price impact in a short period around those rating changes. This is consistent

with the process of assimilating new information which is a shock rather than the process of

assessment which may take place when a firm has been previously placed on a credit watch

list.

16
References

Antweiler, W., & Frank, M. Z. (2004). Is All That Talk Just Noise? The Information Content of Internet
Stock Message Boards. The Journal of Finance, 59(3), 1259-1294.
Baker, M., & Wurgler, J. (2006). Investor Sentiment and the Cross-Section of Stock Returns. The Journal
of Finance, 61(4), 1645-1680.
Bannier, C. E., & Hirsch, C. W. (2010). The Economic Function of Credit Rating Agencies - What Does
the Watchlist Tell Us? Journal of Banking and Finance, 34(12), 3037-3049.
Bedendo, M., Cathcart, L., & El-Jahel, L. (2018). Reputational shocks and the information content of
credit ratings. Journal of Financial Stability, 34, 44-60.
Chan, H., Faff, R., Hill, P., & Scheule, H. (2011). Are watch procedures a critical informational event in
the credit ratings process? An empirical investigation. Journal of Financial Research, 34(4),
617-640.
Chen, H., De, P., Hu, Y. J., & Hwang, B. H. (2014). Wisdon of Crowds: The Value of Stock Openions
Transmitted through Social Media. The Review of Financial Studies, 27(5), 1367-1403.
Chung, K. H., Frost, C. A., & Kim, M. (2012). Characteristics and Information Value of Credit Watches.
Financial Management, 41(1), 119-158.
Da, Z., Engelberg, J., & Gao, P. (2015). The Sum of All FEARS Investor Sentiment and Asset Prices. The
Review of Financial Studies, 28(1), 1-32.
Danbolt, J., Siganos, A., & Vagenas-Nanos, E. (2015). Investor Sentiment and Bidder Announcement
Abnormal Returns. Journal of Corporate Finance, 33, 164-179.
Followill, R. A., & Martell, T. (1997). Bond Review and Rating Change announcements: An Examination
of Informational Value and Market Efficiency. Journal of Economics and Finance, 21(2), 75-82.
Galil, K., & Soffer, G. (2011). Good news, bad news and rating announcements: An empirical
investigation. Journal of Banking & Finance, 35(11), 3101-3119.
Hand, J. R., Holthausen, R. W., & Leftwich, R. W. (1992). The Effect of Bond Rating Agency
Announcements on Bond and Stock Prices. The Journal of Finance, 47(2), 733-752.
He, Y., Wang, J., & Wei, K. J. (2011). Do bond rating changes affect the information asymmetry of stock
trading?. Journal of Empirical Finance, 18(1), 103-116.
Holthausen, R. W., & Leftwich, R. W. (1986). The Effect of Bond Rating Changes on Common Stock
Prices. Journal of Financial Economics, 17(1), 57-89.
Jorion, P., & Zhang, G. (2007). Information Effects of Bond Rating Changes: The Role of the Rating Prior
to the Announcement. The Journal of Fixed Income, 45-59.
Jorion, P., & Zhang, G. (2010). Information Transfer Effects of Bond Rating Downgrades. The Financial
Review, 45(3), 683-706.
Joseph, K., Wintoki, M. B., & Zhang, Z. (2011). Forecasting Abnormal Stock Returns and Trading Volume
Using Investor Sentiment: Evidence from Online Search. International Journal of Forecasting,
27(4), 1116-1127.

17
Kahneman, D., & Tversky, A. (1979). Prospect Theory: An Analysis of Decision under Risk.
Econometrica, 47(2), 263-291.
Karampatsas, N., Malekpour, S., & Mason, A. (2018). Beyond Market Mood: Firm Sentiment and the
Response to Corporate Earnings Announcements. Available at SSRN:
http://dx.doi.org/10.2139/ssrn.2926368
Kiesel, F., & Kolaric, S. (2018). Measuring the effect of watch-preceded and direct rating changes: a
note on credit markets. Review of Quantitative Finance and Accounting, 50(2), 653-672.
Li, X., & Luo, D. (2017). Investor Sentiment, Limited Arbitrage, and the Cash Holding Effect. Review of
Finance, 21(6), 2141-2168.
Mian, G. M., & Sankaraguruswamy, S. (2012). Investor Sentiment and Stock Market Response to
Earnings News. The Accounting Review, 87(4), 1357-1384.
Purda, l. (2007). Stock Market Reaction to Anticipated versus Surprise Rating Changes. The Journal of
Financial Research, 30(2), 301-320.

