The Effect of Media Competition on Analyst Forecast Properties: Cross-Country Evidence

Abstract

We examine the effect of media competition on analyst forecast properties in an international setting using 113,436 firm-year observations from 32 countries spanning 2000 through 2012. We find that firms in countries with stronger media competition enjoy more accurate, less optimistically biased, and less dispersed analyst forecasts. The effects of media competition on the properties of analyst forecasts are stronger for firms with lower institutional ownership, for firms followed by fewer analysts or by analysts from smaller brokerage houses, and for firms with weaker financial performance. This suggests that media competition plays a more pronounced role in shaping the information environment when information from nonmedia channels is likely to be limited or of lower quality. Finally, we find that analysts in countries with stronger media competition tend to follow more firms, suggesting that stronger media competition reduces analysts’ information acquisition costs, which in turn, improves the properties of their forecasts.

Keywords

media competition analyst forecasts information environment

Introduction

The public media serves as one of the main intermediaries that discovers, collects, interprets, and disseminates information to capital market participants in many countries around the world (Bushee et al., 2010; The World Bank Institute, 2002). Prior studies focus mainly on media penetration or media coverage, which capture the amount of information available to capital market participants.¹ In this article, we focus on media competition, which is related to the quality of the information produced by the media, and investigate the extent to which the level of media competition within a country influences the quality of the information available to capital market participants, proxied for by the properties of analysts’ earnings forecasts. Our study adds to the growing body of research on the important role played by media characteristics in the capital markets.

Theory and empirical evidence suggest that media competition enhances information quality (Dyck & Zingales, 2002; Fan, 2013; Gentzkow and Shapiro 2006, 2008; Lacy & Simon, 1993). Other studies find, however, that pressures arising from competition can reduce information quality because the media may be forced to cut investments and costs in response to lower profits (Baron, 2010; Becker et al., 2009; Cagé, 2020; Hollifield et al., 2004; Yildirim et al., 2013). Because the media is an important information source for financial analysts, media competition can improve or reduce the quality of analyst forecasts, making the influence of media competition on the properties of analyst forecasts an empirical question.

We examine this question by taking advantage of cross-country variation in media competition. In our main tests, we construct two country-level media competition variables using data from Djankov et al. (2003)—The primary competition measure is based on the market share of the top five newspapers in the country and the secondary one is based on the ultimate ownership of the top five newspapers in the country. Our focus on media competition is consistent with research that uses newspaper circulation to measure media coverage or media penetration (e.g., Bandyopadhyay, 2006; Bushman et al., 2004; Dyck & Zingales, 2004; Lang et al., 2012; Qi et al., 2010) and with studies that demonstrate the importance of newspapers in disseminating information and influencing the capital markets (e.g., Engelberg & Parsons, 2011; Fang & Peress, 2009; Peress, 2014).

Using 113,436 analyst forecasts for 19,490 unique firms in 32 countries, spanning 2000 through 2012, controlling for the effects of media penetration and media freedom, we find that analyst forecasts are more accurate, less optimistically biased, and less dispersed for firms from countries with stronger media competition. These findings suggest that stronger media competition improves the information environment, thereby facilitating better analyst forecasts. The media effects seem economically significant; for a one-standard deviation change in media competition, the change in forecast properties ranges from 4.2% to 19.1%. In additional analyses, we find that the effect of media competition on the properties of analyst forecasts is stronger for firms with lower institutional ownership, for firms followed by fewer analysts or by analysts from smaller brokerage houses, and for firms with weaker financial performance. Together, these findings suggest that the media plays a more pronounced role in shaping the information environment when information from nonmedia channels is likely to be limited or of lower quality.

Because the media data from Djankov et al. (2003) are time-invariant, we conduct additional analyses based on a subsample of countries with more recent country-level media data. We find that the rankings of country-level media competition either do not vary or vary little for the majority of countries in the subsample. Moreover, our inferences hold when we conduct analyses using the more recent data. We conduct additional tests to address the potential endogeneity due to omitted correlated variables and find our main inferences hold. Nevertheless, we acknowledge that we cannot completely rule out the influence of omitted correlated variables on our results and discuss this issue in more detail in “Empirical Results” section. Finally, to shed light on how media competition affects analyst forecast behavior, we provide evidence that analysts in countries with stronger media competition follow more firms on average. This lends support to the view that by encouraging better quality information, media competition reduces analysts’ information acquisition costs and allows them to follow more firms.

Our study adds to the literature investigating cross-country determinants of analyst forecast properties.² While prior studies generally focus on how differences in accounting and disclosure practices across countries (e.g., differences in the enforcement of accounting standards, generally accepted accounting principles, and the importance of nonfinancial disclosures) affect the properties of analyst forecasts, our study demonstrates that the intensity of country-level media competition also affects analyst forecast properties, presumably because media competition enhances the quality of information used by financial analysts.

We also contribute to research on the capital market impacts of the media by examining media features other than media penetration. Prior empirical studies mainly focus on the aggregate level of media penetration within a country, measured by the circulation of all major newspapers. We take a step further by considering variation in market share and ownership structures of the major players in each country. We provide evidence that the intensity of media competition has a positive impact on a country’s information environment and this impact is incremental to that of media penetration. Thus, our study suggests that the media affects the capital markets not only by reaching a broader audience (as captured by media penetration), but also by improving the quality of information because of competition.

Finally, our study sheds light on the relation between media and financial analysts, which are two important information intermediaries in the capital markets. Prior research examines whether financial analysts and the media are complements or supplements, and investigates their relative importance (Bushee et al., 2010; Hillert et al., 2014; Kothari et al., 2009; Li et al., 2011). We argue that the media and analysts act as complements because the media provides a wide range of information that can be useful to analysts. Thus, our study contributes to research on the impact of the media on financial analysts (Bradshaw et al., 2017; Huang & Mamo, 2016) and provides new evidence on the interactions between these important information intermediaries. Overall, our study has important implications for media regulators in understanding the role of media competition in capital market development, and for investors and financial analysts in understanding the country-level factors that influence the quality of the information environment and hence, the efficiency of their judgments and investment decisions.

Background and Empirical Predictions

The Media

The media’s focus on business news has increased substantially over the past few decades; not only are more newspapers dedicated specifically to business topics, but the number of business news articles in each newspaper has also increased dramatically (Chouliaraki & Morsing, 2010; Mazza & Pedersen, 2004). The role of the business press has also changed from simply reproducing statements made by companies and providing information about stock markets to offering additional interpretations and opinions about company transactions, corporate leaders, and operations, as well as creating new information through journalism activities (Bushee et al., 2010).

Although internet-based news media is becoming an important source of information, print newspapers are still a fundamental source of public information in domestic and global settings, and according to the World Association of Newspapers and News Publications (a nonprofit, nongovernmental organization that represents approximately 18,000 publications), the circulation of printed newspapers remained substantial in many countries over our sample period (World Association of Newspapers and News Publications [WAN-IFRA], 2010, 2014). In addition, traditional newspapers still deliver the same type of content to readers, but in digital format, and survey evidence suggests that approximately 46% of internet users worldwide read newspapers but in digital format (WAN-IFRA, 2014).³ Overall, we view both the new internet-based media and the traditional media as important providers of news to the capital markets, and we view our empirical measures, which are based on traditional media, as proxies for competition in the overall media industry.⁴

Media and Financial Analysts

The media provides a wide range of firm-specific, industry-wide, and market-level information that can be used by financial analysts. For example, Bradshaw et al. (2017) find that for a sample of analyst reports about Standard & Poor’s 500 companies in 2012, 30% refer to print newspapers to support their research. They also find that firm-specific news coverage in America’s top 10 print newspapers is positively associated with analysts’ subsequent stock recommendation revisions, and that the informativeness of analysts’ stock recommendations is associated with the “soft” (qualitative) information in news articles. In addition to firm-level news, the media is also a major source of industry-level and macro-economic news, as well as news about country-specific risks and political risks that are important to financial analysts (Agarwal & Hess, 2012; Hope & Kang, 2005; Hugon et al., 2016).⁵

Media Competition and Analyst Forecast Properties

Stronger media competition is traditionally viewed by regulators and researchers as welfare improving (Lacy & Simon, 1993). Analytical research shows that stronger media competition can effectively improve the quality of information (Fan, 2013), as well as discipline information bias arising from government attempts to manipulate information (Besley & Prat, 2006) or from the media’s incentives to confirm to their consumers’ prior beliefs (Gentzkow & Shapiro, 2006). Rennhoff and Wilbur (2012) provide empirical evidence that ownership concentration (a proxy of low media quality) of the local media in the United States is associated with lower news ratings.

However, media competition can also reduce information quality. For example, Cagé (2020) provides both analytical and empirical evidence suggesting that stronger media competition can reduce the quantity and quality of information because of a reduction in the supply of journalists. In addition, using data from emerging countries, Hollifield et al. (2004) find that media firms in highly competitive markets may reduce financial commitment to news production and opt for low-cost, low-quality news products. Similarly, Becker et al. (2009) find that while moderate competition leads to higher quality news products, higher levels of competition can actually reduce news quality. Given these opposing views, it is not clear ex ante how the strength of media competition will affect the properties of analyst forecasts.⁶

Empirical Model and Variables

The empirical model that we use in our main analyses is as follows:

Forecast properties = f (MediaCompetition, Media Penetration, MediaFreedom, control variables) .

