Politically Connected firms, Legal Enforcement and Analysts’ Forecast Attributes

advertisement
Politically Connected firms, Legal Enforcement and Analysts’
Forecast Attributes
Charles J. P. Chen
City University of Hong Kong
Yuan Ding*
China Europe International Business School (CEIBS)
Chansog (Francis) Kim
City University of Hong Kong
Old Dominion University
April 14, 2007
Please do not cite or quote without permission
Comments welcome.
* Address for correspondence: Yuan Ding, China Europe International Business School (CEIBS), 699 Hongfeng
Road, 201206 Shanghai, P. R. China. Email: dyuan@ceibs.edu. The authors would like to thank Junghun (Jay)
Lee (the discussant), Giorgio Gotti (the discussant), Henri-Claude de Bettignies, Chun Chang, Yan Gao, Kari
Lukka, Bala Ramasamy, Hannu Schadewitz, Bin Xu, Yimin Zhang, Xinge Zhao, Tian Zhu and workshop
participants at Seoul National University (November 2006), 13th Annual Mid-Year Conference of the
International Accounting Section of the American Accounting Association (Charleston, USA, February 2007),
research seminar at Turku School of Economics (Finland, March 2007) and research seminar at China Europe
International Business School (Shanghai, April 2007) for their helpful comments and advice. Charles Chen and
Yuan Ding acknowledge the financial support of Procore - France/Hong Kong Joint Research Scheme. Authors
are gratefully to Mara Faccio for generously providing her list of politically connected firms. The research
assistance of Ching Tung Yu is greatly appreciated. Part of the research was conducted when the second author
was affiliated with HEC School of Management, Paris.
1
Politically Connected firms, Legal Enforcement and Analysts’
Forecast Attributes
Abstract
We study the association between a firm’s political connection and analysts’ earnings
forecasts in an international context. We find that analysts experience more difficulties in
predicting earnings of firms with political connection than firms without it. However, in
those jurisdictions that have effective law enforcement, earnings forecasts are less influenced
by the firms’ political connections. Our findings contribute to the literature by identifying
political connection as an additional dimension to the forecasting difficulty and by showing
that effective legal enforcement can lessen the effect of political connection and improve
financial analysts’ forecasts.
Key words: analysts’ earnings forecasts; political connection; legal enforcement; forecast
attributes.
2
I. INTRODUCTION
This study investigates whether corporate political connection affects the attributes of
financial analysts’ earnings forecasts. This is an important research question not only because
attributes of analysts’ forecasts are of interest to finance and accounting researchers; but also
because political connection plays an important role in many of the world’s largest and most
important economies (Fisman, 2001, p. 138). Extant literature has identified both firm-level
and institutional factors that affect the properties of analysts’ forecasts (O'Brien, 1990; Brown,
1993; Ghosh and Whitecotton, 1997; Kwon, 2002; Hope, 2003; Clarke and Subramanian,
2006). Extending this line of research, we propose that in addition to the inherent uncertainty
of earnings, political connection by itself complicates the task of forecasting earnings both
because political favoritism is often granted covertly and because the effects of political
dealings on future earnings are complex and very difficult to forecast. This conjecture, which
we refer to as analyst task difficulty hypothesis, suggests that analyst earnings forecasts are
less accurate, more optimistic, and more divergent for firms with political connection than for
firms without it.1 Therefore, there should be systematic differences in the analysts’ forecast
attributes between connected and non-connected firms.
As enforcement of law can be expected to deter politicians from frequently engaging
in granting large amounts of benefits to connected firms (Faccio, 2006a), we also propose that
effective enforcement of law can mitigate the impact of political connection on financial
analysts’ forecast attributes.
This conjecture, which we refer to as law enforcement
effectiveness hypothesis, predicts that the effect of political connection on analysts’ forecast
attributes attenuate in jurisdictions with more effective enforcement of law.
Corporate executives tend to conceal information about expected future benefits
arising from their connections with politicians unless it is required to be disclosed and non-
1
One may contend that politicians could use their influence to help the connected firms smooth their earnings by
transferring political favors when earnings are low. In addition, analysts tend to cover winners and to stop
covering losers (McNichols and O'Brien, 1997). Firms with political connection have bigger sizes, more analyst
followings, more media coverage and more investors’ attentions than firms without it. Connected firms have
richer information environments which improve analysts’ capabilities to predict earnings of connected firms
more accurately. This conjecture, which we refer to as information environment hypothesis, suggests that
analyst earnings forecasts are more accurate and less divergent for firms with political connection than for those
without it. A priori, it is not clear whether political connection can mitigate or exacerbate information problems
faced by analysts. Therefore, it is an empirical question.
3
compliance with such requirement is expected to be costly.
This is so because public
knowledge of political favoritism not only often leads to unfavorable publicity for the firm,
but it also may cause troubles for the politicians involved. Therefore, it is more costly for
financial analysts to obtain information about political benefits that a connected firm may
expect to receive. Even if such information were available to the analysts, the exact impact of
politically motivated transactions on future earnings will be much more difficult to estimate as
compared to that of normal business transactions. First, the realization of political benefits is
often dependent upon the swing of the political pendulum which falls outside the expertise
area of financial analysts which normally does not include estimation of the outcome of
political events. Second, both financial institutions and their employed financial analysts
could be intimidated by the firm’s political influence and have strong incentives to exclude
the effect of expected political favoritism in their earnings forecasts, because showing the
effect of political favoritism would be similar to confessing their possession of private
information about politicians’ under-the-table deals with corporations. This would offend
both the corporate executives and their powerful political allies. There are at least three
obvious undesirable consequences: 1) the corporate executives may cease to co-operate with
financial analysts, making their future forecasts less accurate2; 2) the corporation may shun
away from the analysts’ employer in their future financial dealings, directly reducing the
revenue of the financial institution involved 3 ; 3) the politicians concerned may retaliate
against the financial analysts and their employers, especially in jurisdictions with weak
enforcement of law. The above discussion suggests that we should investigate whether
political connection per se affects the attributes of financial analysts’ earnings forecasts and
the effectiveness of enforcement of law must be incorporated into this investigation as a
mediating institutional factor.
Although political connection is an important social economic issue with broad
implications, there were only few studies in this area (Krueger, 1974; Roberts, 1990; Fisman,
2001) until Faccio (Faccio, 2006b, 2006a; Faccio et al., 2006; Faccio and Parsley, 2006) made
a breakthrough. The previous lack of systematic research in this area was mainly due to
difficulties in defining political connections, collecting the appropriate data and estimating the
value of such connections. By identifying 541 connected firms in 35 countries out of 20,202
2
See Francis and Philbrick (1993) and Lim (2001) for analysts’ management relation incentive to retrieve firms’
inside information from managers of firms they follow.
3
See Dugar and Nathan (1995) and Lin and McNichols (1998) for the investment banking incentives of analysts.
4
publicly traded firms in 47 countries between 1997 and 2003, Faccio (2006b) was able to
investigate the common characteristics shared by countries with widespread political
connections and whether such connections add to company value.
Faccio et al. (2006)
analyzed the likelihood of government bailouts of 450 politically-connected (but publiclytraded) firms from 35 countries over the period 1997 through 2002 and found that connected
firms are more likely to be rescued by government. Furthermore, based on their examination
of a large number of sudden deaths of politicians around the world, Faccio and Parsley (2006)
found that these incidences were associated with a 2% decline in market value of politically
connected companies.
Adding to this strand of emerging literature, we identify political connection as an
additional dimension of financial analysts’ forecast difficulties and find that, after controlling
for inherent differences in earnings quality, connected firms are associated with less accurate
analysts’ forecasts, more optimistic forecast biases and a higher level of forecast divergence.
These findings are consistent with the extant literature that shows 1) a positive relationship
between the complexity of the forecasting task and the error of analysts’ earnings forecasts
(Brown et al., 1987; Bhushan, 1989; Lang and Lundholm, 1996; Clement, 1999), 2) less
predictable earnings leads to more optimistic biases in analysts’ earnings forecasts (Das et al.,
1998; Lim, 2001; Duru and Reeb, 2002), and 3) both uncertainty and opacity increase analysts
forecast divergence (Daley et al., 1988; Imhoff Jr. and Lobo, 1992; Barron and Stuerke, 1998;
Park, 2005).4
The remainder of the paper proceeds as follows: section two develops hypotheses;
section three presents variable definitions and empirical models; section four identifies data
source and reports descriptive statistics; section five discusses empirical analysis results, and
section six summarizes robustness check results followed by conclusions in section seven.
II. HYPOTHESIS DEVELOPMENT
The main objective of the paper is to investigate how political connection influences
analysts’ forecast attributes. Political connection can be expected to either increase or
4 While prior studies have used the dispersion in analysts' forecast as a proxy for divergence of opinions
(Ziebart, 1990; Atiase and Bamber, 1994; Diether et al., 2002) and for uncertainty (Imhoff Jr. and Lobo, 1992;
Barron and Stuerke, 1998; Zhang, 2006), Barron et al. (1998) show that dispersion is a function of diversity of
opinions and uncertainty. We use in this paper the dispersion in analysts’ forecasts as a proxy for uncertainty.
However, we also tested a more refined measure of uncertainty by using Barron et al. (1998) metrics. The
results, not reported, maintain the same toner as those reported in this paper.
5
decrease the firm’s earnings volatility. On the one hand, it can be argued that political
favoritism often comes in a windfall manner that creates crests in a firm’s earnings time
series, amplifying the variance of earnings. On the other hand, politicians could use their
influence to help the connected firms smooth their earnings by transferring political favors
when earnings are extremely low. Ex ante, both arguments appear to be sensible. It is
therefore an empirical issue whether political connection systematically increases or decreases
earnings volatility.
Though the descriptive statistics of our sample in Table 3 show a lower disclosure
level, low accrual quality, and a higher level of cash flow from operations (CFO) for
politically connected firms than for non-connected ones, we believe that, even after
controlling for this effect, the political connection may still affect analysts’ forecast attributes.
There are at least two explanations. First, the impact of political connection on firm’s
performance is often abrupt. Since forthcoming transactions carrying political favoritism are
often not disclosed or not completely disclosed, analysts have difficulty predicting the firm’s
future earnings based on historical data and other publicly available information. Second,
even when these forthcoming transactions were disclosed, it would be extremely difficult to
predict their impact on the firm’s future earnings because of their sudden and interruptive
nature.
However, one may question the link between forecast attributes and the forecasting
task complexity associated with political connection since the extant literature shows that
firms may manipulate theirs earnings to meet the analysts’ forecasts as well5. For example,
using U.S. data, Degeorge, Patel and Zeckhauser (1999) find that companies apparently strive
to meet or exceed the analyst consensus forecast for quarterly earnings (consensus fixation).
Graham, Harvey and Rajgopal (2005) show that top U.S. executives are willing to give up
positive NPV projects to meet earnings benchmarks.
However, managers of politically
connected firms can be expected to behave differently, because they enjoy preferential
government treatments and they are relatively immune to merger and acquisition threats. In
other words, they have more job security which makes avoiding poor firm performance less
an incentive for them to manipulate earnings to meet analysts’ targets.