18
 Figure 1
Cumulative Abnormal Returns by Rating Changes and Firm-Specific Investor Sentiment

0.050
0.040
0.030
0.020
0.010
0.000
-0.010
-0.020
CARs

-0.030
-0.040
-0.050
-0.060
-0.070
-0.080
-0.090
-0.100
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10
Days

Downgrades / Positive Sentiment Firms Downgrades / Negative Sentiment Firms

Upgrades / Positive Sentiment Firms Upgrades / Negative Sentiment Firms

19
 Figure 2
Cumulative Abnormal Returns of Rating Downgrades by Rating Grades and Firm-Specific
Investor Sentiment

0.050
0.040
0.030
0.020
0.010
0.000
-0.010
CARs

-0.020
-0.030
-0.040
-0.050
-0.060
-0.070
-0.080
-0.090
-0.100
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10
Days

Downgrades / Speculative-Grade Firms / Positive Sentiment Firms

Downgrades / Speculative-Grade Firms / Negative Sentiment Firms
Downgrades / Investment-Grade Firms / Positive Sentiment Firms
Downgrades / Investment-Grade Firms / Negative Sentiment Firms

20
 Figure 3
Cumulative Abnormal Returns of Rating Downgrades by CreditWatch List Placements and
Firm-Specific Investor Sentiment

0.050
0.040
0.030
0.020
0.010
0.000
-0.010
CARs

-0.020
-0.030
-0.040
-0.050
-0.060
-0.070
-0.080
-0.090
-0.100
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10
Days

Direct Downgrades / Positive Sentiment Firms

Direct Downgrades /Negative Sentiment Firms
Watch-Preceded Downgrades / Positive Sentiment Firms
Watch-Preceded Downgrade / Negative Sentiment Firms

21
 Table 1
Summary Statistics by Rating Changes, Announcement Year, Stock Exchange, and Sector
This table presents summary statistics by rating changes, announcement year, market, and sector. Rating changes are based on
Standard & Poor’s (S&P) credit rating information. Announcement year is the credit rating changes announcement calendar year
and stock exchange is the market that stocks are traded on. Sector is classified based on Global Industry Classification Standard
(GICS). See Appendix A for detailed definitions of the variables. The sample includes stocks that are traded on the NYSE, NASDAQ,
and AMEX over the period of 2011-2016.

N Percent N Percent
Number of Obs 1,214 100% Announcement Year
2011 124 10.21%
2012 136 11.20%
Rating Changes 2013 189 15.57%
Downgrades 569 46.87% 2014 186 15.32%
Upgrades 645 53.13% 2015 241 19.85%
2016 338 27.84%

Stock Exchange
Rating Downgrades by Absolute Magnitude NYSE 902 74.30%
1 431 35.50% NASDAQ 300 24.71%
2 103 8.48% AMEX 12 0.99%
3 10 0.82%
4 7 0.58% Sector
5 10 0.82% Energy 259 21.33%
>5 8 0.66% Materials 89 7.33%
Industrials 146 12.03%
Consumer Discretionary 232 19.11%
Rating Upgrades by Absolute Magnitude Consumer Staples 62 5.11%
1 580 47.78% Health Care 64 5.27%
2 33 2.72% Financials 119 9.80%
3 7 0.58% Information Technology 101 8.32%
4 4 0.33% Telecommunication Services 19 1.57%
5 9 0.74% Utilities 43 3.54%
>5 12 0.99% Real Estate 80 6.59%

22
 Table 2
Descriptive Statistics
This table presents descriptive statistics of the key variables in our study: cumulative abnormal returns (CARs), rating changes
(DOWN), cumulative firm-specific investor sentiment (CSI), pre-event rating grade (SPECULATIVE), rating days (RATEDAYS), loss
(Loss), market-to-book ratio (MB), firms size (Size), profitability ratio (PROFITABILITY), leverage (Leverage), convertible debt
(CONVERTIBLEDEBT), cost of debt (COSTDEBT), cash holding ratio (CASHHOLDING), and volatility index (VIX). See Appendix A for
detailed definitions of the variables. The sample includes stocks that are traded on the NYSE, NASDAQ, and AMEX over the period
of 2011-2016. All variables except the dummy variables and log transformed variables are winsorized to mitigate the impact of
outliers.