(1)

We estimate pooled ordinary least squares regressions for each of the three analyst forecast properties—analyst forecast error (AbsFError), analyst forecast bias (FBias), and analyst forecast dispersion (FDisp). We control for heteroscedasticity and within-firm serial correlation using Rogers standard errors, clustered by firm (Gow et al., 2010; Petersen, 2009),⁷ and we winsorize continuous variables at ±1% to reduce the influence of outliers. We provide detailed descriptions of all variable measurements in Appendix A.

Analyst Forecast Properties

Following Dhaliwal et al. (2012), we measure forecast error (AbsFError) as the absolute value of the difference between actual Earnings Per Share(EPS) and the mean analyst EPS forecast, scaled by stock price at the beginning of the fiscal year.⁸ Forecast bias (FBias) is measured as the actual EPS minus the mean analyst EPS forecast, scaled by stock price at the beginning of the fiscal year. As such, a negative value for FBias indicates that the forecast is optimistic. Finally, forecast dispersion (FDisp) is measured as the standard deviation of the analysts’ most recent EPS forecasts for the firm-year, scaled by stock price at the beginning of the fiscal year.

Media Characteristics

We construct two country-level, time-invariant measures of media competition using data from Djankov et al. (2003); these data were collected for the year 1999.⁹ Our primary measure of media competition is constructed using the Herfindahl index, and is calculated using the market shares of the top five newspapers in the country.¹⁰ Thus, MediaCompetition_HH is computed as $(\sum_{i = 1}^{5} Market {share}_{i}^{2}) \times (- 1)$ so that less negative values represent stronger media competition. Our secondary measure, MediaCompetition_OC, is based on the ownership concentration of the top five newspapers in the country. It is calculated as 5 divided by the number of unique “ultimate owners” of the top five newspapers in the country multiplied by −1, where an “ultimate owner” is defined as the shareholder that, directly or indirectly, controls the greatest number of voting rights (Djankov et al. 2003). This variable ranges from −5 to −1, inclusive, and less negative values represent a lower degree of ownership concentration and hence, stronger media competition. We view MediaCompetition_OC as a secondary measure because it has a limited degree of variation and because its ability to reflect market competition is less direct than that of the market-based Herfindahl measure. We also create a composite media competition variable (MediaCompetition_AVE) by taking the average of the scaled rankings of these two measures.

Empirical evidence suggests the country-level media penetration facilitates the market’s processing of public news (Griffin et al., 2011; Haw et al., 2012). Thus, we control for media penetration (MediaPenetration) in all of our analyses, measured by the average newspaper circulation per 1,000 adults based on data from The World Bank Institute from 1997 through 2004. We also control for media freedom (MediaFreedom) because low freedom of the press can prevent the media from delivering complete and unbiased news (Djankov et al., 2003). We use data from the 2010 World Press Freedom Index issued by Reporters Without Borders.

Control Variables

We follow prior studies in building our empirical models of analyst forecast properties and we include an array of firm-, analyst-, and country-level characteristics as control variables (e.g., Bae et al., 2008; Bilinski et al., 2013). We control for total assets (Size) because larger firms should have better information environments; return on equity (ROE), accounting losses (Loss), earnings volatility (VarEarn), and leverage (Leverage) because firms experiencing poor performance or more volatile performance face more uncertainty, making forecasting more difficult; earnings opacity (Opacity) because analysts face more information asymmetry when earnings are opaque (Byard & Shaw, 2003); the use of a Big 4 auditor (Big4) because committing to high-quality auditing services should lead to higher financial reporting quality; institutional ownership (InstitutionalOwnership) because higher institutional ownership is associated with greater forecast accuracy and less forecast optimism (Ljungqvist et al., 2007); Chief Executive Officer (CEO) ownership (CeoOwnership) because of the potential impact of equity-incentives on firms’ financial reporting activities and hence, on the information environment; the number of analysts following the firm (Nanalyst) because analyst following may affect management’s disclosure behavior and because greater analyst following could improve information availability by enhancing information discovery and analyses; and the number of stock exchanges that the firm is listed on (StkExch) because firms listed on multiple exchanges should face greater pressure to improve corporate transparency and provide higher quality information.

Our models include the number of years of forecasting experience for analysts following the firm (Gexp) because more experienced analysts should provide more accurate forecasts;¹¹ the median number of firms (Nfirm) and countries (Ncountry) followed by analysts following the firm, because analysts who follow more firms should produce less accurate forecasts because of time constraints (Clement, 1999) and because country-specialist analysts produce more accurate forecasts (Sonney, 2009); the median of the number of forecasts made in the year by analysts following the firm (Nforecast) because this proxies for analysts’ diligence and/or the amount of information generated by analysts; the median of brokerage house size (Bsize) for analysts following the firm because larger brokerage houses should provide analysts with access to more resources (Clement, 1999; Jacob et al., 1999); and the median forecast horizon (Horizon) for all forecasts made by analysts following the firm, because forecasts made closer to earnings announcements are more accurate (Clement, 1999).

We control for a country’s financial disclosure requirements (Disclosure) and the effectiveness of the country’s judiciary system (RuleLaw) because more stringent disclosure requirements and more effective judiciary systems should improve the information environment (Hail & Leuz, 2006; Lau et al., 2012). We control for whether the firm uses international financial reporting standards (IFRS) because IFRS adoption improves analyst forecast accuracy. We also control for earnings surprise management in the country (Ratio) because Hope (2003) finds that analyst forecasts are less dispersed and more accurate in countries with greater management of earnings surprises. Finally, we control for a country’s annual gross domestic product (GDP) because a country’s economic development can influence its capital market development and business practices, which in turn affect the information environment and hence analyst forecasts.

Sample Selection and Descriptive Statistics

We restrict the sample to observations with earnings forecasts and actual earnings available from I/B/E/S, stock price data available from the Center for Research in Security Prices (for U.S. firms) or from COMPUSTAT Global (for non-U.S. firms), and other required financial data available from COMPUSTAT Global. We require that the three country-level media characteristics variables (i.e., media competition, media penetration, and media freedom) be available for the country in which the firm is domiciled. We also require that country-level control variables Disclosure and RuleLaw be available from La Porta et al. (2006). Our final sample consists of 113,436 firm-year observations from 19,490 unique firms in 32 countries from 2000 through 2012. For forecast dispersion tests, we further require that the firm is followed by at least three analysts, which reduces the sample to 79,678 firm-year observations.

Table 1 presents summary statistics for the average analyst forecasts properties and for the country-level media variables, by country. In our full sample, 35,500 observations (31%) are from the United States. Samples from Japan, the United Kingdom, and Canada are also relatively large sample but none dominates the full sample. The country-level average number of analysts following is 8.06, with Spain, South Korea, and the Netherlands having the highest average number of analysts following (at approximately 13).

Table 1.

Summary Statistics by Country.

	Country	N	AbsFError	FBias	FDisp	NAnalyst	MediaCompetition_HH	MediaCompetition_OC	MediaFreedom	MediaPenetration
1	Argentina	328	0.158	−0.076	0.071	2.86	−0.11	−1.00	−55	52.60
2	Australia	4,658	0.054	−0.025	0.021	6.19	−0.07	−1.67	−18	197.50
3	Austria	466	0.047	−0.021	0.022	7.21	−0.18	−1.25	−7	354.30
4	Belgium	823	0.043	−0.013	0.020	7.75	−0.10	−1.25	−14	199.25
5	Brazil	1,597	0.060	−0.016	0.036	7.91	−0.01	−1.25	−58	54.00
6	Canada	7,385	0.051	−0.025	0.023	6.75	−0.02	−1.25	−21	293.90
7	Denmark	823	0.049	−0.015	0.029	7.54	−0.13	−1.67	−11	388.50
8	Finland	1,208	0.041	−0.013	0.023	10.91	−0.06	−1.67	−1	600.80
9	France	4,399	0.042	−0.015	0.022	10.98	−0.02	−1.00	−44	176.80
10	Germany	4,175	0.067	−0.035	0.029	10.30	−0.23	−1.25	−17	319.00
11	Greece	731	0.088	−0.055	0.043	7.07	−0.06	−1.00	−70	82.20
12	India	2,932	0.032	−0.009	0.015	8.70	−0.01	−1.00	−122	78.15
13	Indonesia	718	0.089	−0.045	0.036	8.10	−0.04	−1.00	−117	31.35
14	Ireland	700	0.062	−0.018	0.024	5.89	−0.07	−1.25	−9	233.75
15	Italy	1,713	0.048	−0.017	0.023	8.60	−0.03	−1.00	−49	146.20
16	Japan	15,017	0.028	−0.014	0.014	5.99	−0.04	−1.00	−11	644.00
17	Malaysia	2,328	0.052	−0.016	0.024	6.61	−0.04	−1.00	−141	201.75
18	Mexico	1,026	0.066	−0.030	0.033	6.68	−0.04	−1.25	−136	1.20
19	Netherlands	1,703	0.047	−0.018	0.028	12.91	−0.05	−2.50	−1	353.25
20	New Zealand	729	0.036	−0.018	0.016	4.90	−0.11	−1.67	−8	251.60
21	Norway	1,396	0.086	−0.038	0.045	7.53	−0.05	−1.67	−1	665.45
22	Portugal	335	0.046	−0.015	0.025	9.09	−0.05	−1.25	−40	88.95
23	Singapore	1,757	0.066	−0.023	0.026	5.80	−0.22	−5.00	−136	362.60
24	South Africa	1,398	0.039	−0.015	0.016	5.38	−0.08	−1.25	−38	61.55
25	South Korea	1,638	0.064	−0.030	0.029	13.30	−0.05	−1.00	−42	47.50
26	Spain	1,102	0.037	−0.015	0.020	13.45	−0.02	−1.00	−39	166.45
27	Sweden	2,161	0.050	−0.023	0.024	9.21	−0.04	−1.67	−1	546.05
28	Switzerland	1,809	0.036	−0.009	0.020	9.41	−0.03	−1.00	−1	473.55
29	Thailand	1,320	0.073	−0.032	0.037	7.52	−0.01	−1.25	−153	249.80
30	Turkey	682	0.070	−0.012	0.044	7.32	−0.06	−1.67	−138	147.35
31	UK	10,879	0.052	−0.020	0.019	7.09	−0.07	−1.00	−19	387.95
32	USA	35,500	0.036	−0.014	0.014	9.04	−0.00	−1.00	−20	267.30
	Total	113,436
	Average		0.057	−0.023	0.027	8.06	−0.07	−1.47	−48	253.89

Note.Table 1 presents summary statistics for the analyst forecast variables and the media variables using 113,436 firm-year observations (79,678 observations for FDisp) representing 19,490 unique firms from 32 countries from 2000 through 2012. All variables are defined in Appendix A. The average values in the last row are calculated using country-level data.