Of course, an
opposing argument also exists. Due to government-provided protection from public scrutiny,
managing earnings may cost managers of politically connected firms less, because even if
their wrong doings were exposed, politicians could lend them a hand to minimize the adverse
5
See Levitt (1998) and Fuller and Jensen (2002).
6
effect on the firm and its executives. It remains an empirical question whether the reduced
incentive to manage earnings can out-weight the effect of decreases in monitoring
mechanisms for politically connected firms6.
Political Connection and Forecast Error
The error of an earnings forecast depends on the difficulty or complexity of the
forecasting task (Duru and Reeb, 2002). We propose that there are at least three reasons that
political connection makes the analysts’ forecasting task more difficult.
connection adds a new dimension to the earnings generating process.
First, political
Krueger (1974)
suggests that entrepreneurs expend resources on politicians to compete for economic rent
which could be granted by the government in the form of favorable tax treatment, profitable
projects, preferential access to markets, cheap financing, government subsidies, and etc. The
pay-backs of political connection usually come in a windfall fashion which inevitably affects
the pattern of the flow of the reported earnings of the connected firm, making analysts’
forecast task more complex. On the other hand, when government officials lose their political
influence, the cash flow into the connected firm will dwindle and so will firm value as is
indicated by findings in Fisman (2001) and Faccio and Parsley (2006). The unpredictable
nature of the adverse effect of falling out of political favor complicates the earnings
forecasting task as well.7
Another reason that political connection can make earnings forecasting task more
difficult is due to the possibility that connected firms may request the government to help
them smooth their earnings by granting political favors when earnings shrink.
Though
government aids may fill up the valleys in the earnings pattern, they also increase the
forecasting difficulty. First, there is high uncertainty about when and how much government
aid the connected firm may receive, if it comes at all. The uncertainty arises from the fact that
granting government aid would offset the equilibrium status of the political arena leading to
redistribution of wealth among stakeholders. Obviously, this is not an easy task that can be
accomplished without much political maneuver. Furthermore, either the success or the failure
6
Our results are not driven by earnings management effect since the consensus fixation effect is only visible in
highly developed capital market (Degeorge et al., 2004). The results reported in this study remain qualitatively
similar whether U S and UK firms are included or not.
7
We assume that each analyst has similar forecasting capabilities. Consequently, we do not control for analysts’
experience, their employers’ size, and the number of firms and industries covered by analysts. We acknowledge
that analysts’ characteristics may affect their forecasting performance (Jacob et al., 1999; Cowen et al., 2006).
However, we do not see, a priori, that the differential forecasting abilities of analysts adversely affect their
forecasting attributes of the connected firms in a systematical way.
7
in securing government aid will increase the fluctuation of earnings in the future. In receiving
government aids, the connected firm will have in its earnings a more transit component that is
not expected to persist into the future, making the crystal ball more smudged for financial
analysts to look into. If it turns out that government does not grant the favors in time to be
reported in the current earnings, future earnings will be contingent upon the outcome of the
connected firm’s political efforts, which often times is not an area in which many financial
analysts have expertise or experience8.
Last but not the least, our proposal that government decisions significantly influence
the error of earnings forecasts of connected firms is consistent with prior studies that
document that the equity value of politically connected firms can be easily affected by
political events (Roberts, 1990; Fisman, 2001; Faccio, 2006b). Fisman (2001) argued that “in
Southeast Asia, political connectedness, rather than fundamentals such as productivity, was
the primary determinant of profitability and this had led to distorted investment decisions”.
At the same time, political connection often links to greater opacity at firm level. Due to
government-provided shielding from market monitoring mechanisms (e.g., regulatory
disclosure requirements, investors’ demand for transparency, etc.), managers enjoy more
financial disclosure discretion in firms with political connection, which increases information
asymmetry between analysts and management. Therefore, analysts are less likely to forecast
future earnings of these firms accurately9.
H1: Firms with political connection are associated with less accurate analysts’ earnings
forecasts, ceteris paribus.
Political Connection and Optimistic Bias
There are several reasons for us to expect that politically connected firms are
associated with more optimistic bias in financial analysts’ forecasts. First, as previously
8
The government aid may come either in the form of politically favored projects that affect operating income or
in the form of direct fiscal transfer or tax breaks that do not affect operating income. Either way, the increased
uncertainty complicates the earnings forecast task. We acknowledge that analysts tend to forecast the EPS from
continuing operations. However, classification of earnings into recurring and non-recurring items is fuzzy.
Hand (1990) identifies about 2/3 of his sample firms classified gains from debt-equity swaps as net income from
continuing operations. Besides, management has incentives to classify good news as continuing operations and
bad news as nonrecurring.
9
One may contend that Firms with political connection are associated with more accurate analysts’ earnings
forecasts. The connections are more common in countries with transparent economies (Faccio, 2006b). Faccio
(2006b) conjectures that the transparent economies are better able to tolerate connections because a misuse
would be more likely to be detected. However, whether these country-level transparency variables in economic
systems would lead to firm-level transparency is an empirical question. Besides, her study does not intend to
imply any causality.
8
discussed, earnings of politically connected firms are more difficult to predict. Lim (2001)
shows that analysts issue more favorable forecasts for firms with less predictable earnings in
order to maintain favorable relations with managers who allow them to obtain access to
private information.
Extant literature also shows that financial community tends to be
friendlier with politically connected firms. For instance, Hutchcroft (1998) describes how
troubled banks that lent to Philippines President Marcos and his cronies enjoyed important
privileges, including “emergency loans and generous equity infusions from state banks.” The
cozy relationship between the financial community and connected firms can be expected to
contribute to the optimistic bias in analysts’ forecasts.
Second, analysts may overweigh their private information on political connection
relative to other public accounting information.
Since the excess weight on private
information is more pronounced in good news than in bad news, it leads to optimistic bias in
analyst forecasts. Both economic incentives and behavioral bias can affect the degree of the
deviation from efficient weighting on private and public information (Chen and Jiang, 2006).
It can be reasoned that private information about politically connected firms more often than
not is positive and the overweight placed on it leads to an upward bias. Especially, the
behavioral bias (e.g., overconfidence) may be more severe in some countries with lessdeveloped analyst industry (see Jiang et al., 2005 for the review on overconfidence).
Third, empirical evidence suggests that financial institutions often treat connected
firms more optimistically than others. For instance, Faccio et al. (2006) find politically
connected (but publicly-traded) firms are more likely to be bailed out than are their nonconnected peers, although the former exhibit significantly poorer operating performance than
the latter at the time of the bailout and over the following two years. Furthermore, they also
confirm former systematic and anecdotal evidence that politically connected firms make
greater use of debt financing than their non-connected peers.
Fourth, unfavorable forecasts may upset not only the firm but also its political
connection, leading to undesirable consequences for both the individual analysts and their
affiliated securities firms, particularly, if the enforcement of law is so poor that politicians can
act without serious concerns for check and balance imposed by the rule of law. Therefore, we
hypothesize:
H2: Firms with political connection are associated with more optimistically biased analysts’
earnings forecasts, ceteris paribus.
9
Political Connection and Analysts’ Forecast Dispersion
Analysts’ forecast dispersion has been employed by prior studies as a proxy for the
effect of information asymmetry among investors of the forecasted firm.
Some early
researches show that forecast dispersion is likely to reflect uncertainty about the price
irrelevant component of firms' financial reports (Daley et al., 1988; Imhoff Jr. and Lobo,
1992). Others conclude later that, if forecast dispersion after (i.e., conditional on) an earnings
announcement reflects uncertainty about firms' future cash flows and this uncertainty causes
investors to desire additional information, then dispersion will be positively associated with
both (a) the level of demand for more information and (b) the magnitude of price reactions
around the subsequent earnings release (Abarbanell et al., 1995). Furthermore, Barron and
Stuerke (1998) found the dispersion in analysts' earnings forecasts serves as a useful indicator
of uncertainty about the price relevant component of firms' future earnings.
Benefits of political connection are often transferred in a covert way to avoid
unfavorable media coverage or even public outcries. The predisposition to keep the political
connection secret is likely to increase the information asymmetry for investors of firms that
benefit from such relations.
This situation is likely to lead to differences in investors'
expectations that will in turn reflect in the dispersion in analysts' forecasts (Park, 2005).
H3: Firms with political connection are more likely to be associated with more analysts’
forecast dispersion, ceteris paribus.
Political Connection, Legal Enforcement and Analysts’ Earnings Forecasts
Faccio (2006a) argues that politicians do not have equal influence in all countries and
finds that (1) leverage is significantly higher for connected firms in Malaysia, Russia, and
Thailand; (2) connected firms are subject to lower tax rates in most countries, but significantly
lower only in Russia, where connected firms enjoy an amazing 73.27 percent tax rate
discount; (3) connected firms exhibit lower ROE in all countries, but the difference is only
significant in Russia and Thailand. Interestingly, countries in which connected firms enjoy
significant political benefits are those that are often regarded as nations with less effective
legal system (La Porta et al., 1998). Likewise, we propose that the benefits and costs of
acquiring political favoritism may vary across jurisdictions. In particular, the extent of the
rule of law, the effectiveness of legal enforcement and the level of corruption can significantly
affect the distribution of costs and benefits of political connection. Ex ante, a high level of
10
rule of law increases the expected cost and decreases the expected benefit of maintaining
political connection. Ex post, effective enforcement of law imposes a higher penalty cost on
political favoritism which in turn reduces its net benefit. Similarly, corruption level serves to
indicate how easy it is to reap the benefit of political connection. In jurisdictions with a low
level of corruption, the executive branch of the government is subject to a high level of
oversight such as media monitoring and check and balance imposed by the legislative and
judiciary branches. As the net benefit of political connection decreases, so will its effect on
forecast difficulty, because connected firms will tend to resort to government favoritism less
frequently and do so at a smaller scale. Since the effectiveness of enforcement of law is
positively correlated with the extent of the rule of law and negatively correlated with
corruption level, we will hereafter use the term enforcement of law collectively to refer to
these three institutional factors.
Consistent with this view, Hope (2003) finds that financial disclosures are positively
related to forecast accuracy, suggesting that such disclosures provide useful information to
analysts. He also finds that strong legal enforcement is associated with higher forecast
accuracy. Besides, Hung (2001) reports that in countries with strong shareholder protection,
the information based on accrual accounting is more value relevant. So we expect effective
enforcement of law to mitigate the impact of political connection on firm’s earnings
generating process and therefore alleviate the forecasting task difficulty inflicted by such
connection. Therefore, we hypothesize the following:
H1a: The effect of political connection on analysts’ forecast error decreases in jurisdictions
with more effective enforcement of law, ceteris paribus.
H2a: The effect of political connection on analysts’ forecast optimistic bias decreases in
jurisdictions with more effective enforcement of law, ceteris paribus.
H3a: The effect of political connection on analysts’ forecast dispersion decreases in
jurisdictions with more effective enforcement of law, ceteris paribus.
11
III. VARIABLE MEASURMENT AND EMPIRICAL MODELS
Measuring Political Connection and Law Enforcement
We use Faccio’s (2006) definition of political connection, namely, a firm is deemed to
be connected “if one of the company’s large shareholders or top officers is: (a) a member of
parliament (MP), (b) a minister or the head of state, or (c) closely related to a top official”
(Faccio, 2006b). This variable assumes a value of 1 for connected firms and 0 otherwise.