N Mean Median Std. Dev. Min Max

CAR(0,+2) 1,214 -0.0043 0.0001 0.0656 -0.2893 0.2031

DOWN 1,214 0.4687 0.0000 0.4992 0.0000 1.0000

CSI(-2,-1) 1,214 0.4590 0.1586 0.9551 -1.9988 2.5734

SPECULATIVE 1,214 0.6474 1.0000 0.4780 0.0000 1.0000

RATEDAYS 1,214 6.4738 6.6561 1.2969 0.6931 9.2537

LOSS 1,214 0.2801 0.0000 0.4492 0.0000 1.0000

MB 1,208 2.5419 1.8146 6.0340 -27.4919 31.9605

SIZE 1,214 8.1036 8.0842 1.6326 3.1976 13.3480

PROFITABILITY 1,161 0.0753 0.1062 0.2052 -0.9278 0.3808

CAR(-202,-3) 1,201 -0.0887 -0.0057 0.4409 -1.5074 0.9517

LEVERAGE 1,211 0.3989 0.3714 0.2439 0.0000 1.2988

CONVERTIBLEDEBT 1,167 0.0146 0.0000 0.0426 0.0000 0.2209

COSTDEBT 1,158 0.0608 0.0578 0.0248 0.0028 0.1539

CASHHOLDING 1,202 0.5283 0.1804 1.0149 0.0000 6.5569

VIX 1,214 16.9293 15.4700 4.3639 12.2964 32.8291

23
 Table 3
Correlation Matrix
This table presents the correlations of the variables, cumulative abnormal returns (CARs), rating changes (DOWN), cumulative firm-specific investor sentiment (CSI), pre-event rating grade
(SPECULATIVE), rating days (RATEDAYS), loss (Loss), market-to-book ratio (MB), firms size (Size), profitability ratio (PROFITABILITY), leverage (Leverage), convertible debt (CONVERTIBLEDEBT),
cost of debt (COSTDEBT), cash holding ratio (CASHHOLDING), and volatility index (VIX). See Appendix A for detailed definitions of the variables. The sample includes stocks that are traded on
the NYSE, NASDAQ, and AMEX over the period of 2011-2016. All variables except the dummy variables and log transformed variables are winsorized to mitigate the impact of outliers.

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15)
(1) CAR(0,+2) 1.0000
(2) DOWN -0.1367 1.0000
(3) CSI(-2,-1) 0.0656 -0.1284 1.0000
(4) SPECULATIVE -0.0632 -0.0394 0.0180 1.0000
(5) RATEDAYS 0.0843 -0.1070 0.0700 -0.3647 1.0000
(6) LOSS -0.1420 0.3259 -0.0665 0.2990 -0.3379 1.0000
(7) MB 0.0443 -0.0785 0.0319 -0.1048 0.1073 -0.1244 1.0000
(8) SIZE 0.0816 -0.1831 -0.0058 -0.5929 0.3990 -0.4452 0.1982 1.0000
(9) PROFITABILITY 0.1343 -0.1857 0.0911 -0.1838 0.3581 -0.4833 0.2013 0.2994 1.0000
(10) CAR(-202,-3) 0.1422 -0.4534 0.1173 -0.0996 0.1960 -0.2490 0.0609 0.2254 0.1808 1.0000
(11) LEVERAGE -0.0194 0.1745 -0.0215 0.3461 -0.3004 0.3372 -0.0764 -0.4554 -0.3302 -0.2055 1.0000
(12) CONVERTIBLEDEBT -0.0593 -0.0505 -0.0157 0.1544 -0.0831 0.0862 -0.0757 -0.1002 -0.1521 0.0003 0.0068 1.0000
(13) COSTDEBT -0.0749 -0.0628 -0.0354 0.3815 -0.2400 0.2437 -0.0696 -0.4394 -0.1449 -0.0644 0.0962 0.1052 1.0000
(14) CASHHOLDING 0.0419 -0.1215 0.0078 -0.1625 0.1247 -0.1853 0.0443 0.2075 0.1480 0.1050 -0.4669 0.0305 0.0983 1.0000
(15) VIX -0.0052 0.2182 0.0163 -0.0314 -0.0088 0.0599 -0.0045 -0.0004 -0.0775 -0.2330 -0.0278 -0.0105 0.0625 -0.0066 1.0000