Newspapers in the United States face the most competition (MediaCompetition_HH = −0.003), followed by newspapers in India, Brazil, Thailand, Spain, France, and Canada. The newspaper industry is least competitive in Germany, Singapore, and Austria. Thirteen countries, including the United States, have an ownership concentration (MediaCompetition_OC) score of −1, indicating that the top five newspapers in these countries are owned by five different entities or individuals (so competition is high). The top five newspapers in most other countries have three or four ultimate owners, with Singapore having the highest ownership concentration (at −5) and least competition. Overall, countries with high media penetration also have relatively high media freedom, but these countries do not necessarily have high media competition. In fact, some countries with media penetration above the median (e.g., Denmark, Singapore, Austria, and Germany) have relatively low MediaCompetition_HH, and some countries with low media penetration (e.g., Brazil and India) have relatively high MediaCompetition_HH.

Table 2 presents descriptive statistics for our main variables. The mean (median) absolute forecast error is 0.045 (0.014), and the mean (median) forecast dispersion is 0.020 (0.007). Analyst forecasts are optimistic on average (with a mean FBias of −0.018) but at the median, forecast bias is close to zero (−0.002). Approximately 79% of sample observations report positive earnings but their profitability is moderate, with a mean (median) ROE of 0.01 (0.09). Institutions and CEOs own 22.0% and 1.3% of shares, respectively, for the mean observation, but own much less for the median observation (at 7.20% and 0%, respectively). The average leverage ratio is approximately 50%. Eighty-four percent of sample observations are audited by Big 4 auditors, and the number of analysts following ranges from 1 to 38 inclusive, with a mean (median) of 8.14 (5). Firms cross-list their stocks on 0.15 foreign stock exchanges (StkExch), on average. The average analyst has worked in the industry for between 6 and 7 years (Gexp), and issues approximately four forecasts on average for each firm followed during the year (Nforecast), with an average forecast horizon of approximately 200 days (Horizon). The average brokerage house employs between 36 and 50 analysts. Finally, approximately 22% of sample observations report under IFRS.

Table 2.

Descriptive Statistics.

	N	M	1%	25%	Median	75%	99%
Dependent variables
AbsFError	113,436	0.045	0.001	0.006	0.014	0.035	0.746
FBias	113,436	−0.018	−0.568	−0.017	−0.002	0.006	0.185
FDisp	79,678	0.020	0.000	0.003	0.007	0.018	0.284
Media variables
MediaCompetition_HH	113,436	−0.04	−0.23	−0.06	−0.04	−0.00	−0.00
MediaCompetition_OC	113,436	−1.20	−5.00	−1.25	−1.00	−1.00	−1.00
MediaPenetration	113,436	321.33	31.35	249.80	267.30	387.95	665.45
MediaFreedom	113,436	−30.84	−153.00	−21.00	−20.00	−17.00	−1.00
ForeignOwner	113,436	0.07	0.00	0.00	0.00	0.00	1.00
All other variables
Size	113,436	1.90	0.63	1.63	1.92	2.21	2.70
ROE	113,436	0.01	−2.40	0.00	0.09	0.17	1.34
Loss	113,436	0.21	0.00	0.00	0.00	0.00	1.00
VarEarn	113,436	−0.07	−2.56	−0.91	−0.25	0.62	3.89
Leverage	113,436	0.50	0.05	0.33	0.51	0.65	1.14
Opacity	113,436	0.42	0.00	0.00	0.00	1.00	1.00
BIG4	113,436	0.84	0.00	1.00	1.00	1.00	1.00
InstitutionalOwnership	113,436	0.22	0.00	0.00	0.07	0.36	1.00
CeoOwnership	113,436	0.01	0.00	0.00	0.00	0.00	0.37
NAnalyst	113,436	8.14	1.00	2.00	5.00	11.00	38.00
StkExch	113,436	0.15	0.00	0.00	0.00	0.00	3.00
GExp	113,436	6.65	1.00	4.00	6.00	8.00	26.00
NFirm	113,436	14.42	2.00	10.00	14.00	17.00	44.00
NCountry	113,436	1.66	1.00	1.00	1.00	2.00	7.00
NForecast	113,436	3.84	1.00	2.00	4.00	5.00	11.00
BSize	113,436	50.08	1.00	19.00	36.00	65.00	237.00
Horizon	113,436	200.41	29.50	173.00	197.50	231.00	372.00
Disclosure	113,436	0.81	0.25	0.750	0.83	1.00	1.00
RuleLaw	113,436	1.67	−0.37	1.82	1.92	1.93	2.22
IFRS	113,436	0.22	0.00	0.00	0.00	0.00	1.00
Ratio	113,436	1.06	0.95	1.01	1.05	1.11	1.16
GDP	113,436	0.036	0.001	0.031	0.039	0.046	0.073

Note.Table 2 reports the distributions of variables used in our analyses for all firm-year observations in our sample. All variables are defined in Appendix A. InstitutionalOwnership and CeoOwnership are winsorized at 100% and divided by 100 for ease of exposition.

Table 3 reports correlations for the variables of interest and main firm-level control variables. Absolute forecast error and forecast bias are negatively correlated, suggesting that absolute forecast errors tend to be driven by optimistic forecasts. In addition, forecast dispersion is positively correlated with absolute forecast error, but negatively correlated with forecast bias. The two measures of media competition are positively correlated but they are generally not significantly correlated with either MediaPenetration or MediaFreedom.¹² Finally, MediaCompetition, MediaPenetration, and MediaFreedom are all negatively correlated with absolute forecast error and with forecast dispersion but are positively correlated with forecast bias. These results demonstrate the importance of controlling for media penetration and media freedom when examining the effect of media competition on analyst forecast properties.

Table 3.

Correlation Matrix (Pearson\Spearman).

	AbsFError	FBias	FDisp	MediaCompetition_HH	MediaCompetition_OC	MediaPenetration	MediaFreedom	Size	ROE	Loss	VarEarn	Leverage	Opacity	BIG4	InstituteOwn.	CeoOwn.	NAnalyst	StkExch
AbsFError		−0.68	0.66	−0.07	−0.04	−0.04	−0.05	−0.10	−0.27	0.38	0.13	0.15	0.08	–0.04	–0.10	–0.01	–0.15	–0.04
FBias	–0.29		–0.45	0.04	0.02	0.01	0.01	0.06	0.26	–0.39	–0.08	–0.12	–0.06	0.04	0.09	0.01	0.10	0.03
FDisp	0.74	–0.23		–0.07	–0.06	−0.04	–0.07	–0.04	–0.25	0.36	0.14	0.13	0.08	−0.00	–0.09	–0.01	–0.09	–0.01
MediaCompetition_HH	–0.16	0.04	–0.23		0.51	−0.14	−0.10	0.07	–0.03	0.03	0.01	–0.02	0.01	0.03	0.18	–0.04	0.03	–0.08
MediaCompetition_OC	–0.15	0.05	–0.22	0.41		−0.26	0.15	0.07	–0.03	0.02	0.05	0.03	0.01	–0.06	0.08	–0.06	0.02	–0.03
MediaPenetration	–0.03	–0.00	–0.05	−0.14	–0.37		0.52	0.21	–0.02	–0.02	0.06	–0.00	0.03	0.00	–0.08	–0.02	–0.08	−0.06
MediaFreedom	–0.05	0.01	–0.06	−0.23	–0.33	0.69		–0.12	–0.08	0.09	0.09	0.02	0.01	0.09	0.07	–0.08	0.02	0.00
Size	–0.14	0.05	–0.02	–0.05	0.12	0.11	0.09		0.20	–0.31	–0.05	0.26	–0.07	0.15	0.06	–0.09	0.37	0.18
ROE	–0.36	0.34	–0.32	–0.05	–0.05	–0.11	–0.08	0.16		–0.47	–0.09	0.03	–0.06	0.04	0.07	0.02	0.11	0.06
Loss	0.46	–0.36	0.39	0.07	–0.03	0.01	0.01	–0.29	–0.58		0.15	–0.02	0.09	–0.06	–0.11	–0.03	–0.17	–0.06
VarEarn	0.26	–0.06	0.24	0.03	0.01	0.07	0.06	–0.09	–0.22	0.18		0.09	0.12	–0.04	0.00	–0.02	–0.09	–0.02
Leverage	0.11	–0.05	0.13	–0.04	–0.02	–0.01	0.02	0.28	0.13	–0.04	0.08		0.07	0.07	−0.00	–0.04	0.11	0.05
Opacity	0.12	–0.04	0.11	0.01	0.02	0.02	0.01	–0.07	–0.04	0.09	0.12	0.06		–0.05	–0.03	0.02	–0.05	–0.04
BIG4	–0.07	0.04	–0.02	0.02	–0.09	0.01	0.03	0.13	0.04	–0.06	–0.04	0.07	–0.05		0.05	–0.12	0.18	0.07
InstituteOwn.	–0.17	0.12	–0.16	0.14	0.05	–0.01	–0.02	0.08	0.13	–0.13	–0.01	–0.01	–0.03	–0.03		0.03	0.18	0.12
CeoOwn.	–0.07	0.06	–0.10	0.12	0.06	0.01	–0.04	–0.06	0.06	–0.05	–0.01	–0.05	0.01	–0.10	0.56		–0.07	–0.03
NAnalyst	–0.27	0.10	–0.11	0.14	0.03	–0.09	–0.05	0.37	0.18	–0.19	–0.12	0.11	–0.06	0.22	0.21	0.03		0.23
StkExch	–0.03	0.03	0.05	–0.10	–0.16	–0.04	–0.01	0.18	0.08	–0.05	–0.02	0.05	–0.04	0.07	0.25	0.08	0.17