As discussed above, the impact of political connection on analysts’ forecast task is
influenced by the expected net benefit of political favoritism which in turn is jointly
determined by the rule of law, efficiency of the judiciary system and corruption level of the
country. However, these three constructs are interrelated and as a result, empirical proxies for
them will inevitably be correlated. In order to mitigate problems arising from collinearity and
to construct a parsimony empirical model, we adopt the country-level law enforcement
measure from Leuz & Nanda & Wysocki (2003), which is the mean score across three
variables (rule of law, efficiency of judicial system, corruption index) used in La Porta et al.
(1998).
Measuring Forecast Error, Bias and Dispersion
Three empirical proxies are employed to test our hypotheses. Following prior studies
(Lang and Lundholm, 1996; Duru and Reeb, 2002; Hope, 2003), we define forecast error as
the absolute value of the difference between the forecasted and actual earnings, scaled by the
stock price at time t-1:
FORCASTt t 1  EARN t
Fcst _ Errort 
PRICEt 1
(1)
Similarly, we define analysts’ forecast optimism as follows:
Fcst _ Optimt 
FORECASTt t 1  EARN t
PRICEt 1
(2)
The standard deviation of analysts’ forecasts scaled by the stock price at time t-1 is
adopted as our measure for analysts’ forecast dispersion:
Fcst _ Disp t 
t
(3)
PRICEt 1
Where Fcst_Errort is analysts’ consensus forecast error in period t, FORECASTt t 1 is
the consensus forecast of period t earnings per share made at period time t-1, EARNt is the
12
actual earnings per share before extraordinary items for period t, PRICEt-1 is the stock price at
time of the forecast (t-1), Fcst_Optimt is analysts’ forecast optimism in period t, Fcst_Dispt.is
our measure for forecast dispersion, and σt is standard deviation of analysts’ forecasts.
Control Variables
Table 1 summarizes definition and data sources of control variables. They can be
divided into country and firm-level groups. At country level, we follow Hope (2003) and La
Porta et al. (1998) to include a number of variables to control for innate cross-country
differences that have been shown to influence financial analysts’ forecast attributes. At firm
level, our choice of control variables is guided by Francis et al. (2004), Hope (2003), Duru
and Reed (2002) and Lang and Lundholm (1996). Disclose is the Center for International
Financial Analysis and Research (CIFAR) firm-level disclosure index (Hope, 2003).
Accrual_Q is the standard deviation of the firm’s residuals from annual cross-sectional
estimations of the modified Dechow-Dichev model over the previous 5 years (Chaney et al.,
2007). We include the number of analysts following (N_Fcst) as a proxy for incentives for
analysts to reduce forecast error in their competition for reputation.
Firm size (natural
logarithm of total assets) controls for difference in information environment since large firms
tend to attract more public attention and media coverage. A number of variables are adopted
from Francis et al. (2004) to control for innate difference in earnings quality. They are
standard deviation of the firm’s rolling ten-year cash flow from operations ( σ (CFO)),
standard deviation of the firm’s rolling ten-year sales revenues (σ(Sales)), the proportion of
years reporting losses over the prior ten years (Neg_Earn), advertising expenditures over sales
revenue, and capital intensity (Capital_Intens) which is the ratio of net book value of PP&E
to total assets.10 AvgFcst is the mean of analysts’ forecasts.
Therefore, the following model is employed to test our hypotheses:
Fcst_Attribute = β0 + β1 Polit_Conn + β2 Enforcement + β3 Polit_Conn * Enforce
+ β4 N_Fcst + β5 Asset + β6 Disclose + β7 Accrual_Q+ β8 σ(CFO)
+ β9 σ(sales)+ β10 Neg_Earn + β11 Capital_Intens+ β12AvgFcst
Where the left-hand side variable, Fcst_Attribute, takes on the value of analysts’ forecast error
(Fcst_Error)), optimism (Fcst_Optim), and dispersion (Fcst_Disp), respectively; Polit_Conn
is an indicator variable with a value of 1 for politically connected firms and 0 otherwise;
Disclose is a measure for financial transparency at firm level.
10
In the interest of space, please see the cited references for a detailed discussion about these control variables.
13
IV. DATA SOURCE AND SAMPLE DESCRIPTION
Sample Selection
We constructed a one-to-all, by-country, by-industry and by-year matched sample for
empirical tests. The sample excluded industries in which there were no politically connected
observations in the country or in the year. This sampling approach minimizes the probability
that our findings may be due to industry differences between connected and non-connected
firms. The sample of this study is retrieved from Global Vantage and International I/B/E/S
databases. We start with a name list of politically connected firms generously provided by
professor Mara Faccio. There are 541 companies in 35 countries on this list. We were able to
identify 398 of them with Global Vantage information.
Those without Global Vantage
identifier (GVKEY) were not included in our sample. We further excluded companies that do
not have sufficient information either for computing analysts’ forecast attributes or innate
accounting information quality measures. As a result, our sample consists of 114 politically
connected firms with 349 firm-year observations in 17 countries between 1997 and 2000.
There are 5,368 firm-year observations of 1,895 non-connected firms as indicated in Table 2.
Apparently, there are far more non-connected firms than connected ones.
As will be
discussed, a bootstrapping procedure and one-to-one, one-to-three and one-to-five matched
samples are employed in the robustness check section to address the concern that the number
of politically connected observations may not be sufficient. It is possible that some connected
firms are misclassified into the non-connected group. However, either the small sample size
or the misclassification errors of this type would not be biased towards finding significant
results. Therefore, we do not consider them a threat to internal validity unless our results are
not statistically significant in the expected direction. Table 2 also shows UK sample represent
one third of all politically connected firms in our study and only less than 0.1% of US sample
firms have political connection which is substantially lower than the sample average of over
4%. We address the issue of possible sampling bias effect by testing our hypotheses on a
sample without UK and US firms in the robustness check section. Panel B of Table 2
indicates our sample is well distributed across different industries.
--Insert Table 2 about here --
Descriptive Statistics
14
Table 3 presents descriptive statistics after winsorizing for extreme values at one
percentile. The sample has 349 (114) politically connected firm-year observations (firms) and
5,368 (1,895) non-connected firm-year observations (firms). The mean values of all three
forecast attribute variables are well within the expected range.
Consistent with our expectation that political connection complicates the task of
earnings forecast, the mean values of forecast error, forecast optimism and forecast dispersion
are significantly larger for politically connected firms than for non-connected firms. However,
we should interpret these univariate results with caution because there is no control for
confounding factors. In addition, our hypothesized relations are based on the assumption that
the impact of political connection on forecast attributes is incremental to that of the effect of
innate earnings quality (σ(CFO), σ(Sales), Neg_Earn, Accrual_Q and Capital_Intens).
Countries with less effective law enforcement have a higher proportion of firms with political
connection as the mean and median values are significantly smaller for the connected group.
This is consistent with our argument that effective law enforcement deters corruption and
reduces the net benefit of political connection. The number of analysts following is
significantly higher for politically connected firms than for non-connected firms. Considering
the fact that the mean and median values of firm size (Asset) are higher for politically
connected firms, the difference in the number of analysts following is probably due to the fact
that connected firms are generally larger in size and more conspicuous than the non-connected
ones.
As expected, the innate accounting quality ( σ (CFO), σ (Sales), Accrual_Q and
Neg_Earn) is significantly better for the non-connected firms than the connected ones.
Politically connected firms are also more profitable as the mean and median values of the
proportion of years reporting negative earnings (Neg_Earn) are significantly smaller for the
connected group. These results are consistent with our argument that connected firms enjoy
political benefits that complicate the earnings generating process.
correlation analysis results.
--Insert Table 3 about here—
V. EMPIRICAL ANALYSIS
Correlation Analysis Results
15
Next we discuss the
Table 4 presents correlation analysis results. Spearman (Pearson) correlations are
reported in the upper (lower) diagonal. All three measures of forecast attribute (Fcst_Optim,
Fcst_Error and Fcst_Disp) are significantly positively correlated, suggesting the existence of
a common theme underlying these forecast attribute variables. Since our measure of interest
(Polit_Conn) is a dichotomous variable, we will focus on its Spearman correlations with
forecast-attribute variables. Though political connection is significantly positively correlated
with all three forecast attributes as predicted, we do not attempt to make conclusive remarks,
because these results are obtained without controlling for the effect of several factors that
have been identified by prior studies to affect analysts forecast attributes evidently. After all,
our hypotheses emphasize the incremental effect of political connection. In order to test
whether political connection incrementally complicates the earnings forecast task, we need to
employ a multiple regression analysis approach. Results concerning control variables are
consistent with our expectations. They suggest that connected firms are more prevalent in
jurisdictions with less effective law enforcement; they are followed by more analysts, larger
in size, and less transparent (Disclosure) as compared to non-connected firms. In addition,
they are also more profitable and more capital intensive but have lower accruals quality.
--Insert Table 4 about here—
Regression Results
Table 5 and 6 present regression results using analysts’ forecast attributes, namely,
Forecast Error (Panel A), Forecast Optimism (Panel B) or Forecast Dispersion (Panel C) as
dependent variables. Results in Table 5 are obtained by using Newey-West regression to test
the incremental effect of political connection (Polit_Conn) on analysts’ forecast attributes.
Results in Table 6 test the effect of political connection and the mitigating effect of
enforcement of law on analysts’ forecast attributes (Polit_Enforce). To control for clustering
effect of law enforcement score at jurisdiction level, we employ SURVEYREG procedure as
suggested by Petersen (2006) in obtaining results reported in Table 6.
Four models are reported in Panels A and C in both tables: Model 1 has only the
experimental variables (Polit_Conn in Table 5, Politi_Conn, Enforcement and Polit_Enforce
in Table 6). Model 2 includes two control variables (number of analysts’ forecasts and asset),
Model 3 includes two more control variables (disclosure level and leverage) and Model 4
includes all variables. For Panel B related to forecast optimism, we also add Model 5 which
has an additional control variable of forecast mean.
16
As shown in Table 5, consistent with our hypotheses, we find that firms with political
connection are associated with less accurate earnings forecasts, more optimistic forecast bias
and larger forecast dispersion after controlling for other variables that have been shown by
prior studies to significantly affect forecast attributes. Therefore, these results support our
hypotheses 1 to 3. In addition, as reported in Table 6, effective enforcement of law is shown
to mitigates this effect as is evidenced by the significantly negative sign of the estimated
coefficient of the interaction term between political connection and enforcement
(Polit_Enforce) in all panels. These results lend support to hypotheses 1a, 2a and 3a.
As far as the control variables are concerned, consistent with prior studies, analysts’
coverage (N_Fcst) is significantly negative. The proportion of years reporting losses in the
prior ten years (Neg_Earn) consistently shows a significantly positive sign, suggesting that
earnings of less profitable firms are more difficult to forecast. The variability of sales (σ
(Sales)) has a significantly positive effect on all analysts forecast attributes. The variability of
operating cash flows (σ (CFO)) has a significantly positive effect on analysts forecast
optimism.
Overall, the evidence provided in these tables supports our hypotheses.
Considering that these results are obtained after controlling for confounding factors at firmlevel, they suggest that political connection does have an incremental effect on complicating
the task of earnings forecast.