24
 Table 4
CARs and Firm-Specific Investor Sentiment
This table presents the results of OLS regressions investigating the effect of firm-specific investor sentiment (CSI) on
announcement-period abnormal returns (CARs). Other variables are rating changes (DOWN), pre-event rating grade
(SPECULATIVE), rating days (RATEDAYS), loss (Loss), market-to-book ratio (MB), firms size (Size), profitability ratio
(PROFITABILITY), leverage (Leverage), convertible debt (CONVERTIBLEDEBT), cost of debt (COSTDEBT), cash holding ratio
(CASHHOLDING), and volatility index (VIX). See Appendix A for detailed definitions of the variables. The sample includes
stocks that are traded on the NYSE, NASDAQ, and AMEX over the period of 2011-2016. All variables except the dummy
variables and log transformed variables are winsorized to mitigate the impact of outliers. All regressions control for year and
sector fixed effects whose coefficients are suppressed. The t-statistics reported in parentheses are adjusted for
heteroskedasticity and stock clustering. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level,
respectively.

(1) ( 2)

DOWN*CSI(-2,-1) 0.0084* 0.0100**

(1.91) (2.09)
DOWN -0.0204*** -0.0176***
(-4.40) (-3.19)
CSI(-2,-1) -0.0005 -0.0025
(-0.33) (-1.42)
SPECULATIVE -0.0090** -0.0059
(-2.49) (-1.17)
RATEDAYS 0.0020 0.0012
(0.98) (0.62)
LOSS -0.0091
(-1.23)
MB 0.0000
(0.13)
SIZE -0.0011
(-0.57)
PROFITABILITY 0.0258
(1.39)
CAR(-202,-3) 0.0137**
(2.27)
LEVERAGE 0.0304**
(2.21)
CONVERTIBLEDEBT -0.0628
(-0.70)
COSTDEBT -0.1260
(-1.46)
CASHHOLDING 0.0035**
(2.03)
VIX 0.0008
(1.35)

Year F.E. Yes Yes

Sector F.E. Yes Yes
N 1,214 1,078
Adjusted R2 0.0253 0.0392

25
 Table 5
CARs and Firm-Specific Investor Sentiment for Speculative/Investment-Grade Firms
This table presents the results of OLS regressions of cumulative abnormal returns (CARs) on cumulative firm-specific investor
sentiment (CSI) for non-investment-grade/investment-grade firms. The firms are classified based on the pre-event rating
grade. See Appendix A for detailed definitions of the variables. The sample includes stocks that are traded on the NYSE,
NASDAQ, and AMEX over the period of 2011-2016. All variables except the dummy variables and log transformed variables
are winsorized to mitigate the impact of outliers. All regressions control for year and sector fixed effects whose coefficients
are suppressed. The t-statistics reported in parentheses are adjusted for heteroskedasticity and stock clustering. *, **, and
*** indicate statistical significance at the 10%, 5%, and 1% level, respectively.