Note.Table 3 reports Pearson (upper) and Spearman (lower) correlations. All variables are defined in Appendix A. The correlations in bold are statistically significant at p value .10 or lower. All correlations are based on firm-year observations except for the correlations among the country-level media variables (MediaCompetition_HH, MediaCompetition_OC, MediaPenetration, and MediaFreedom) which are based on country-level data.

Empirical Results

Main Results

We begin our empirical tests by comparing analyst forecast properties between subsamples partitioned on media competition. Specifically, we partition the full sample into “Low” and “High” subsamples based on whether a firm-year observation is from a country with media competition that is below or above the median value for all countries. Results in Table 4 Panels A (Panel B) suggest that analyst forecasts are more accurate, less optimistically biased, and less dispersed for firms in countries characterized by stronger media competition when competition is measured by market share (ownership concentration). Specifically, the forecast error is 0.056 (0.04) for the observations in the above-median (below-median) subsample, representing a decrease of 28.6% (0.016/0.056). The magnitude of the change is similar for the other two forecast properties (29.1% for forecast bias and 25% for forecast dispersion).

Table 4.

Comparative Statistics.

Panel A
MediaCompetition_HH	Low		High
MediaCompetition_HH	N	M	N	M	Diff (High − Low)	t-statistic
AbsFError	31,330	0.056	82,106	0.040	−0.016***	−20.41
FBias	31,330	−0.024	82,106	−0.016	0.007***	11.15
FDisp	19,832	0.024	59,846	0.018	−0.006***	−13.95
Panel B
MediaCompetition_OC	Low		High
	N	M	N	M	Diff (High − Low)	t-statistic
AbsFError	34,342	0.056	79,094	0.040	−0.016***	−22.26
FBias	34,342	−0.024	79,094	−0.016	0.008***	13.36
FDisp	23,933	0.027	55,745	0.017	−0.010***	−28.03

Note.Table 4 reports values of the analyst forecast properties for subsamples based on the median value of media competition. The “Low” (“High”) subsample represents the observations where the country-level value of media competition is below (above) the sample median. All variables are defined in Appendix A. We compare the mean values of each analyst forecast property for the low versus high subsamples and report the difference in the last two columns.

, **, and *** represent significance at 10%, 5%, and 1%, respectively.

Results in Table 5, where the dependent variable is absolute forecast error, reveal a consistently negative association between absolute forecast error (AbsFError) and each of our media competition measures (p values < .01), revealing that analyst forecasts are more accurate in countries with greater media competition. Thus, stronger media competition improves, rather than harms, the information environment. MediaPenetration and MediaFreedom are also negatively associated with analyst forecast error. The signs on the control variable coefficient estimates are generally consistent with prior studies. For example, we find that firms with low accounting earnings (ROE), loss firms (Loss), firms with more volatile accounting earnings (VarEarn), or more opaque earnings (Opacity), and firms with higher leverage (Leverage) experience greater absolute forecast errors.

Table 5.

Media and Analyst Forecast Error (AbsFError).

	Coeff.	t-stat.	Coeff.	t-stat.	Coeff.	t-stat.
Intercept	0.063***	5.92	0.069***	6.50	0.058***	5.46
MediaCompetition_HH	−0.105***	−8.00
MediaCompetition_OC			−0.006***	−5.39
MediaCompetition_AVE					−0.034***	−12.48
MediaPenetration	−0.020***	−5.04	−0.020***	−5.19	−0.017***	−4.35
MediaFreedom	−0.198***	−8.10	−0.186***	−6.90	−0.201***	−8.66
Size	−0.001	−0.42	−0.001	−0.85	0.000	0.24
ROE	−0.029***	−14.00	−0.029***	−14.05	−0.029***	−14.12
Loss	0.079***	48.42	0.079***	48.26	0.079***	48.28
VarEarn	0.005***	18.04	0.005***	18.50	0.005***	18.34
Leverage	0.070***	28.10	0.070***	28.14	0.071***	28.47
Opacity	0.004***	6.30	0.004***	6.21	0.005***	6.62
BIG4	−0.003***	−3.03	−0.003***	−3.84	−0.004***	−4.05
InstituteOwn.	−0.004***	−2.91	−0.003***	−2.68	−0.004***	−2.79
CeoOwn.	0.000	0.02	0.001	0.19	0.003	0.50
NAnalyst	−0.001***	−22.36	−0.001***	−22.02	−0.001***	−21.56
StkExch	−0.002***	−2.78	−0.002***	−2.90	−0.002***	−2.74
GExp	−0.000**	−2.04	−0.000*	−1.76	−0.000	−1.30
NFirm	0.000	1.37	0.000	1.24	0.000*	1.72
NCountry	0.003***	6.19	0.003***	6.47	0.003***	6.46
NForecast	−0.000	−1.61	−0.000**	−2.36	−0.001***	−3.19
BSize	−0.037***	−3.80	−0.031***	−3.20	−0.037***	−3.79
Horizon	0.140***	20.88	0.142***	21.08	0.137***	20.47
Disclosure	−0.021***	−5.35	−0.034***	−9.62	−0.016***	−4.04
RuleLaw	−0.007***	−2.75	−0.003	−1.27	0.009***	−3.89
IFRS	−0.005***	−3.56	−0.003**	−2.17	−0.008***	−6.07
Ratio	−0.048***	−5.28	−0.048***	−5.19	−0.021**	−2.11
GDP	0.343***	4.94	0.166**	2.57	0.411***	5.90
Year indicators	yes		yes		yes
Industry indicators	yes		yes		yes
N	113,436		113,436		113,436
Adj. R²	0.234		0.233		0.235

Note.Table 5 reports the results from regressing absolute forecast errors (AbsFError) on the media characteristics and control variables. We multiply the coefficients on MediaPenetration, MediaFreedom, BSize, and Horizon by 1,000 for ease of exposition. In all tests, the coefficient estimates are from pooled, cross-sectional regressions with year and industry indicators. Following Hope (2003), the industry indicators are based on one-digit SIC codes. The t-statistics are based on Rogers standard errors, which are clustered at the firm level and control for serial correlation and heteroscedasticity (Petersen, 2009). All variables are defined in the Appendix A.

, **, and *** represent significance at 10%, 5%, and 1%, respectively.

Table 6 presents the results where the dependent variable is forecast bias (FBias), defined as actual EPS minus forecasted EPS so that a negative value indicates analyst forecast optimism. Consistent with expectations, for all three measures of media competition, forecast optimism is lower when MediaCompetition is higher (two with p values < .01 and one with p value < .05). Table 7 reports the results where forecast dispersion (FDisp) is the dependent variable. Here we find that all media competition variables are negatively associated with FDisp (two with p values < .01 and one with p value < .05). Results for most control variables in Tables 6 and 7 are similar to those in Table 5.

Table 6.

Media and Analyst Forecast Bias (FBias).