In addition to the statistically significant results of our experimental variables (Polit_Conn,
Polit_Enforce), the magnitude of their estimated regression coefficients also worth discussion.
The value of estimated coefficient of Polit_Conn is much larger than most control variables,
though it ranges only between 1.15 to 3.08% in Table 5, suggesting that political connection
is a more important factor in determining the attributes of analysts’ forecasts. For results in
Table 6, more interestingly, there are two similar, yet none identical interpretations. First,
effective law enforcement could almost completely offset the adverse effect of political
connection on analysts’ forecast attributes, as the value of the estimated coefficient in all three
panels is very similar for both Polit_Conn and Ploit_Enforce. This can be interpreted as that
political connection complicates the earnings forecast process only when the law enforcement
is ineffective. Second, alternatively, we could focus on enforcement of law and its interaction
term with political connection.
Here the interpretation would be that effective law
enforcement can improve the attributes of analysts’ forecasts. However, it is effective only
when firms are politically connected. In other words, the effect of law enforcement on
forecast attributes will diminish if the firm being forecasted is not politically connected. Both
interpretations are plausible and they do not necessarily contradict each other. Moreover, they
17
all point to the important role that political connection plays in the examination of financial
analysts’ forecast attributes.
--Insert Tables 5 and 6 about here-VI. ROBUSTNESS CHECKS
In this section we discuss robustness tests results. To mitigate concerns for sampling bias
effect, we replicate our tests reported in Table 6 by excluding all UK and US firms as a large
proportion of politically connected firms in our sample are from UK and the proportion of
connected firms is unusually low for the US sub-sample. Though this procedure substantially
reduces the sample size, the regression results in Table 7 do not qualitatively differ from our
main results reported in Table 6. Indeed, the estimated coefficients of political connection
(Polit_Conn) are significantly positive and its interactions with Enforcement are significantly
negative throughout all panels of Table 7. The estimated coefficients of control variables are
not qualitatively different from those in Table 6, either. Therefore, we conclude that our main
results are not driven by the large proportion of politically connected firms in the UK sample
or the lack of politically connected firms in the US.
---Insert Table 7 about here---
As only less than ten percent of our sample observations are classified as politically
connected, there is concern about non-proportional sampling effect. To mitigate this concern,
we test our hypotheses on bootstrapped samples with firm-level control variables. First, we
randomly select 13,308 observations with replacement from the firm-level model sample to
construct a random sub-sample. This random sampling procedure is repeated to produce one
thousand such sub-samples. Second, each sub-sample is used in an OLS regression analysis
to generate a set of estimated coefficients of variables in the model. Finally, we investigate
the distribution of these one thousand individually estimated coefficients for each variable and
determine the significance level by examining the sign of the estimated coefficient at
respective percentile. For example, an estimated coefficient is deemed to be significantly
greater (smaller) than zero at p < 0.01 level if its one (ninety nine) percentile value is greater
(smaller) than zero.
The bootstrapping results (not tabulated) show that except for the
interaction term between Polit_Conn and Enforcement in the forecast dispersion model, the
estimated coefficients for Polit_Conn and its interaction with Enforcement are in the expected
18
direction and significant at conventional levels throughout these one thousand regressions.
Therefore, bootstrapping procedure results are mostly consistent with our main results. To
further mitigate the concern for non-proportional sample, we also use matched sample tests.
The connected observations are matched on one-to-one, one-to-three and one-to-five basis by
country, industry, year and firm size. Results (not tabulated) remain qualitatively similar to
that in Table 5 and 6.
In order to investigate whether our main results are driven by unidentified factors that are
correlated with our classification of political connection, we randomly assign 349
observations as politically connected to match the actual distribution of politically connected
observations in our sample by country, by year and by industry. This process is repeated
1,000 times to generate as many random sub-samples. We then run 1,000 regression analyses
by using the model reported in Table 6 and sort the 1,000 estimated coefficients of each fo the
experimental variables (Polit_Conn, Polit_Enforce) in ascending order. The results (not
tabulated) do not suggest that results in Table 6 are obtained by chance. Specifically, the sign
of at least 25% of the coefficients estimated this way are not in the expected direction.
Finally, as an alternative empirical approach, we conduct Fama-MacBeth (1973) analysis
as a sensitivity test to minimize the probability of reporting inflated t-values. Table 8
summarizes the average value of coefficients estimated by seven by-year regressions. As
expected, these results are largely consistent with those reported in Table 6. Therefore, based
on these sensitivity tests results, we conclude that our main findings are not likely driven by
statistical bias or large presence of politically connected UK firms or the usually high
proportion of non-connected firms in the US.
---Insert Table 8 about her---
VII. SUMMARY AND CONCLUSION
Firms’ political connection has become a topical research issue following recent
publication of a series of important studies (Faccio and Masulis, 2005; Faccio, 2006b, 2006a;
Faccio and Parsley, 2006; Leuz and Oberholzer-Gee, 2006). We argue that the often
unexpected windfall profits arising from political connection complicate the earnings
generating process, making analysts’ forecast task more difficult. Therefore, forecast
attributes are predicted to be associated with political connection systematically. Thanks to
the availability of a list of politically connected firms generously provided by Mara Faccio,
we are able to construct an international sample to test our hypotheses. After controlling for
19
both country-level and firm-level factors that have been identified to affect forecast attributes,
we find that analysts’ forecast results are less accurate, more optimistically biased and more
diverged for connected firms than for non-connected ones. These results suggest that future
research in this area should view political connection as an important source of influence for
analysts’ forecast outcome. Another interesting result of our study is that the adverse effect of
political connection on analysts’ forecast attributes is lessened by effective enforcement of
law. This is consistent with our argument that the net benefit of political connection is
reduced by effective law enforcement which is accompanied by less corruption and more
developed market mechanisms that can serve to set boundaries on politicians’ nepotism.
However, this study is not without limitations. We are aware of the fact that our sample of
politically connected firms is relatively small and exercise extensive efforts including
matched sample tests, bootstrapping estimation, and exclusion of observations from the UK
and US to mitigate this concern. Since politically connected firms are minority in nature all
over the world, the sampling issue should not distract our main findings.
20
REFERENCES:
Abarbanell, J. S., Lanen, W. N., Verrecchia, R. E., 1995. Analysts' forecasts as proxies for
investor beliefs in empirical research. Journal of Accounting and Economics 20, 3160.
Atiase, R. K., Bamber, L. S., 1994. Trading volume reactions to annual accounting earnings
announcements: The incremental role of predisclosure information asymmetry.
Journal of Accounting and Economics 17, 309-329.
Barron, O. E., Kim, O., Lim, S. C., Stevens, D. E., 1998. Using analysts' forecasts to measure
properties of analysts' information environment. The Accounting Review 73, 421-433.
Barron, O. E., Stuerke, P. S., 1998. Dispersion in analysts' earnings forecasts as a measure of
uncertainty. Journal of Accounting, Auditing & Finance 13, 245-270.
Bhushan, R., 1989. Firm characteristics and analyst following. Journal of Accounting and
Economics 11, 255-274.
Brown, L. D., 1993. Earnings forecasting research: Its implications for capital markets
research. International Journal of Forecasting 9, 295-320.
Brown, L. D., Richardson, G. D., Schwager, S. J., 1987. An information interpretation of
financial analyst superiority in forecasting earnings. Journal of Accounting Research
25, 49-67.
Chaney, P. K., Faccio, M., Parsley, D. C., 2007. The quality of accounting information in
politically connected firms. http://ssrn.com/abstract=966379.
Chen, Q., Jiang, W., 2006. Analysts' weighting of private and public information. Review of
Financial Studies 19, 319-355.
Clarke, J., Subramanian, A., 2006. Dynamic forecasting behavior by analysts: Theory and
evidence. Journal of Financial Economics 80, 81-113.
Clement, M. B., 1999. Analyst forecast accuracy: Do ability, resources, and portfolio
complexity matter? Journal of Accounting and Economics 27, 285-303.
Core, J. E., Guay, W. R., Rusticus, T. O., 2006. Does weak governance cause weak stock
returns? An examination of firm operating performance and investors' expectations.
Journal of Finance 61, 655-687.
Cowen, A., Groysberg, B., Healy, P., 2006. Which types of analyst firms are more optimistic?
Journal of Accounting and Economics 41, 119-146.
Daley, L. A., Senkow, D. W., Vigeland, R. L., 1988. Analysts' forecasts, earnings, variability,
and option pricing: Empirical evidence. Accounting Review 63, 563.
Das, S., Levine, C. B., Sivaramakrishnan, K., 1998. Earnings predictability and bias in
analysts' earnings forecasts. The Accounting Review 73, 277-294.
Degeorge, F., Ding, Y., Jeanjean, T., Stolowy, H., 2004. Do financial analysts curb earnings
management? International evidence. Working paper, University of Lugano and HEC
School of Management, Paris.
Degeorge, F., Patel, J., Zeckhauser, R., 1999. Earnings management to exceed thresholds.
Joumal of Business 72, 1-33.
Diether, K. B., Malloy, C. J., Scherbina, A., 2002. Differences of opinion and the cross
section of stock returns. Journal of Finance 57, 2113-2141.
Dugar, A., Nathan, S., 1995. The effect of investment banking relationships on financial
analysts' earnings forecasts and investment recommendations. Contemporary
Accounting Research 12, 131-160.
Duru, A., Reeb, D. M., 2002. International diversification and analysts' forecast accuracy and
bias. The Accounting Review 77, 415-433.
Faccio, M., 2006a. The characteristics of politically connected firms.
http://ssrn.com/abstract=918244.
Faccio, M., 2006b. Politically connected firms. American Economic Review 96, 369-386.
21
Faccio, M., Masulis, R. W., 2005. The choice of payment method in European mergers and
acquisitions. Journal of Finance 60, 1345-1388.
Faccio, M., Masulis, R. W., McConnell, J. J., 2006. Political connections and corporate
bailouts. Journal of Business forthcoming.
Faccio, M., Parsley, D. C., 2006. Sudden deaths: Taking stock of political connections. ECGI
- Finance Working Paper No. 113/2006, http://ssrn.com/abstract=875808.
Fama, E. F., MacBeth, J. D., 1973. Risk, return, and equilibrium: Empirical tests. Journal of
Political Economy 81, 607-636.
Fisman, R., 2001. Estimating the value of political connections. American Economic Review
91, 1095.
Francis, J., LaFond, R., Olsson, P. M., Schipper, K., 2004. Costs of equity and earnings
attributes. Accounting Review 79, 967-1010.
Francis, J., Philbrick, D., 1993. Analysts' decisions as products of a multi-task environment.
Journal of Accounting Research 31, 216-230.
Fuller, J., Jensen, M., 2002. Just say No to wall street. Journal of Applied Corporate Finance
14, 41-46.
Ghosh, D., Whitecotton, S. M., 1997. Some determinants of analysts' forecast accuracy.
Behavioral Research in Accounting 9, 50.
Graham, J. R., Harvey, C. R., Rajgopal, S., 2005. The economic implications of corporate
financial reporting. Journal of Accounting and Economics Forthcoming.