SPECULATIVE INVESTMENT
(1) (2)

DOWN*CSI(-2,-1) 0.0133** 0.0055

(1.99) (0.87)
DOWN -0.0222*** -0.0077
(-2.85) (-1.01)
CSI(-2,-1) -0.0035 -0.0015
(-1.49) (-0.65)
RATEDAYS 0.0016 -0.0014
(0.62) (-0.65)
LOSS -0.0148* 0.0106
(-1.76) (0.85)
MB 0.0001 0.0000
(0.55) (0.03)
SIZE -0.0025 0.0032
(-1.00) (1.49)
PROFITABILITY 0.0305 -0.0275
(1.48) (-1.20)
CAR(-202,-3) 0.0165** -0.0199
(2.38) (-1.32)
LEVERAGE 0.0430*** -0.0048
(2.74) (-0.20)
CONVERTIBLEDEBT -0.0779 0.0612
(-0.78) (1.25)
COSTDEBT -0.1557 -0.0301
(-1.43) (-0.24)
CASHHOLDING 0.0062*** 0.0011
(2.67) (0.50)
VIX 0.0006 0.0006
(0.70) (0.76)

Year F.E. Yes Yes

Sector F.E. Yes Yes
N 723 355
Adjusted R2 0.0524 0.0168

26
 Table 6
Summary Statistics by Credit Watch Placement
This table presents summary statistics by credit watch placement prior rating changes. Rating changes and additions to credit
watch lists are from Standard & Poor’s (S&P). Rating changes are classified based on whether a firm is placed on S&P
CreditWatch reviews within a 90-day/180-day/365-day timeframe prior to rating changes. The sample includes stocks that are
traded on the NYSE, NASDAQ, and AMEX over the period of 2011-2016.

N Percent N Percent

Number of Obs 1,214 100%

CreditWatch in 90 Days prior the Rating Changes

Surprise Downgrades 493 40.61% Surprise Upgrades 568 46.79%

Watch-Preceded Downgrades 76 6.26% Watch-Preceded Upgrades 77 6.34%

CreditWatch in 180 Days prior the Rating Changes

Surprise Downgrades 459 37.81% Surprise Upgrades 550 45.30%

Watch-Preceded Downgrades 110 9.06% Watch-Preceded Upgrades 95 7.83%

CreditWatch in 365 Days prior the Rating Changes

Surprise Downgrades 430 35.42% Surprise Upgrades 529 43.57%

Watch-Preceded Downgrades 139 11.45% Watch-Preceded Upgrades 116 9.56%

27
 Table 7
CARs and Firm-Specific Investor Sentiment for Direct/Watch-Preceded Rating Changes
This table presents the results of OLS regressions of cumulative abnormal returns (CARs) on cumulative firm-specific investor
sentiment (CSI) for sub-samples of surprise/watch-preceded rating changes. Rating changes are classified based on whether a firm is
placed on S&P CreditWatch reviews within a 90-day/180-day/365-day timeframe prior to rating changes. See Appendix A for detailed
definitions of the variables. The sample includes stocks that are traded on the NYSE, NASDAQ, and AMEX over the period of 2011-
2016. All variables except the dummy variables and log transformed variables are winsorized to mitigate the impact of outliers. All
regressions control for year and sector fixed effects whose coefficients are suppressed. The t-statistics reported in parentheses are
adjusted for heteroskedasticity and stock clustering. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level,
respectively.

90 DAYS 180 DAYS 365 DAYS

DIRECT WATCH DIRECT WATCH DIRECT WATCH

(1) (2) (3) (4) (5) (6)