	Coeff.	t-stat.	Coeff.	t-stat.	Coeff.	t-stat.
Intercept	−0.019**	−2.34	−0.025***	−2.98	−0.019**	−2.27
MediaCompetition_HH	0.067***	6.43
MediaCompetition_OC			0.002**	2.01
MediaCompetition_AVE					0.016***	7.44
MediaPenetration	0.017***	5.69	0.019***	6.08	0.017***	5.43
MediaFreedom	0.060***	3.22	0.085***	4.10	0.075***	4.27
Size	−0.010***	−7.83	−0.010***	−7.70	−0.011***	−8.27
ROE	0.022***	13.60	0.023***	13.62	0.023***	13.67
Loss	−0.075***	−55.59	−0.075***	−55.52	−0.075***	−55.54
VarEarn	−0.001***	−5.25	−0.001***	−5.55	−0.001***	−5.48
Leverage	−0.044***	−22.51	−0.044***	−22.39	−0.045***	−22.66
Opacity	−0.001***	−2.64	−0.001***	−2.63	−0.002***	−2.83
BIG4	0.005***	5.29	0.005***	5.77	0.005***	5.96
InstitutionOwn.	−0.001	−0.73	−0.001	−0.96	−0.001	−0.87
CeoOwn.	−0.013***	−2.61	−0.013***	−2.79	−0.014***	−2.93
NAnalyst	0.000***	11.93	0.000***	11.93	0.000***	11.42
StkExch	0.001	1.28	0.001	1.32	0.001	1.24
GExp	0.000	0.45	0.000	0.31	0.000	0.01
NFirm	−0.000***	−3.68	−0.000***	−3.46	−0.000***	−3.80
NCountry	0.001***	3.37	0.001***	3.13	0.001***	3.15
NForecast	0.000	0.83	0.000	1.27	0.000*	1.85
BSize	0.030***	4.20	0.027***	3.76	0.029***	4.09
Horizon	−0.095***	−17.33	−0.096***	−17.53	−0.094***	−17.12
Disclosure	0.009***	3.08	0.020***	6.93	0.010***	3.27
RuleLaw	0.000	−0.15	−0.004**	−2.30	−0.000	−0.12
IFRS	−0.002*	−1.64	−0.003***	−3.32	−0.001	−0.54
Ratio	0.046***	6.41	0.048***	6.57	0.034***	4.59
GDP	−0.033	−0.59	0.078	1.53	−0.035	−0.63
Year indicators	yes		yes		yes
Industry indicators	yes		yes		yes
N	113,436		113,436		113,436
Adj. R²	0.204		0.204		0.204

Note.Table 6 reports the results from regressing forecast bias (FBias) on the media characteristics and control variables. We multiply the coefficients on MediaFreedom, MediaPenetration, BSize, and Horizon by 1,000 for ease of exposition. All variables are defined in the Appendix A. In all tests, the coefficient estimates are from pooled, cross-sectional regressions with year and industry indicators. Following Hope (2003), the industry indicators are based on one-digit SIC codes. The t-statistics are based on Rogers standard errors, which are clustered at the firm level and control for serial correlation and heteroscedasticity (Petersen, 2009).

, **, and *** represent significance at 10%, 5%, and 1%, respectively.

Table 7.

Media and Analyst Forecast Dispersion (FDisp).

	Coeff.	t-stat.	Coeff.	t-stat.	Coeff.	t-stat
Intercept	0.038***	7.15	0.039***	7.21	0.036***	6.73
MediaCompetition_HH	−0.016**	−2.44
MediaCompetition_OC			−0.002***	−2.97
MediaCompetition_AVE					−0.010***	−7.13
MediaPenetration	−0.005**	−2.24	−0.005**	−2.17	−0.004*	−1.78
MediaFreedom	−0.138***	−11.61	−0.125***	−9.07	−0.127***	−11.29
Size	0.001	0.95	0.001	0.76	0.001	1.23
ROE	−0.011***	−11.12	−0.011***	−11.15	−0.011***	−11.19
Loss	0.033***	36.54	0.033***	36.50	0.033***	36.54
VarEarn	0.003***	18.35	0.003***	18.55	0.003***	18.46
Leverage	0.023***	19.16	0.023***	19.16	0.024***	19.36
Opacity	0.002***	5.26	0.002***	5.23	0.002***	5.50
BIG4	0.001**	2.08	0.001*	1.67	0.001	1.47
InstitutionOwn.	−0.003***	−4.65	−0.003***	−4.62	−0.003***	−4.69
CeoOwn.	0.002	0.58	0.002	0.61	0.002	0.81
NAnalyst	−0.000***	−13.11	−0.000***	−12.72	−0.000***	−12.36
StkExch	−0.001**	−2.37	−0.001**	−2.43	−0.001**	−2.33
GExp	−0.000	−0.33	−0.000	−0.25	−0.000	−0.03
NFirm	−0.000	−1.36	−0.000	−1.33	−0.000	−0.86
NCountry	0.001***	5.39	0.001***	5.44	0.001***	5.47
NForecast	0.001***	5.30	0.000***	4.83	0.000***	4.10
BSize	−0.019***	−3.85	−0.017***	−3.46	−0.019***	−3.72
Horizon	0.041***	10.75	0.041***	10.81	0.040***	10.36
Disclosure	−0.022***	−11.37	−0.023***	−13.40	−0.018***	−9.44
RuleLaw	−0.002*	−1.69	−0.002*	−1.86	−0.004***	−3.60
IFRS	−0.002***	−2.71	−0.002**	−2.51	−0.003***	−4.74
Ratio	−0.023***	−5.14	−0.023***	−5.06	−0.015***	−3.33
GDP	0.208***	5.66	0.183***	5.29	0.252***	6.72
Year indicators	yes		Yes		yes
Industry indicators	yes		Yes		yes
N	79,678		79,678		79,678
Adj. R²	0.227		0.228		0.227

Note. Table 7 reports the results from regressing forecast dispersion (FDisp) on the media characteristics and control variables. We multiply the coefficients on MediaPenetration, MediaFreedom, BSize, and Horizon by 1,000 for ease of exposition. All variables are defined in the Appendix A. In all tests, the coefficient estimates are from pooled, cross-sectional regressions with year and industry indicators. Following Hope (2003), the industry indicators are based on one-digit SIC codes. The t-statistics are based on Rogers standard errors, which are clustered at the firm level and control for serial correlation and heteroscedasticity (Petersen, 2009).

, **, and *** represent significance at 10%, 5%, and 1%, respectively.

We also examine the economic impact of media competition on analyst forecast properties. For a one-standard deviation change in MediaCompetition_AVE (0.0522, untabulated), the forecast error, forecast bias, and forecast dispersion change by 12.3%, −19.1%, and 4.2%, respectively; these amounts appear to be economically significant. Overall, we find that media competition improves the quality of analyst forecasts; when competition is stronger, analyst forecasts are more accurate, less optimistically biased, and generally less dispersed.

Endogeneity and Correlated Omitted Variables

Like many studies that use international data, our study faces endogeneity concerns because it is difficult to control for all of the country-level factors that potentially influence analyst forecast properties. We adopt several approaches to address this issue. First, we follow Isidro et al. (2016) to enhance our controls for unobserved country-level factors. Isidro et al. (2016) identify four latent factors that explain a substantial amount of the variation in 72 country-level variables from prior studies. These latent factors reflect the country’s economic structure and development, legal origin and legal systems, regulatory systems, and sociological characteristics. In addition, Isidro et al. (2016) show that individual country-level variables have little incremental explanatory power once the four latent factors are included. Because financial reporting quality largely reflects firm-level transparency and the quality of the information environment (which are vital to analyst forecasts), we use these latent variables to control for correlated factors that influence both analyst forecast properties and media competition, and report the results in Table 8. Consistent with Isidro et al. (2016), the four country-level factors have significant explanatory power for analyst forecast properties in most cases. More importantly, our media competition variables remain statistically significant in all but one case (when we use MediaCompetition_CC in Panel B).

Table 8.

Media and Analyst Forecast Properties—Controlling for Country-Level Latent Factors Following Isidro et al. (2016).

Panel A Dependent variable: Analyst Forecast Error (AbsFError)
	Coeff.	t-stat.	Coeff.	t-stat.	Coeff.	t-stat.
MediaCompetition_HH	−0.099***	−8.94
MediaCompetition_OC			−0.003***	−2.67
MediaCompetition_AVE					−0.028***	−10.57
MediaPenetration	−0.010*	−1.86	−0.019***	−3.59	−0.018***	−3.47
MediaFreedom	−0.098***	−2.78	−0.118***	−3.27	−0.150***	−4.32
Factor1	−0.003	−1.53	0.001	0.41	−0.002	−0.94
Factor2	−0.002***	−2.89	−0.002***	−3.03	−0.003***	−3.70
Factor3	−0.005***	−5.44	−0.006***	−7.09	−0.002**	−2.47
Factor4	0.006***	6.72	0.006***	5.35	0.002**	2.16
Firm-level control variables	yes		yes		yes
Year and industry dummies	yes		yes		yes
Adj. R ²	0.233		0.232		0.233
Panel B Dependent variable: Analyst Forecast Bias (FBias)
	Coeff.	t-stat.	Coeff.	t-stat.	Coeff.	t-stat.
MediaCompetition_HH	0.078***	8.77
MediaCompetition_OC			0.001	1.47
MediaCompetition_AVE					0.020***	9.82
MediaPenetration	0.014***	3.40	0.020***	5.03	0.020***	5.03
MediaFreedom	0.019	0.75	0.051*	1.88	0.062**	2.39
Factor1	0.000	0.21	−0.004**	−2.32	−0.001	−0.73
Factor2	0.000	0.91	0.000	0.67	0.001	1.61
Factor3	0.003***	4.52	0.004***	6.56	0.001**	1.97
Factor4	−0.003***	−4.14	−0.003***	−3.48	−0.000	−0.10
Firm-level control variables	yes		yes		yes
Year and industry dummies	yes		yes		yes
Adj. R ²	0.204		0.203		0.204
Panel C Dependent variable: Analyst Forecast Dispersion (FDisp)
	Coeff.	t-stat.	Coeff.	t-stat.	Coeff.	t-stat.
MediaCompetition_HH	−0.029***	−5.32
MediaCompetition_OC			−0.001 **	−2.12
MediaCompetition_AVE					−0.011***	−8.67
MediaPenetration	0.003	1.08	0.000	0.11	0.000	0.12
MediaFreedom	−0.100***	−6.05	−0.101***	−5.70	−0.113***	−6.91
Factor1	0.000	0.12	0.001	1.03	−0.000	−0.06
Factor2	−0.003***	−7.90	−0.003***	−7.79	−0.003***	−8.70
Factor3	−0.002***	−3.99	−0.002***	−4.97	−0.000	−1.15
Factor4	0.003***	7.71	0.003***	6.33	0.002***	3.71
Firm-level control variables	yes		yes		yes
Year and industry dummies	yes		yes		yes
Adj. R ²	0.224		0.224		0.225

Note.Table 8 reports the results from regressing forecast properties on the media characteristics, all firm-level control variables, industry and year indicators, and four country-level latent factors constructed by Isidro et al. (2016). Specifically, the country-level latent factors are based on a factor analysis of 72 individual country-level factors used in prior research to measure differences across countries, and these latent variables reflect a country’s economic structure and development, legal origin and legal systems, regulatory systems, and sociological characteristics. All other variables are as defined in the Appendix A. The coefficient estimates are from pooled, cross-sectional regressions. The t-statistics are based on Rogers standard errors clustered at the firm level and control for serial correlation and heteroscedasticity (Petersen, 2009).