Hand, J. R. M., 1990. A test of the extended functional fixation hypothesis. The Accounting
Review 65, 740-763.
Healy, P. M., Palepu, K. G., 2001. Information asymmetry, corporate disclosure, and the
capital markets: A review of the empirical disclosure literature. Journal of Accounting
and Economics 31, 405-440.
Hope, O.-K., 2003. Accounting policy disclosures and analysts' forecasts. Contemporary
Accounting Research 20, 295-321.
Hung, M., 2001. Accounting standards and value relevance of financial statements: An
international analysis. Journal of Accounting and Economics 30, 401-420.
Hutchcroft, P. D., 1998. Booty capitalism: The politics of banking in the philippines. Cornell
University Press (Ithaca and London).
Imhoff Jr., E. A., Lobo, G. J., 1992. The effect of ex ante earnings uncertainty on earnings
response coefficients. Accounting Review 67, 427-439.
Jacob, J., Lys, T. Z., Neale, M. A., 1999. Expertise in forecasting performance of security
analysts. Journal of Accounting and Economics 28, 51-82.
Jiang, G. H., Lee, C. M. C., Zhang, Y., 2005. Information uncertainty and expected returns.
Review of Accounting Studies 10, 185-221.
Krueger, A. O., 1974. The political economy of the rent-seeking society. American Economic
Review 64, 291-303.
Kwon, S. S., 2002. Financial analysts' forecast accuracy and dispersion: High-tech versus
low-tech stocks. Review of Quantitative Finance & Accounting 19, 65.
La Porta, R., Lopez-de-Silanes, F., Shleiffer, A., Vishny, R., 1998. Law and finance. Journal
of Political Economy 106, 1113-1155.
Lang, M. H., Lundholm, R. J., 1996. Corporate disclosure policy and analyst behavior. The
Accounting Review 71, 467-492.
Leuz, C., Nanda, D., Wysocki, P. D., 2003. Earnings management and investor protection: An
international comparison. Journal of Financial Economics 69, 505-527.
Leuz, C., Oberholzer-Gee, F., 2006. Political relationships, global financing, and corporate
transparency: Evidence from indonesia. Journal of Financial Economics 81, 411-439.
Levitt, A., 1998. The 'numbers game'. CPA Journal 68, 14-18.
22
Lim, T., 2001. Rationality and analysts' forecast bias. Journal of Finance 56, 369-385.
Lin, H. W., McNichols, M. F., 1998. Underwriting relationships, analysts' earnings forecasts
and investment recommendations. Journal of Accounting and Economics 25, 101-127.
McNichols, M., O'Brien, P. C., 1997. Self-selection and analyst coverage. Journal of
Accounting Research 35, 167-199.
O'Brien, P. C., 1990. Forecast accuracy of individual analysts in nine industries. Journal of
Accounting Research 28, 286-304.
Park, C., 2005. Stock return predictability and the dispersion in earnings forecasts. Journal of
Business 78, 2351-2375.
Petersen, M. A., 2006, Estimating standard errors in finance panel data sets: comparing
approaches, Working Paper, Kellogg School of Management, Northwestern
University.
Roberts, B. E., 1990. A dead senator tells No lies: Seniority and the distribution of federal
benefits. American Journal of Political Science 34, 31.
Zhang, X. F., 2006, Information uncertainty and analyst forecast behavior. Contemporary
Accounting Research, 23, 565-590.
Ziebart, D. A., 1990, The association between consensus of beliefs and trading activity
surrounding earnings announcements, Accounting Review, 65, 477-488.
23
Table 1
Variables used in the study
Variable
Fcst_Error
Fcst_Optim
Fcst_Disp
Polit_Conn
Enforcement
Disclose
N_Fcst
Leverage
Accrual_Q
Assets
σ(CFO)
σ(Sales)
Neg_Earn
Capital_Intens
AvgFcst
Explanation
The absolute value of the difference between mean forecast and
actual EPS scaled by stock price at the beginning of fiscal year
The difference between mean forecast and actual EPS scaled by
stock price at the beginning of fiscal year
The standard deviation of analysts’ forecasts scaled by stock price
at the beginning of fiscal year
Indicator variable for politically connected firms
Country level law enforcement measure calculated as the mean
score across three variables (rule of law, efficiency of judicial
system, corruption index) used in La Porta et al. (1998).
Firm-level total annual report disclosure scores
Number of analysts’ forecasts
The ratio of long-term debt to shareholder’s equity
Standard deviation of the firm’s residuals from year t-4 to t from
time-series cross-sectional estimations of fitted values regressing
total current accruals on change in sales, GPPE, and industries and
time dummies
Log of total assets
Standard deviation of the firm’s rolling ten-year cash flows from
operations
Standard deviation of the firm’s rolling ten-year sales revenues
Proportion of losses over the prior ten years
The ratio of the net book value of PP&E to total assets
Mean forecast
24
Data Source(s)
I/B/E/S
I/B/E/S
I/B/E/S
Faccio (2006)
Leuz et al (2003)
Hope (2003), CIFAR
I/B/E/S
Chaney et al. (2007)
Francis et al. (2004)
Francis et al. (2004)
Francis et al. (2004)
Francis et al. (2004)
Francis et al. (2004)
I/B/E/S
Table 2
Sample Distribution
This table reports distributions of politically-connected and non-connected firms across countries and industry breakdowns.
Panel A shows a distribution of 349 politically-connected firm-year observations and 5,368 non-connected firm-year
observations across countries.
Panel A: Countries, Connected and Non-Connected Firms, Legal Enforcement, and Country-level Disclosure Scores:
Connected Firms
Countries
Australia
Canada
France
Germany
Hong Kong
Indonesia
Italy
Japan
Korea
Malaysia
Singapore
Spain
Sweden
Switzerland
Taiwan
UK
USA
Total
# of firms
1
1
7
2
1
4
2
17
1
23
3
1
1
3
4
35
8
114
Firm-years
2
1
19
7
1
8
4
58
1
78
7
4
3
5
7
112
32
349
Non-Connected Firms
# of firms
4
4
36
13
1
7
4
768
2
50
8
3
4
19
11
282
679
1,895
Firm-Years
7
4
82
23
1
19
7
2,089
2
124
16
11
6
28
18
800
2,131
5,368
Legal
Enforcement
9.5
9.8
8.7
9.1
8.9
2.9
7.1
9.2
5.6
7.7
8.9
7.1
10
10
7.4
9.2
9.5
Disclosure
0.75
0.92
0.75
0.42
0.92
0.50
0.67
0.75
0.75
0.92
1.00
0.50
0.67
0.58
0.75
0.83
1.00
25
Table 2 (Cont’d)
Sample Distribution
Panel B: Industry Breakdown11
Industry
Agriculture
Automobiles and Truck
Business Services
Chemicals
Computers
Construction
Construction Materials
Consumer Goods
Defense
Electronic Equipment
Entertainment
Food Products
Machinery
Personal Services
Petroleum and Natural Gas
Pharmaceutical Products
Printing and Publishing
Recreational Products
Restaurants, Hotel, Motel
Retail
Rubber and Plastic Products
Steel Works, Etc.
Telecommunications
Textiles
Transportation
Wholesale
Total
11
Political Connection
Firm Obs.
Firm-year
Obs.
2
4
4
11
1
4
2
4
4
15
2
5
3
9
5
15
1
1
2
2
13
42
5
18
4
16
5
11
4
9
1
5
3
8
4
9
1
5
8
19
4
5
4
21
11
31
9
30
7
29
5
21
114
349
Non-Political Connection
Firm Obs.
Firm-year
Obs.
3
5
7
24
13
39
2
3
11
29
48
135
134
448
99
337
1
1
3
3
49
130
38
132
63
207
127
307
54
64
4
13
81
273
13
27
5
15
401
1,055
40
42
238
772
131
425
186
458
137
402
7
22
1,895
5,368
Note that Utilities (4900-4949) and Finance (6000-6999) are not included in our sample.
26
Table 3
Descriptive Statistics
This table reports mean, median, and standard deviation values of variables used in regression analyses from 1997 to 2001.
*, **, *** denote statistical significance at the 10 percent, 5 percent, and 1 percent levels, respectively, (two-tailed).
Fcst_Error is the absolute value of the difference between mean forecast and actual EPS scaled by stock price at the
beginning of fiscal year. Fcst_Optim is the difference between mean forecast and actual EPS scaled by stock price at the
beginning of fiscal year. Fcst_Disp is the standard deviation of analysts’ forecasts scaled by stock price at the beginning of
fiscal year. Polit_Conn equals 1 for politically connected firms and 0 otherwise. Enforcement is the country-level law
enforcement measure from Leuz, Nanda & Wysocki (2003). Disclose is a measure for CIFAR firm-level disclosure scores.
N_Fcst is the number of analysts’ forecasts. Asset is the natural log of total assets. Leverage is the ratio of long-term debt to
shareholder’s equity. Accrual_Q is the standard deviation of the firm’s residuals from years t-4 to t from time-series crosssectional estimations of the model from Chaney et al. (2007). σ(CFO) is the standard deviation of the firm’s rolling tenyear cash flows from operations. σ(Sales) is the standard deviation of the firm’s rolling ten-year sales revenues. Neg_Earn
is the proportion of number of years reporting losses over the prior ten years. Capital_Intens is the ratio of the net book
value of PP&E to total assets.
Fcst_Error
Fcst_Optim
Fcst_Disp
Enforcement
Disclose
N_Fcst
Asset
Leverage
Accrual_Q
σ(CFO)
σ(Sales)
Neg_Earn
Capital_Intens
No. of Obs
(No. of firms)
Mean
0.0636
0.0406
0.0224
0.8639
0.8240
9.3160
6.5215
0.6766
0.0779
0.1202
0.2115
0.1338
0.3325
Political
Median
0.0154
0.0051
0.0072
0.9200
0.8333
6.4444
6.7321
0.4175
0.0601
0.0952
0.1521
0.0000
0.2915
349
(114)
Std
0.11
0.11
0.04
0.12
0.12
8.30
2.10
1.04
0.05
0.09
0.19
0.21
0.22
Mean
0.0365
0.0224
0.0107
0.9245
0.8629
6.5918
5.9215
0.5857
0.0724
0.1118
0.2245
0.2100
0.2990
Non-political
Median
Std
0.0112
0.08
0.0041
0.08
0.0044
0.02
0.9200
0.05
0.8333
0.12
4.0000
6.63
5.9865
1.71
0.2486
1.20
0.0581
0.05
0.0849
0.10
0.1581
0.20
0.1000
0.28
0.2565
0.21
5,368
(1,895)
t-stat.
4.35***
2.92***
5.28***
-9.69***
-5.75***
6.01***
5.23***
1.57
1.84*
1.72*
-1.25
-6.39***
2.77***
z-stat.