DOWN*CSI(-2,-1) 0.0116** 0.0019 0.0110* 0.0039 0.0106* 0.0070

(2.16) (0.22) (1.92) (0.53) (1.77) (1.01)
DOWN -0.0203*** 0.0032 -0.0208*** -0.0034 -0.0218*** -0.0050
(-3.19) (0.29) (-3.14) (-0.36) (-3.10) (-0.50)
CSI(-2,-1) -0.0026 -0.0010 -0.0024 -0.0010 -0.0021 -0.0038
(-1.33) (-0.24) (-1.18) (-0.25) (-1.04) (-1.08)
SPECULATIVE -0.0063 -0.0016 -0.0067 -0.0066 -0.0045 -0.0124
(-1.09) (-0.14) (-1.12) (-0.58) (-0.74) (-1.21)
RATEDAYS 0.0007 0.0015 -0.0002 0.0045 -0.0011 0.0045
(0.30) (0.40) (-0.08) (1.25) (-0.44) (1.35)
LOSS -0.0105 -0.0033 -0.0109 -0.0006 -0.0077 -0.0144
(-1.25) (-0.19) (-1.28) (-0.04) (-0.87) (-0.84)
MB 0.0001 0.0003 0.0002 -0.0006 0.0002 -0.0002
(0.41) (0.46) (0.63) (-0.76) (0.62) (-0.30)
SIZE -0.0012 -0.0000 -0.0008 -0.0027 -0.0007 -0.0023
(-0.55) (-0.01) (-0.39) (-0.82) (-0.30) (-0.70)
PROFITABILITY 0.0249 -0.0092 0.0247 0.0396 0.0237 0.0486
(1.27) (-0.25) (1.24) (0.98) (1.18) (1.19)
CAR(-202,-3) 0.0110 0.0256 0.0095 0.0254* 0.0063 0.0345***
(1.59) (1.55) (1.31) (1.94) (0.85) (2.84)
LEVERAGE 0.0286* 0.0402 0.0273* 0.0350 0.0223 0.0576**
(1.91) (1.33) (1.83) (1.23) (1.45) (2.16)
CONVERTIBLEDEBT -0.0668 0.0001 -0.0680 0.0085 -0.0913 0.0815
(-0.66) (0.00) (-0.66) (0.08) (-0.86) (1.02)
COSTDEBT -0.1937* 0.2962 -0.1698* 0.0593 -0.1937* 0.1390
(-1.96) (1.27) (-1.68) (0.38) (-1.87) (1.00)
CASHHOLDING 0.0041** 0.0013 0.0036* 0.0030 0.0036* 0.0035
(2.15) (0.30) (1.85) (0.68) (1.83) (0.98)
VIX 0.0011* -0.0045* 0.0012* -0.0019 0.0012* -0.0015
(1.79) (-1.83) (1.82) (-1.24) (1.70) (-1.17)

Year F.E. Yes Yes Yes Yes Yes Yes

Sector F.E. Yes Yes Yes Yes Yes Yes
N 939 139 891 187 847 231
Adjusted R2 0.0424 -0.0054 0.0396 0.0095 0.0311 0.1074

28
Appendix A: Variables Definition

Variable Definition
Cumulative abnormal returns over the 3-day event window (0,+2), where 0 is the rating announcement date.
The returns are calculated using the market adjusted model parameters estimated over the period between
CAR(0,+2)
300 and 46 days prior to the rating announcement. The CRSP value-weighted index return is the market
return.
DOWN An indicator variable equal to 1 for rating downgrades and 0 for rating upgrades.
Cumulative firm-specific investor sentiment Index over the two-day window from 2 days before the rating
CSI(-2,-1) announcement date until 1 day before the date of announcement, where Sentiment Index (SI) is measured
as natural logarithm of (1+Bulish Intensity)/(1+Bearish Intensity).
An indicator variable equal to 1 for speculative-grade firms and 0 for investment-grade firms. Grades are
SPECULATIVE
defined based on pre-event ratings.
RATEDAYS The natural logarithm of number of days between current rating date and lag rating date.
An indicator variable equal to 1 for firms reporting negative net income in the year prior the rating
LOSS
announcement.

MB The market value of equity divided by the book value of equity in the year prior the rating announcement.

SIZE The natural logarithm of share price times shares outstanding in the year prior the rating announcement.

The ratio of earnings before interest, tax, depreciation, and amortization to total assets in the year prior the
PROFITABILITY
rating announcement.
Cumulative Abnormal Return relative to value-weighted market return over the (-202,-3) day interval prior
CAR(-202,-3)
the rating announcement.
The sum of long term debt and the debt in current liabilities divided by total assets in the year prior the rating
LEVERAGE
announcement.
CONVERTIBLEDEBT Convertible debt divided by total assets in the year prior the rating announcement.
Interest expenses divided by the sum of long term debt and the debt in current liabilities in the year prior the
COSTDEBT
rating announcement.
Cash and short-term investments divided by the sum of long term debt and the debt in current liabilities in
CASHHOLDING
the year prior the rating announcement.
VIX The average of the CBOE Volatility Index (VIX) in the month prior the rating announcement.

29