, **, and *** represent significance at 10%, 5%, and 1%, respectively.

We also conduct analyses on changes in media competition using available data. Specifically, a database developed by World Press Trends provides cross-country information about the circulation of top newspapers from 2010 through 2012. This allows us to compute media competition based on the Herfindahl index for 23 of the 32 countries in our sample. As shown in Columns I and II of Appendix B, the raw values of MediaCompetition_HH based on Djankov et al. (2003) and on the World Press Trends are highly comparable for the majority of the countries. Only Germany, Denmark, and Ireland experience large changes in the competition measure. The Pearson correlation coefficient is 0.69 (p value < .01), and is 0.88 when excluding Germany.¹³ Thus, our results based on Djankov et al. (2003) should reflect current media competition and its impact on financial analysts. In Table 9, Panel A the coefficient on media competition variable remains statistically significant in all regressions, supporting inferences from our main analyses.¹⁴

Table 9.

Changes in Media Competition and Forecast Properties.

Panel A Media competition measured as a rank variable
	AbsFError		FBias		FDisp
	Coeff.	t-stat.	Coeff.	Coeff.	t-stat.	Coeff.
Intercept	0.073***	(6.87)	−0.024***	(–2.67)	0.044***	(7.81)
Media Competition_RANK	−0.026***	(–12.48)	0.014***	(7.79)	−0.007***	(–6.31)
Media Penetration	−0.029***	(–7.21)	0.007	(1.35)	0.007*	(1.89)
Media Freedom	−0.250***	(–10.87)	0.060***	(2.90)	−0.122***	(–9.01)
All control variables	Yes		Yes		Yes
N	113,436		113,436		79,678
Adj. R ²	0.235		0.186		0.218
Panel B Using the sample from 2010 through 2012
	AbsFError		FBias		FDisp
	Coeff.	t-stat.	Coeff.	t-stat.	Coeff.	t-stat.
MediaCompetition_HH	−0.019***	(–4.90)	0.015***	(4.32)	−0.005**	(–2.20)
MediaPenetration	0.005	(0.61)	0.015**	(2.00)	−0.000	(–0.04)
MediaFreedom	0.023	(0.62)	−0.024	(–0.84)	−0.023	(–1.38)
All control variables	Yes		Yes		Yes
Adj. R2	0.189		0.165		0.192
N	19,899		19,899		14,468
Panel C Improvement (deterioration) in media competition and changes in forecast properties
	AbsFError		FBias		FDisp
	Coeff.	t-stat.	Coeff.	t-stat.	Coeff.	t-stat.
Intercept	0.147***	(11.90)	−0.064***	(–6.46)	0.069***	(11.45)
Post	−0.012***	(–7.09)	0.015***	(10.94)	−0.001	(–1.53)
Improve	0.009*	(1.90)	−0.011***	(–2.74)	0.001	(0.54)
PostImprove*	−0.015***	(–2.69)	0.015***	(3.40)	−0.001	(–0.50)
Worsen	−0.033***	−12.52)	0.023***	(12.35)	−0.011***	(–9.25)
PostWorsen*	0.023***	(9.63)	−0.021***	−10.68)	0.006***	(4.73)
All control variables	Yes		Yes		Yes
Adj. R2	0.233		0.199		0.218
N	63,530		63,530		45,608

Note.Table 9 Panel A reports results, using the full sample, from regressing forecast properties on the rank of MediaCompetition_HH. In Panel B, the results are based on a subsample of observations with media competition data available from 2010 through 2012. The media competition variable is the rank of the Herfindahl measure of media competition. Panel C reports the results from regressing forecast properties on indicators of improvement or deterioration of media competition ranks. The sample consists of firm-year observations from 2000 through 2003 and from 2010 through 2012 from 23 countries that have media competition data available for both periods. Post is set to 1 if the observation is from 2010 through 2012, and zero otherwise. Improve is set to 1 if a country’s media competition rank improved by 5 or more over the two periods (i.e., for Germany and Denmark), and 0 otherwise. Worsen is set to 1 if a country’s media competition rank dropped by 5 or more over the two periods (i.e., for Malaysia, Japan, and Ireland), and 0 otherwise. All other variables are defined in the Appendix A. In all panels, the coefficient estimates are from pooled, cross-sectional regressions. The t-statistics are based on Rogers standard errors clustered at the firm level (Petersen, 2009).

, **, and *** represent significance at 10%, 5%, and 1%, respectively.

We perform two additional tests using more recent media competition data. First, we replicate our main results using a subsample of observations from 23 countries with media competition data available from 2010 through 2012 (Panel B). Second, we perform a regression using changes in media competition. The last two columns in Appendix B show that the annual change in media competition is small for most countries. To address this moderate year-to-year change, we form a new sample using observations from 2000 through 2003 (the pre period) and from 2010 through 2012 (the post period). We identify countries with a relatively large change in their media competition rankings, and use Improve and Worsen, respectively, to indicate improvements and deteriorations in the media competition rankings of these countries.¹⁵ Next, we regress the forecast properties on Improve and Worsen and on their interactions with Post (which is an indicator variable for the latter sample period). We report the results in Panel C. Consistent with the main findings, Post*Improve is negatively related to AbsFError (p values < .01), suggesting that analyst forecast errors are significantly smaller when country-level media competition becomes stronger. In contrast, the coefficient on Post*Worsen is positive and significant (p values < 01), suggesting that forecast errors are larger when the media industry becomes less competitive at the country-level. The results for forecast bias (FBias) also support the main findings, but results for forecast dispersion (FDisp) are only significant when media competition is weakened (i.e., for Post*Worsen). Overall, results from the analyses in Table 9 generally support our main results.

Cross-Sectional Variation in Media Impact

In this subsection, we further investigate whether and how the effect of media competition on the properties of analyst forecasts varies with the degree of information uncertainty faced by analysts and/or the ability of analysts to obtain high-quality, firm-specific information. To do this, we identify situations where information from nonmedia channels is likely to be of lower quality, and for each situation, we partition the full sample into three subsamples based on the quality of the information environment. We then test whether the effect of the media characteristics on the properties of analyst forecasts differs for firms in the top and the bottom terciles.

First, prior studies document that institutional investors monitor firms’ financial reporting quality, thereby reducing earnings management and enhancing voluntary disclosure (Ajinkya et al., 2005; Ayers et al., 2011). To the extent that greater institutional ownership fosters a better firm-specific information environment, thereby reducing analysts’ reliance on information provided by the media, we expect media competition to have a smaller effect on the properties of analyst forecasts for firms with higher institutional ownership. Table 10, Panel A reveals that our previous findings on the effects of media competition on analyst forecast error and analyst forecast bias are more pronounced for firms with lower institutional ownership and the difference in the coefficients on MediaCompetition_AVE across the two subsamples are statistically significant (p values < .01).

Table 10.

Cross-Sectional Tests on AbsFError, FBias, and FDisp.

Panel A Institutional ownership
	Low				High				Diff
	Coeff.	t-stat.	N	Adj. R²	Coeff.	t-stat.	N	Adj. R²
AbsFError	−0.048***	−10.54	47,859	0.256	−0.015***	−4.49	37,811	0.186	0.033***
FBias	0.025***	7.00	47,859	0.223	0.007**	2.49	37,811	0.146	−0.018***
FDisp	−0.009***	−4.52	32,104	0.241	−0.005**	−2.74	31,770	0.183	0.004
Panel B Number of analysts following
	Low				High				Diff
	Coeff.	t-stat.	N	Adj. R²	Coeff.	t-stat.	N	Adj. R²
AbsFError	−0.060***	−10.52	33,346	0.245	−0.013***	−4.45	39,142	0.217	0.047***
FBias	0.026***	5.54	33,346	0.234	0.007***	3.12	39,142	0.187	−0.019***
FDisp	−0.018***	−6.91	40,806	0.207	−0.007**	−3.41	38,872	0.242	0.011***
Panel C Brokerage house size
	Low				High
	Coeff.	t-stat.	N	Adj. R²	Coeff.	t-stat.	N	Adj. R²	Diff
AbsFError	−0.039***	−8.66	37,805	0.236	−0.019***	−4.24	37,995	0.231	0.020***
FBias	0.012***	3.48	37,805	0.229	0.014***	3.85	37,995	0.180	0.002
FDisp	−0.010***	−4.94	19,134	0.211	−0.009***	−3.68	30,673	0.239	0.001
Panel D Accounting return on equity (ROE)
	Low				High				Diff
	Coeff.	t-stat.	N	Adj. R²	Coeff.	t-stat.	N	Adj. R²
AbsFError	−0.064***	−8.76	37,811	0.261	−0.009***	−2.81	37,812	0.278	0.055***
FBias	0.041***	6.67	37,811	0.218	0.001	0.45	37,812	0.212	−0.040***
FDisp	−0.013***	−3.75	23,747	0.227	−0.003**	−2.26	28,845	0.263	0.010***

Note.Table 10 reports the results from regressing analyst forecast properties on media competition (MediaCompetition_AVE) and control variables for subsamples partitioned on factors related to uncertainty in the information environment, specifically institutional ownership (in Panel A), the number of analysts following (in Panel B), brokerage house size (in Panel C), and accounting return on equity (ROE) (in Panel D). The “Low” (“High”) subsample consists of firm-year observations with values of the partitioning factor in the bottom (top) tertile. All control variables in the main regression model are included. “Diff” is the difference in the estimated coefficients on media competition between the low and high subsamples. The last two columns report the change in the forecast property (scaled by the mean value of that forecast property) for a one-standard deviation change in media competition. In all tests, the coefficient estimates are from pooled, cross-sectional regressions with year and industry indicators. Following Hope (2003), the industry fixed effects are based on one-digit SIC codes. The t-statistics are based on Rogers standard errors, which are clustered at the firm level and control for serial correlation and heteroscedasticity (Petersen, 2009). All variables are defined in the Appendix A.