3.92***
1.76*
6.80***
-16.11***
-3.93***
6.05***
6.53***
4.69***
1.82*
3.86***
-1.09
-5.18***
2.79***
27
Table 4
Correlation Analysis Results
This table reports the Pearson (above the diagonal) and the Spearman (below the diagonal) correlations among variables used in the paper for 13,419 firm-year observations in panel A
and for 5,717 firm-year observations in panel B from 1997 to 2001. Significance levels are in parentheses. Fcst_Error is the absolute value of the difference between mean forecast
and actual EPS scaled by stock price at the beginning of fiscal year. Fcst_Optim is the difference between mean forecast and actual EPS scaled by stock price at the beginning of fiscal
year. Fcst_Disp is the standard deviation of analysts’ forecasts scaled by stock price at the beginning of fiscal year. Polit_Conn equals 1 for politically connected firms and 0
otherwise. Enforcement is the country-level law enforcement measure from Leuz, Nanda & Wysocki (2003). N_Fcst is the number of analysts’ forecasts. Asset is the natural log of
total assets Leverage is the ratio of long-term debt to shareholder’s equity. Disclose is a measure CIFAR firm-level disclosure scores. Accrual_Q is the standard deviation of the
firm’s residuals from years t-4 to t from time-series cross-sectional estimations of the model from Chaney et al. (2007). σ(CFO) is the standard deviation of the firm’s rolling ten-year
cash flows from operations. σ(Sales) is the standard deviation of the firm’s rolling ten-year sales revenues. Neg_Earn is the proportion of number of years reporting losses over the
prior ten years. Capital_Intens is the ratio of the net book value of PP&E to total assets.
1
Fcst_Error
2
Fcst_Optim
3
Fcst_Disp
4
Polit_conn
5
Enforcement
6
N_Fcst
7
Asset
8
Disclose
9
Leverage
10
ACCRUAL_Q
11
σ(CFO)
12
σ(Sales)
13
Neg_Earn
14
Capital_Intens
1
1.0000
0.4803
(<.0001)
0.6640
(<.0001)
0.0519
(<.0001)
-0.3205
(<.0001)
-0.4217
(<.0001)
-0.1757
(<.0001)
-0.2981
(<.0001)
0.1615
(<.0001)
0.0076
(0.5649)
-0.0570
(<.0001)
-0.0075
(0.5721)
0.2042
(<.0001)
0.1146
(<.0001)
2
0.7980
(<.0001)
1.0000
0.2613
(<.0001)
0.0233
(0.0787)
-0.1905
(<.0001)
-0.2162
(<.0001)
-0.0595
(<.0001)
-0.2248
(<.0001)
0.0803
(<.0001)
-0.0120
(0.3634)
-0.0819
(<.0001)
0.0106
(0.4239)
0.0208
(0.1153)
0.0240
(0.0702)
3
0.6043
(<.0001)
0.3770
(<.0001)
1.0000
0.0999
(<.0001)
-0.3994
(<.0001)
-0.2727
(<.0001)
-0.1135
(<.0001)
-0.3612
(<.0001)
0.2233
(<.0001)
0.0071
(0.6276)
-0.0282
(0.0547)
-0.0363
(0.0135)
0.2047
(<.0001)
0.2210
(<.0001)
4
0.0807
(<.0001)
0.0534
(<.0001)
0.1248
(<.0001)
1.0000
-0.2130
(<.0001)
0.0801
(<.0001)
0.0864
(<.0001)
-0.0520
(<.0001)
0.0620
(<.0001)
0.0241
(0.0683)
0.0510
(0.0001)
-0.0144
(0.2768)
-0.0685
(<.0001)
0.0369
(0.0053)
5
-0.2305
(<.0001)
-0.1357
(<.0001)
-0.3269
(<.0001)
-0.2444
(<.0001)
1.0000
0.2350
(<.0001)
0.0447
(0.0007)
0.7729
(<.0001)
-0.1970
(<.0001)
0.0147
(0.2675)
0.2195
(<.0001)
0.0540
(<.0001)
0.2487
(<.0001)
-0.2572
(<.0001)
6
-0.2054
(<.0001)
-0.1529
(<.0001)
-0.1395
(<.0001)
0.0962
(<.0001)
-0.0099
(0.4549)
1.0000
0.4082
(<.0001)
0.2982
(<.0001)
0.0165
(0.2119)
0.0157
(0.2358)
0.1329
(<.0001)
0.0648
(<.0001)
-0.1454
(<.0001)
-0.0239
(0.0707)
7
-0.1238
(<.0001)
-0.0804
(<.0001)
-0.1228
(<.0001)
0.0824
(<.0001)
0.1783
(<.0001)
0.4095
(<.0001)
1.0000
0.0377
(0.0043)
0.2231
(<.0001)
-0.1872
(<.0001)
-0.1091
(<.0001)
-0.1906
(<.0001)
-0.1888
(<.0001)
0.0998
(<.0001)
8
-0.1733
(<.0001)
-0.1577
(<.0001)
-0.1749
(<.0001)
-0.0760
(<.0001)
0.3749
(<.0001)
0.2626
(<.0001)
0.1230
(<.0001)
1.0000
-0.2723
(<.0001)
0.1327
(<.0001)
0.4240
(<.0001)
0.1376
(<.0001)
0.2145
(<.0001)
-0.2639
(<.0001)
9
0.1236
(<.0001)
0.0671
(<.0001)
0.1453
(<.0001)
0.0183
(0.1672)
-0.0656
(<.0001)
-0.0471
(0.0004)
-0.0246
(0.0625)
-0.1828
(<.0001)
1.0000
-0.1278
(<.0001)
-0.1696
(<.0001)
-0.0164
(0.2165)
0.0490
(0.0002)
0.3912
(<.0001)
10
0.0347
(0.0087)
0.0197
(0.1374)
0.0426
(0.0038)
0.0262
(0.0480)
-0.0312
(0.0183)
-0.0125
(0.3439)
-0.1874
(<.0001)
0.0940
(<.0001)
-0.0478
(0.0003)
1.0000
0.5126
(<.0001)
0.3311
(<.0001)
0.0688
(<.0001)
-0.2347
(<.0001)
11
0.0222
(0.0927)
-0.0182
(0.1699)
0.0432
(0.0033)
0.0209
(0.1135)
0.0221
(0.0945)
0.0395
(0.0028)
-0.1521
(<.0001)
0.2897
(<.0001)
-0.0984
(<.0001)
0.5016
(<.0001)
1.0000
0.3229
(<.0001)
0.2797
(<.0001)
-0.3151
(<.0001)
12
0.0485
(0.0002)
0.0404
(0.0022)
0.0289
(0.0495)
-0.0157
(0.2346)
0.0024
(0.8561)
0.0153
(0.2475)
-0.1799
(<.0001)
0.0580
(<.0001)
0.0695
(<.0001)
0.3046
(<.0001)
0.2554
(<.0001)
1.0000
0.0408
(0.0021)
-0.2387
(<.0001)
13
0.1469
(<.0001)
0.0551
(<.0001)
0.1395
(<.0001)
-0.0670
(<.0001)
0.1234
(<.0001)
-0.1481
(<.0001)
-0.2177
(<.0001)
0.2664
(<.0001)
0.0843
(<.0001)
0.0754
(<.0001)
0.4102
(<.0001)
-0.0040
(0.7608)
1.0000
(-0.1088)
<.0001
14
0.0490
(0.0002)
-0.0063
(0.6327)
0.1058
(<.0001)
0.0378
(<.0042)
-0.1521
(<.0001)
0.0326
(0.0137)
0.0746
(<.0001)
-0.1809
(<.0001)
0.2499
(<.0001)
-0.2168
(<.0001)
-0.2845
(<.0001)
-0.2200
(<.0001)
-0.1373
(<.0001)
1.0000
28
Table 5
Newey-West Regression Results
Fcst_Error is the absolute value of the difference between mean forecast and actual EPS scaled by stock price at the
beginning of fiscal year. Fcst_Optim is the difference between mean forecast and actual EPS scaled by stock price at the
beginning of fiscal year. Fcst_Disp is the standard deviation of analysts’ forecasts scaled by stock price at the beginning of
fiscal year. Polit_Conn equals 1 for politically connected firms and 0 otherwise. Enforcement is the country-level law
enforcement measure from Leuz, Nanda, and Wysocki (2003). N_Fcst is the number of analysts’ forecasts. Asset is the
natural log of total assets. Disclose_firm is a measure for CIFAR firm-level disclosure scores. Leverage is the ratio of
long-term debt to shareholder’s equity. Accrual_Q is the standard deviation of the firm’s residuals from years t-4 to t from
time-series cross-sectional estimations of the model from Chaney et al. (2007). σ(CFO) is the standard deviation of the
firm’s rolling ten-year cash flows from operations. σ(Sales) is the standard deviation of the firm’s rolling ten-year sales
revenues. Neg_Earn is the proportion of number of years reporting losses over the prior ten years. Capital_Intens is the
ratio of the net book value of PP&E to total assets. Year dummies are included in the regression but not tabulated.
Reported are the coefficient estimates and t-statistics in parentheses. t-values are based on clustered standard errors
(Petersen, 2006). *, **, *** denote statistical significance at the 10 percent, 5 percent, and 1 percent levels, respectively,
(one-tailed).
Panel A: Dependent Variable is Forecast Error (Fcst_Error)
Independent
Variables
Pred.
Sign
Intercept
Model 1
Model 2
Model 3
Model 4
+
0.0279***
(10.56)
0.0605***
(11.03)
0.1097***
(9.64)
0.1059***
(8.49)
Polit_Conn
+
0.0275***
(4.02)
0.0349***
(5.26)
0.0307***
(4.58)
0.0308***
(4.64)
N_Fcst
-
-0.0023***
(-14.44)
-0.0019***
(-11.23)
-0.0017***
(-10.05)
Asset
-
-0.0025***
(-3.01)
-0.0023***
(-2.89)
0.0000
(0.01)
Disclo_Firm
-
-0.0661***
(-5.94)
-0.1053***
(-8.08)
Leverage
+
0.0064***
(4.33)
0.0037***
(2.46)
Accrual_Q
+
σ(CFO)
+
σ(Sales)
+
0.0245***
(3.39)
Neg_Earn
+
0.0523***
(9.68)
Capital_Intens
+
0.0194***
(3.16)
0.0011***
(2.30)
-0.0111
(-0.63)
Adjusted R2
0.0077
0.0543
0.0757
0.1029
No. of firms
(Connected)
2,009
(114)
2,009
(114)
2,009
(114)
2,009
(114)
No. of obs.
(Connected)
5,717
(349)
5,717
(349)
5,717
(349)
5,717
(349)
29
Table 5 (Cont’d)
Newey-West Regression Results
Panel B: Dependent Variable is Forecast Optimism (Fcst_Opti)
Independent
Variables
Pred.
Sign
Intercept
+
Polit_Conn
+
N_Fcst
Model 1
Model 2
Model 3
Model 4
Model 5
0.0207***
(7.66)
0.0421***
(7.60)
0.1005***
(8.64)
0.1031***
(8.01)
0.1032***
(8.02)
0.0184***
(2.79)
0.0238***
(3.62)
0.0196***
(2.93)
0.0203***
(3.04)
0.0204***
(3.05)
-
-0.0018***
(-11.73)
-0.0014***
(-8.42)
-0.0013***
(-7.56)
-0.0012***
(-7.30)
Asset
-
-0.0013*
(-1.56)
Disclo_Firm
-
-0.0745***
(-6.50)
-0.0947***
(-6.92)
-0.0946***
(-6.92)
Leverage
+
0.0028**
(1.98)
0.0019*
(1.31)
0.0019*
(1.31)
Accrual_Q
+
0.0008*
(1.64)
0.0008*
(1.63)
σ(CFO)
+
-0.0318**
(-1.73)
-0.0320**
(-1.75)
σ(Sales)
+
0.0191***
(2.68)
0.0191***
(2.68)
Neg_Earn
+
0.0270***
(5.20)
0.0259***
(4.60)
Capital_Intens
+
AvgFcst
+
-0.0011*
(-1.40)
0.0002
(0.22)
-0.0069
(-1.11)
0.0002
(0.26)
-0.0069
(-1.11)
-0.0008
(-0.79)
Adjusted R2
0.0037
0.0284
0.0427
0.0501
0.0500
No. of firms
(Connected)
2,009
(114)
2,009
(114)
2,009
(114)
2,009
(114)
2,009
(114)
No. of obs.