, **, and *** represent significance at 10%, 5%, and 1% (two-tailed test), respectively.

Second, research shows that analyst following improves a firm’s information environment. For example, Yu (2008) finds that firms followed by more analysts manage their earnings less, and Anantharaman and Zhang (2011) find that firms provide additional disclosure to increase analyst coverage. Thus, we expect the effect of media competition on properties of analyst forecasts to be greater when firms have relatively low analyst following. Panel B reveals that the effects of media competition on all three forecast properties are more significant for firms with fewer analysts following (p values < .01).

Third, prior studies find that analysts employed by larger brokerage houses provide more accurate forecasts, presumably because of resource availability (Clement, 1999). Thus, we expect the effects of media competition on the properties of analyst forecasts to be stronger for firms that are followed by analysts who work for relatively smaller brokerage houses and hence, should tend to rely more on information obtained from the media. Panel C reveals that the effect of media competition on analyst forecast error (but not on the other forecast properties) is more pronounced for firms followed by analysts from smaller brokerage houses.

Finally, in Panel D, we examine the influence of accounting performance (measured by ROE) on the effects of media competition on analyst forecast properties because prior research finds that firms experiencing weaker financial performance provide relatively low-quality disclosures (Lang & Lundholm, 1993) and that firms with negative earnings have low earnings quality (Dechow & Dichev, 2002; Francis et al., 2004). We find that the effects of media competition on all three forecast properties are significantly greater for firms with relatively low ROE than for firms with relatively high ROE (p values < .01). These results are consistent with analysts relying more on the media for firm-specific information when accounting performance is weaker. Overall, results in Table 10 suggest that when information from nonmedia channels is likely to be of lower quality media competition is more important for the analyst forecast properties.

Additional Tests

We conduct a series of additional tests to confirm the robustness of our main findings, and our results hold in each of the test as described below. First, although the press is a major source of financial news, other types of media can play a role in shaping the information environment. Thus, we construct additional variables that measure television competition, television penetration, and television freedom, and include them in our models. We also include a variable for foreign media ownership given its growing prevalence in the media sector in transition economics (Djankov et al., 2003). Second, because information communicated directly by managers through earnings forecasts and conference calls could be positively associated with media characteristics, we re-estimate our regressions adding indicator variables for the issuance of management earnings forecasts and for management conference calls in the year. Third, because analysts can access information from their local media about the foreign firms that they follow, we re-estimate our models removing all cross-listed firms. Fourth, to ensure that our main results are not driven by U.S. firms, which account for 31% of our sample, we re-estimate our main model after removing U.S. firms. Fifth, because the degree of capital market development may affect the quality of information and the quality of financial analysts in a country, we re-estimate the regression adding the ratio of the country’s stock market capitalization to GDP. Finally, competition in different markets (for example, in the capital and product markets) could be correlated within a country, and high competition in these markets could lead to greater information generation. For example, Haw et al. (2015) show that product market competition positively affects analyst forecast activities. We therefore include the country-level product competition variables in our regressions. Our inferences hold in each of these tests.

Possible Underlying Channel

The positive association between media competition and better analyst forecast properties that we document can occur for at least two reasons. First, strong media competition could lead to a greater amount of information being provided or to better quality information, which would in turn reduce the information acquisition costs faced by financial analysts. This should allow analysts in countries with stronger media competition to follow a greater number of firms. Alternatively, Mullainathan and Shleifer (2005) suggest that greater media competition could lead to less reliable information but readers could use the greater volume of information to form unbiased beliefs; this implies that analysts could improve their forecast quality by collecting and analyzing information from a more comprehensive set of news sources. Because this requires more time and greater effort from analysts, however, we would expect to observe analysts following fewer firms when media competition is stronger. To shed light on the possible underlying channel through which media competition affects analyst forecasts, we examine the effect of media competition on the average number of firms that analysts follow. Results presented in Table 11 reveal that analysts in countries with stronger media competition tend to follow more firms. This is consistent with the argument that stronger media competition leads to a better information environment by providing more information and higher quality information.

Table 11.

The Relation Between Media Competition and the Number of Firms Followed.

	NFirm
	Coeff.	t-stat.
Intercept	−0.858	−1.18
Media Competition_AVE	2.569***	11.40
Media Penetration	8.100***	23.96
Media Freedom	−2.217	−1.26
Size	0.593***	4.77
ROE	0.103**	1.97
Loss	−0.493***	−6.85
VarEarn	−0.034	−1.33
Leverage	0.153	1.03
Opacity	−0.112**	−2.11
BIG4	0.384***	4.61
Institutional Ownership	0.306***	2.90
CeoOwnership	−0.283	−0.51
NAnalyst	−0.137***	−28.02
StkExch	−0.700***	−8.95
GExp	0.291***	34.20
NCountry	1.317***	35.87
NForecast	0.062***	4.05
BSize	−7.171***	−9.55
Horizon	3.764***	9.53
Disclosure	3.704***	11.35
RuleLaw	−0.198	−1.41
IFRS	−2.738***	−29.83
Ratio	3.876***	5.93
Year Indicators	Yes
Industry Indicators	Yes
N	113,436
Adj. R²	0.223

Note.Table 11 reports the results from regressing the median number of firms followed by all analysts following the firm in the year (NFirm) on the media characteristics and control variables. The coefficient estimates are from a pooled, cross-sectional regression with year and industry indicators. The industry indicators are based on one-digit SIC codes. All variables are defined in the Appendix A. The t-statistics are based on Rogers standard errors, which are clustered at the firm level and control for serial correlation and heteroscedasticity (Petersen, 2009).

, **, and *** represent significance at 10%, 5%, and 1%, respectively.

Conclusion

In this study, we examine whether and how media competition affects analyst forecast accuracy, bias, and dispersion in an international setting. Controlling for media penetration and media freedom, we find strong and robust evidence that stronger media competition improves the information environment, as reflected in the superior forecasting performance of financial analysts. We also find that the effects of media competition on the properties of analyst forecasts are generally stronger when information from nonmedia channels is perceived to be of lower quality. Finally, we show that analysts in countries with stronger media competition tend to follow more firms, supporting the conjecture that stronger media competition reduces the information acquisition costs faced by financial analysts. Taken together, our results confirm that media competition is an important determinant of analyst forecast quality. Because analyst forecasts help investors to form expectations about future earnings, our findings should be of interest to media regulators, market participants, and researchers.

Footnotes

Appendix

Appendix B.

Comparing 2003 Measure of Media Competition (MediaCompetition_HH) With Data From 2010 Through 2012.

	Djankov et al. (2003)	World Press Trends (2010-2012)	Rank of (I)	Rank of (II)	Change 2010-2011	Change 2011-2012
	I.	II.	III.	IV.	V.	VI.
USA	−0.003	−0.005	1	2	0.01%	0.17%
India	−0.008	−0.003	2	1	−0.03%	−0.06%
France	−0.016	−0.020	3	4	−0.10%	0.57%
Canada	−0.020	−0.018	4	3	0.03%	0.13%
Spain	−0.025	−0.028	5	6	−0.04%	−0.12%
Switzerland	−0.034	−0.041	6	11	−0.30%	−0.12%
Italy	−0.035	−0.029	7	7	2.67%	0.45%
Sweden	−0.036	−0.035	8	8	0.07%	−1.11%
Japan	−0.040	−0.079	9	18	0.24%	0.18%
Malaysia	−0.044	−0.061	10	16	−0.35%	1.94%
Norway	−0.045	−0.038	11	9	−0.08%	−0.20%
Netherlands	−0.050	−0.038	12	10	−0.08%	2.70%
Portugal	−0.052	−0.056	13	15	1.80%	2.57%
Korea, South	−0.053	−0.052	14	13	−1.26%	1.37%
Finland	−0.055	−0.053	15	14	−0.11%	−0.07%
UK	−0.071	−0.070	16	17	−0.36%	2.75%
South Africa	−0.083	−0.088	17	19	−1.64%	−0.96%
Argentina	−0.109	−0.120	18	21	−1.23%	−1.24%
Denmark	−0.130	−0.047	19	12	1.73%	−1.86%
Austria	−0.177	−0.204	20	22	2.01%	−1.84%
Ireland	−0.200	−0.108	21	20	1.13%	−0.12%
Singapore	−0.220	−0.218	22	23	−1.86%	−0.66%
Germany	−0.230	−0.027	23	5	0.12%	−0.14%
Average (mean)					0.16%	0.25%

Note. Column I reports the value of MediaCompetition_HH based on data from Djankov et al. (2003). Column II reports the average value of MediaCompetition_HH from 2010 through 2012 based on data from World Press Trends. Columns III and IV report the rankings of MediaCompetition_HH based on the Djankov et al. (2003) data and on World Press Trends data, respectively. Columns V and VI report the annual change of MediaCompetition_HH between 2010 and 2012. All variables are defined in the Appendix A.