(Connected)
5,717
(349)
5,717
(349)
5,717
(349)
5,717
(349)
5,717
(349)
30
Table 5 (Cont’d)
Newey-West Regression Results
Panel C: Dependent Variable is Forecast Dispersion (Fcst_Disp)
Independent
Variables
Pred.
Sign
Intercept
+
Polit_Conn
+
N_Fcst
Model 1
Model 2
Model 3
Model 4
0.0073***
(8.39)
0.0178***
(8.91)
0.0330***
(8.46)
0.0299***
(6.90)
0.0118***
(4.52)
0.0134***
(5.19)
0.0117***
(4.48)
0.0115***
(4.41)
-
-0.0004***
(-6.64)
-0.0003***
(-5.29)
-0.0003***
(-4.57)
Asset
-
-0.0011***
(-3.78)
-0.0010***
(-3.32)
-0.0003
(-1.06)
Disclo_Firm
-
-0.0213***
(-5.78)
-0.0328***
(-7.49)
Leverage
+
0.0024***
(4.21)
0.0016***
(2.74)
Accrual_Q
+
0.0004***
(2.78)
σ(CFO)
+
0.0032
(0.53)
σ(Sales)
+
0.0031*
(1.40)
Neg_Earn
+
0.0162***
(7.64)
Capital_Intens
+
0.0123***
(6.10)
Adjusted R2
0.0188
0.0474
0.0765
0.1135
No. of firms
(Connected)
1,622
(105)
1,622
(105)
1,622
(105)
1,622
(105)
No. of obs.
(Connected)
4,631
(308)
4,631
(308)
4,631
(308)
4,631
(308)
31
Table 6
Regression Results Corrected for Clustering Effect at Country Level
Fcst_Error is the absolute value of the difference between mean forecast and actual EPS scaled by stock price at the beginning of fiscal
year. Fcst_Optim is the difference between mean forecast and actual EPS scaled by stock price at the beginning of fiscal year. Fcst_Disp
is the standard deviation of analysts’ forecasts scaled by stock price at the beginning of fiscal year. Polit_Conn equals 1 for politically
connected firms and 0 otherwise. Enforcement is the country-level law enforcement measure from Leuz, Nanda, and Wysocki (2003).
N_Fcst is the number of analysts’ forecasts. Asset is the natural log of total assets. Disclose_firm is a measure for CIFAR firm-level
disclosure scores. Leverage is the ratio of long-term debt to shareholder’s equity. Accrual_Q is the standard deviation of the firm’s
residuals from years t-4 to t from time-series cross-sectional estimations of the model from Chaney et al. (2007). σ(CFO) is the standard
deviation of the firm’s rolling ten-year cash flows from operations. σ(Sales) is the standard deviation of the firm’s rolling ten-year sales
revenues. Neg_Earn is the proportion of number of years reporting losses over the prior ten years. Capital_Intens is the ratio of the net
book value of PP&E to total assets. Year dummies are included in the regression but not tabulated. Reported are the coefficient estimates
and t-statistics in parentheses. t-values are based on clustered standard errors (Petersen, 2006). *, **, *** denote statistical significance
at the 10 percent, 5 percent, and 1 percent levels, respectively, (one-tailed).
Panel A: Dependent Variable is Forecast Error (Fcst_Error)
Independent
Variables
Pred.
Sign
Intercept
+
Polit_Conn
Model 1
Model 2
Model 3
Model 4
0.2755***
(4.36)
0.2897***
(5.41)
0.2756***
(4.92)
0.2856***
(4.62)
+
0.1345***
(2.43)
0.1452***
(3.74)
0.1616***
(4.72)
0.1499***
(4.69)
Enforcement
-
-0.2683***
(-3.92)
-0.2628***
(-4.42)
-0.2326***
(-3.07)
-0.2444***
(-2.97)
Polit_Enforce
-
-0.1426***
(-2.39)
-0.1471***
(-3.37)
-0.1660***
(-4.12)
-0.1524***
(-3.94)
N_Fcst
-
-0.0025***
(- 4.71)
-0.0023***
(- 3.59)
-0.0021***
(- 4.66)
Asset
-
Disclo_Firm
-0.0001
( -0.10)
0.0023**
(2.20)
-
-0.0211
(-0.75)
-0.0595***
(-2.30)
Leverage
+
0.0064***
( 6.37)
Accrual_Q
+
0.0009
(1.27)
σ(CFO)
+
-0.0258
(-0.87)
σ(Sales)
+
0.0271***
(3.43)
Neg_Earn
+
0.0555**
( 1.89)
Capital_Intens
+
0.0111
( 0.88)
0.0000
( 0.02)
0.0038***
(3.67)
R2
0.0580
0.1013
0.1120
0.1407
No. of firms
(Connected)
2,009
(114)
2,009
(114)
2,009
(114)
2,009
(114)
No. of obs.
(Connected)
5,717
(349)
5,717
(349)
5,717
(349)
5,717
(349)
32
Table 6 (Cont’d)
Regression Results Corrected for Clustering Effect at Country Level
Panel B: Dependent Variable is Forecast Optimism (Fcst_Opti)
Independent
Variables
Pred.
Sign
Intercept
+
Polit_Conn
Model 1
Model 2
Model 3
Model 4
Model 5
0.1518***
(2.46)
0.1624***
(3.45)
0.1539***
(3.74)
0.1704***
(3.94)
0.1703***
(3.94)
+
0.1399*
(1.46)
0.1484**
(1.74)
0.1841***
(2.74)
0.1758***
(2.73)
0.1758***
(2.73)
Enforcement
-
-0.1420**
(-2.08)
-0.1392***
(-2.57)
-0.0797*
(-1.54)
-0.0956**
(-1.72)
-0.0955**
(-1.72)
Polit_Enforce
-
-0.1505*
(-1.42)
-0.1546*
(-1.63)
-0.1955***
(-2.60)
-0.1857***
(-2.58)
-0.1857***
(-2.58)
N_Fcst
-
-0.0019***
(- 4.04)
-0.0016***
(- 2.86)
-0.0015***
(- 3.40)
-0.0015***
(- 3.53)
Asset
-
0.0014**
(1.69)
0.0014**
(1.66)
Disclo_Firm
-
-0.0571***
(-2.67)
-0.0748***
(-3.86)
-0.0748***
(-3.87)
Leverage
+
0.0028**
( 1.68)
0.0019**
(1.72)
0.0019**
(1.71)
Accrual_Q
+
0.0007
(1.00)
0.0007
(1.00)
σ(CFO)
+
-0.0393**
(-1.66)
-0.0393**
(-1.66)
σ(Sales)
+
0.0204***
(2.88)
0.0204***
(2.88)
Neg_Earn
+
0.0283*
( 1.33)
0.0281*
( 1.36)
Capital_Intens
+
-0.0107*
( -1.39)
-0.0107*
( -1.39)
AvgFcst
+
0.0003
( 0.26)
0.0000
( 0.02)
-0.0001
( -0.08)
R2
0.0227
0.0466
0.0548
0.0640
0.0640
No. of firms
(Connected)
2,009
(114)
2,009
(114)
2,009
(114)
2,009
(114)
2,009
(114)
No. of obs.
(Connected)
5,717
(349)
5,717
(349)
5,717
(349)
5,717
(349)
5,717
(349)
33
Table 6 (Cont’d)
Regression Results Corrected for Clustering Effect at Country Level
Panel C: Dependent Variable is Forecast Dispersion (Fcst_Disp)
Independent
Variables
Pred.
Sign
Intercept
+
Polit_Conn
Model 1
Model 2
Model 3
Model 4
0.1113***
(6.08)
0.1163***
(6.38)
0.1110***
(5.71)
0.1097***
(5.35)
+
0.0325**
(1.83)
0.0331**
(2.15)
0.0344***
(2.47)
0.0322***
(2.35)
Enforcement
-
-0.1127***
(-5.72)
-0.1127***
(-5.67)
-0.1084***
(-4.12)
-0.1080***
(-3.90)
Polit_Enforce
-
-0.0318**
(-1.74)
-0.0311**
(-1.88)
-0.0327**
(-2.05)
-0.0301**
(-1.92)
N_Fcst
-
-0.0005***
(- 4.68)
-0.0005***
(- 3.78)
-0.0004***
(- 4.99)
Asset
-
-0.0001
(- 0.36)
-0.0001
(- 0.02)
Disclo_Firm
-
-0.0004
(-0.04)
Leverage
+
0.0024***
( 4.25)
Accrual_Q
+
σ(CFO)
+
σ(Sales)
+
0.0049***
(3.82)
Neg_Earn
+
0.0174***
( 2.34)
Capital_Intens
+
0.0089**
( 1.90)
0.0006***
(2.32)
-0.0128
(-1.17)
0.0015***
(3.21)
0.0003*
(1.41)
-0.0028
(-0.26)
R2
0.1146
0.1378
0.1503
0.1852
No. of firms
(Connected)
1,622
(105)
1,622
(105)
1,622
(105)
1,622
(105)
No. of obs.
(Connected)
4,631
(308)
4,631
(308)
4,631
(308)
4,631
(308)
34
Table 7
Regression Results Corrected for Clustering Effect at Country Level without US and UK Firms
Fcst_Error is the absolute value of the difference between mean forecast and actual EPS scaled by stock price at the beginning of fiscal
year. Fcst_Optim is the difference between mean forecast and actual EPS scaled by stock price at the beginning of fiscal year. Fcst_Disp
is the standard deviation of analysts’ forecasts scaled by stock price at the beginning of fiscal year. Polit_Conn equals 1 for politically
connected firms and 0 otherwise. Enforcement is the country-level law enforcement measure from Leuz & Nanda & Wysocki (2003).
N_Fcst is the number of analysts’ forecasts. Asset is the natural log of total assets. Disclose_firm is a measure for CIFAR firm-level
disclosure scores. Leverage is the ratio of long-term debt to shareholder’s equity. Accrual_Q is the standard deviation of the firm’s
residuals from years t-4 to t from time-series cross-sectional estimations of the model from Chaney et al. (2007). σ(CFO) is the standard
deviation of the firm’s rolling ten-year cash flows from operations. σ(Sales) is the standard deviation of the firm’s rolling ten-year sales
revenues. Neg_Earn is the proportion of number of years reporting losses over the prior ten years. Capital_Intens is the ratio of the net
book value of PP&E to total assets. Year dummies are included in the regression but not tabulated. t-values are based on clustered
standard errors (Petersen, 2006). Reported are the coefficient estimates and t-statistics in parentheses. *, **, and *** denote statistical
significance at the 10 percent, 5 percent, and 1 percent levels, respectively, (one-tailed).
Panel A: Dependent Variable is Forecast Error (Fcst_Error)
Independent
Variables
Pred.
Sign
Model 1
Model 2
***
0.3507
(4.26)
***
Model 3
0.2594
(3.70)
***
Model 4
0.2247***
(4.61)
Intercept
+
0.2640
(4.30)
Polit_Conn
+
0.1997***
(3.82)
0.1649***
(2.48)
0.1590**
(2.14)
0.1114***
(2.38)
Enforcement
-
-0.2360***
(-4.46)
-0.2942***
(-3.54)
-0.3033***
(-5.30)
-0.2937***
(-7.16)
Polit_Enforce
-
-0.2107***
(-3.16)
-0.1542**
(-1.97)
-0.1524**
(-1.81)
-0.0998**
(-1.87)
N_Fcst
-
-0.0041***
(- 9.65)
-0.0004***
(- 9.98)
-0.0039***
(- 9.27)
Asset
-
-0.0015
( -0.70)
-0.0012
( -0.74)
0.0002
(0.20)
Disclo_Firm
-
0.1189
( 1.12)
0.1056
( 1.17)
Leverage
+
0.0074***
( 5.28)
0.0014
( 0.91)
Accrual_Q
+
0.0016**
(1.70)
σ(CFO)
+
0.1530*
( 1.55)
σ(Sales)
+
0.0304*
(1.57)
Neg_Earn
+
0.1430***
( 9.63)
Capital_Intens
+
-0.0271***
( -2.60)
R2
0.0467
0.0781
0.0912
0.1533
No. of firms
(Connected)
1,005
(71)
1,005
(71)
1,005
(71)
1,005
(71)
No. of obs.
(Connected)
2,642
(205)
2,642
(205)
2,642
(205)
2,642
(205)
35
Table 7 (Cont’d)
Regression Results Corrected for Clustering Effect at Country Level without US and UK Firms
Panel B: Dependent Variable is Forecast Optimism (Fcst_Opti)
Independent
Variables
Pred.
Sign
Intercept
+
Polit_Conn
Model 1
Model 2
Model 3
Model 4
Model 5
0.1260***
(2.63)
0.1912***
(2.87)
0.1407***
(2.32)
0.1298***
(2.66)
0.1300***
(2.67)
+
0.2129***
(2.30)
0.1868**
(1.83)
0.1848**
(1.76)
0.1476**
(1.77)
0.1477**
(1.77)
Enforcement
-
-0.0878***
(-2.53)
-0.1317***
(-2.27)
-0.1353***
(-3.22)
-0.1382***
(-4.33)
-0.1391***
(-4.27)
Polit_Enforce
-
-0.2305**
(-2.12)
-0.1881*
(-1.60)
-0.1887*
(-1.57)
-0.1476*
(-1.55)
-0.1478*
(-1.55)
N_Fcst
-
-0.0031***
(- 10.11)
-0.0031***
(- 9.08)
-0.0030***
(-8.02)
-0.0030***
(- 7.81)
Asset
-
-0.0011
( -0.38)
-0.0011
( -1.03)
-0.0001
(-0.14)
-0.0001
(-0.13)
Disclo_Firm
-
0.0676
( 0.90)
0.0621
( 0.94)
0.0630
( 0.94)
Leverage
+
0.0025*
( 1.62)
-0.0006
(-0.37)
-0.0006
(-0.37)
Accrual_Q
+
0.0016**
(1.81)
0.0016**
(1.81)
σ(CFO)
+
0.0931
( 1.12)
0.0926
( 1.11)
σ(Sales)
+
0.0154
(0.57)
0.0155
(0.58)
Neg_Earn
+
0.0938***
( 6.08)
0.0940***
( 6.04)
Capital_Intens
+
-0.0427***
( -3.07)
-0.0428***
( -3.03)
AvgFcst
+
0.0002
( 0.36)
R2
0.0172
0.0347
0.0372
0.0688
0.0688
No. of firms
(Connected)
1,005
(71)
1,005
(71)
1,005
(71)
1,005
(71)
1,005
(71)
No. of obs.
(Connected)
2,642
(205)
2,642
(205)
2,642
(205)
2,642
(205)
2,642
(205)
36
Table 7 (Cont’d)
Regression Results Corrected for Clustering Effect at Country Level without US and UK Firms
Panel C: Dependent Variable is Forecast Dispersion (Fcst_Disp)
Independent
Variables
Pred.
Sign
Intercept
+
Polit_Conn
Model 2
Model 3
0.0955***
(6.63)
0.1158***
(7.62)
0.0770***
(3.11)
0.0655***
(3.13)
+
0.0744***
(3.27)
0.0658***
(2.73)
0.0627***
(2.33)
0.0500**
(2.18)
Enforcement
-
-0.0931***
(-6.30)
-0.1074***
(-5.92)
-0.1117***
(-9.27)
-0.1068***
(-8.53)
Polit_Enforce
-
-0.0787***
(-2.71)
-0.0653**
(-2.15)
-0.0637**
(-1.95)
-0.0489**
(-1.72)
N_Fcst
-
-0.0009***
(- 3.13)
-0.0009***
(- 3.51)
-0.0008***
(- 3.05)
Asset
-
-0.0002
(- 0.30)
0.0000
( 0.04)
0.0003
(0.68)
Disclo_Firm
-
0.0504*
( 1.37)
0.0436
( 1.27)
Leverage
+
0.0031***
( 3.65)
0.0006
(0.77)
Accrual_Q
+
0.0007***
(2.53)
σ(CFO)
+
0.0116
( 0.60)
σ(Sales)
+
0.0060***
(2.44)
Neg_Earn
+
0.0545***
( 10.54)
Capital_Intens
+
0.0034
( 0.60)
R2
Model 1
Model 4
0.0791
0.0944
0.1152
0.1610
No. of firms
(Connected)
722
(65)
722
(65)
722
(65)
722
(65)
No. of obs.
(Connected)
1886
(174)
1886
(174)
1886
(174)
1886
(174)
37
Table 8
Fama-MacBeth Regression Results
Fcst_Error is the absolute value of the difference between mean forecast and actual EPS scaled by stock price at the
beginning of fiscal year. Fcst_Optim is the difference between mean forecast and actual EPS scaled by stock price at the
beginning of fiscal year. Fcst_Disp is the standard deviation of analysts’ forecasts scaled by stock price at the beginning of
fiscal year. Polit_Conn equals 1 for politically connected firms and 0 otherwise. Enforcement is the country-level law
enforcement measure from Leuz, Nanda, and Wysocki (2003). N_Fcst is the number of analysts’ forecasts. Asset is the
natural log of total assets. Disclos is a measure for CIFAR firm-level disclosure scores. Leverage is the ratio of long-term
debt to shareholder’s equity. Accrual_Q is the standard deviation of the firm’s residuals from years t-4 to t from time-series
cross-sectional estimations of the model from Chaney et al. (2007). σ(CFO) is the standard deviation of the firm’s rolling tenyear cash flows from operations. σ(Sales) is the standard deviation of the firm’s rolling ten-year sales revenues. Neg_Earn is
the proportion of number of years reporting losses over the prior ten years. Capital_Intens is the ratio of the net book value of
PP&E to total assets. *, **, *** denote statistical significance at the 10 percent, 5 percent, and 1 percent levels, respectively,
(one-tailed). Year dummies are not tabulated.
Constant
Polit_Conn
Enforcement
Polit_enforce
N_fcst
Asset
Disclos
Leverage
Accrual_Q
σ(CFO)
σ(Sales)
Neg_Earn
Capital_Intens
Fcst_Error
Coef.
Coef.
(t-stat.)
(t-stat.)
0.1012***
0.2912***
(5.80)
(23.66)
0.0319***
0.1815**
(4.95)
(2.58)
-0.2624***
(-8.60)
-0.1876**
(-2.50)
-0.0015**
-0.0020***
(-3.27)
(-4.93)
-0.0001
0.0025**
(-0.09)
(2.90)
-0.0983***
-0.0480*
(-4.70)
(-1.81)
0.0039**
0.0043***
(3.65)
(3.98)
0.0012**
0.0012**
(3.03)
(3.07)
-0.0127
-0.0317*
(-0.81)
(-1.94)
0.0244***
0.0284***
(4.32)
(6.15)
0.0620***
0.0664***
(7.35)
(7.11)
0.0155**
0.0062
(2.15)
(0.69)
Coef.
(t-stat.)
0.0986***
(9.57)
0.0224**
(2.63)
-0.0011**
(-2.93)
0.0000
(-0.04)
-0.0901***
(-5.07)
0.0030
(1.43)
0.0009**
(2.61)
-0.0429**
(-2.27)
0.0175**
(3.06)
0.0350***
(5.40)
-0.0120
(-1.07)
Ave_Fcst
Mean Adj. R2
N=5
0.1166
0.0667
Fcst_Optim
Coef.
Coef.
(t-stat.)
(t-stat.)
0.1804***
0.0968***
(5.76)
(8.16)
0.2345**
0.0225**
(2.74)
(2.64)
-0.1185**
(-2.50)
-0.2511**
(-2.76)
-0.0014***
-0.0011**
(-4.23)
(-2.72)
0.0016**
0.0001
(2.18)
(0.09)
-0.0660**
-0.0872***
(-2.60)
(-4.29)
0.0033
0.0030
(1.51)
(1.44)
0.0010**
0.0009**
(2.57)
(2.57)
-0.0526**
-0.0433**
(-2.56)
(-2.30)
0.0203***
0.0178**
(4.01)
(3.23)
0.0372***
0.0331***
(5.42)
(5.80)
-0.0161
-0.0123
(-1.28)
(-1.07)
-0.0024
(-1.21)
0.0683
Coef.
(t-stat.)
0.1781***
(5.60)
0.2338**
(2.76)
-0.1164**
(-2.45)
-0.2502**
(-2.78)
-0.0014***
(-3.95)
0.0016**
(2.36)
-0.0651**
(-2.43)
0.0033
(1.51)
0.0010**
(2.60)
-0.0525**
(-2.56)
0.0205***
(4.10)
0.0365***
(5.84)
-0.0163
(-1.29)
-0.0011
(-0.76)
Fcst_Disp
Coef.
Coef.
(t-stat.)
(t-stat.)
0.0330***
0.1186***
(7.59)
(5.37)
0.0118***
0.0424*
(4.88)
(1.60)
-0.1173***
(-3.92)
-0.0421
(-1.45)
-0.0002
-0.0004**
(-1.41)
(-3.68)
-0.0004
0.0006**
(-0.94)
(2.75)
-0.0329***
-0.0104
(-5.14)
(-0.95)
0.0012*
0.0013**
(1.98)
(2.63)
0.0004**
0.0004***
(3.41)
(4.89)
0.0050
-0.0044
(1.09)
(-1.53)
0.0025
0.0048**
(1.25)
(3.47)
0.0180***
0.0198***
(18.19)
(11.75)
0.0109**
0.0067
(2.51)
(1.26)
0.1286
38
Download