Acknowledgements

We appreciate helpful comments from Rashad Abdel-Khalik, Kris Allee, TJ Atwood, Mark Bradshaw, Kevin Butler, Cory Cassell, Fabio Gaertner, Zhaoyang Gu, Bowe Hansen, Brian Mayhew, Robbie Moon, James Myers, Ken Peasnell, Phil Shane, Terry Warfield, TJ Wong, Steven Young, and other workshop participants at Chinese University of Hong Kong, Florida State University, Lancaster University, University of Arkansas, University of Wisconsin, and York University, as well as participants at the 37th Annual Congress of the European Accounting Association, the 2014 AAA Financial Accounting Reporting Section Midyear Meeting, and the 2014 AAA International Accounting Section Midyear Meeting.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Linda Myers gratefully acknowledges financial support from the Haslam Chair of Business and the William B. Stokely Faculty Research Fellowship at the University of Tennessee, Knoxville.

Data Availability

All data used in this study are publicly available.

ORCID iD

Albert Tsang

Notes

References

Agarwal

Hess

(2012). Common factors in analysts’ earnings revisions: The role of changing economic conditions (Working paper). Georgia State University; University of Cologne.

Ajinkya

Bhojraj

Sengupta

(2005). The association between outside directors, institutional investors and the properties of management earnings forecasts. Journal of Accounting Research, 43, 343–376.

Anantharaman

Zhang

(2011). Cover me: Managers’ responses to changes in analyst coverage in the post-regulation FD period. The Accounting Review, 86, 1851–1885.

Angrist

J. D.

Pischke

J.-S.

(2009). Mostly harmless econometrics: An empiricist’s companion. Princeton University Press.

Ayers

Ramalingegowda

Yeung

P. E.

(2011). Hometown advantage: The effects of monitoring institution location on financial reporting discretion. Journal of Financial Economics, 42, 41–61.

Bae

K. H.

Tan

Welker

(2008). International GAAP differences: The impact on foreign analysts. The Accounting Review, 83, 593–628.

Bandyopadhyay

(2006). Knowledge-driven economic development (Working paper). Oxford University; London School of Economics.

Barniv

Hope

Myring

Thomas

(2010). International evidence on analyst stock recommendations, valuations, and returns. Contemporary Accounting Research, 27, 1131–1167.

Barniv

Myring

Thomas

(2005). The association between the legal and financial reporting environments and forecast performance of individual analysts. Contemporary Accounting Research, 22:, 727–758.

10.

Baron

D. P.

(2010). Business and the environment (6th ed.). Pearson Education.

11.

Becker

L. B.

Hollifield

C. A.

Jacobsson

E.-M.

Vlad

(2009). Is more always better? Examining the adverse effects of competition on media performance. Journalism Studies, 10(3), 368–385.

12.

Besley

Prat

(2006). Handcuffs for the grabbing hand? Media capture and government accountability. American Economic Review, 96, 720–736.

13.

Bhattacharya

Daouk

Welker

(2003). The world price of earnings opacity. The Accounting Review, 78, 641–678.

14.

Bilinski

Lyssimachou

Walker

(2013). Target price accuracy: International evidence. The Accounting Review, 88, 825–851.

15.

Bonner

S. E.

Hugon

Walther

B. R.

(2007). Investor reaction to celebrity analysts: The case of earnings forecast revisions. Journal of Accounting Research, 45(3), 481–513.

16.

Bradshaw

M. T.

Wang

Zhou

(2017). Soft information in the financial press and analysts’ recommendation revisions (Working paper). Boston College; The Ohio State University; Baruch College.

17.

Bushee

B. J.

Core

J. E.

Guay

Hamm

S. J. W.

(2010). The role of the business press as an information intermediary. Journal of Accounting Research, 48, 1–19.

18.

Bushman

R. M.

Piotroski

J. D.

Smith

A. J.

(2004). What determines corporate transparency? Journal of Accounting Research, 42, 207–252.

19.

Byard

Shaw

K. W.

(2003). Corporate disclosure quality and properties of analysts’ information environment. Journal of Accounting, Auditing and Finance, 18(3), 355–378.

20.

Cagé

(2020). Media competition, information provision and political participation: Evidence from French local newspapers and elections, 1944-2014. Journal of Public Economics, 185, Article 104077.

21.

Chouliaraki

Morsing

(2010). Media, organization and identity. Macmillan.

22.

Clement

M. B.

(1999). Analyst forecast accuracy: Do ability, resources and portfolio complexity matter? Journal of Accounting and Economics, 27, 285–303.

23.

Clement

M. B.

Rees

L. L.

Swanson

E. P.

(2004). The influence of culture and corporate governance on the characteristics that distinguish superior analysts. Journal of Accounting, Auditing and Finance, 18(4), 593–618.

24.

Dechow

Dichev

(2002). The quality of accruals and earnings: The role of accrual estimation errors. The Accounting Review, 77, 35–59.

25.

DeFond

Hung

(2007). Investor protection and analysts’ cash flow forecasts around the world. Review of Accounting Studies, 12:, 377–419.

26.

Dhaliwal

D. S.

Radhakrishnan

Tsang

Yang

Y. G.

(2012). Nonfinancial disclosure and analyst forecast accuracy: International evidence on corporate social responsibility disclosure. The Accounting Review, 87, 723–759.

27.

Djankov

Mcliesh

Nenova

Shleifer

(2003). Who owns the media? Journal of Law and Economics, 46, 341–381.

28.

Drake

M. S.

Myers

L. A.

(2011). Analysts’ accrual-related over-optimism: Do analyst characteristics play a role? Review of Accounting Studies, 16, 59–88.

29.

Dyck

Zingales

(2002). The corporate governance role of the media. In Islam

(Ed.), The right to tell: The role of mass media in economic development (pp. 107–140). The World Bank Institute.

30.

Dyck

Zingales

(2004). Private benefits of control: An international comparison. The Journal of Finance, 59, 537–600.

31.

Engelberg

J. E.

Parsons

C. A.

(2011). The causal impact of media in financial markets. The Journal of Finance, 66, 67–97.

32.

Fan

(2013). Ownership consolidation and product characteristics: A study of the US daily newspaper market. American Economic Review, 103(5), 1598–1628.

33.

Fang

Peress

(2009). Media coverage and the cross-section of stock returns. The Journal of Finance, 64, 2023–2052.

34.

Francis

Lafond

Olsson

P. M.

Schipper

(2004). Costs of equity and earnings attributes. The Accounting Review, 79, 967–1010.

35.

Gentzkow

Shapiro

J. M.

(2006). Media bias and reputation. Journal of Political Economy, 114, 280–316.

36.

Gentzkow

Shapiro

J. M.

(2008). Competition and truth in the market for news. Journal of Economic Perspectives, 22, 133–154.

37.

Gow

I. D.

Ormazabal

Taylor

D. J.

(2010). Correcting for cross-sectional and time-series dependence in accounting research. The Accounting Review, 85, 483–512.

38.

Griffin

J. M.

Hirschey

N. H.

Kelly

P. J.

(2011). How important is the financial media in global markets? Review of Financial Studies, 24, 3941–3992.

39.

Hail

Leuz

(2006). International differences in the cost of equity capital: Do legal institutions and securities regulation matter? Journal of Accounting Research, 44(3), 485–531.

40.

Haw

I.-M.

S. S. M.

(2010). Analysts’ forecast properties, concentrated ownership and legal institutions. Journal of Accounting, Auditing and Finance, 25(2), 235–259.

41.

Haw

I.-M.

Lee

(2015). Product market competition and analyst forecasting activity: International evidence. Journal of Banking & Finance, 56, 48–60.

42.

Haw

I.-M.

Lee

(2012). Investor protection and price informativeness about future earnings: International evidence. Review of Accounting Studies, 17(2), 389–419.

43.

Hillert

Jacobs

Muller

(2014). Media makes momentum. Review of Financial Studies, 27(2), 3467–3501.

44.

Hollifield

C. A.

Vlad

Becker

L. B.

(2004). Market and organizational factors affecting the success of media organizations in emerging economies. In Picard

R. G.

(Ed.), Strategic responses to media market changes (pp. 133–153). Jönköping University.

45.

Hope

O.-K.

(2003). Disclosure practices, enforcement of accounting standards, and analysts’ forecast accuracy: An international study. Journal of Accounting Research, 41, 235–272.

46.

Hope

O.-K.

Kang

(2005). The association between macroeconomic uncertainty and analysts’ forecast accuracy. Journal of International Accounting Research, 4(1), 23–38.

47.

Huang

A. G.

Mamo

K. Y.

(2016). Do analysts read the news? (Working paper). Hong Kong University of Science and Technology; University of Waterloo.

48.

Hugon

Kumar

Lin

(2016). Analysts, macroeconomic news, and the benefit of active in-house economists. The Accounting Review, 91(2), 513–534.

49.

Isidro

Nanda

Wysocki

P. D.

(2016). Financial reporting differences around the world: What matters? (Working paper). ISCTE-IUL Instituto Universitário de Lisboa; University of Miami; Boston University.

50.

Jacob

Lys

T. Z.

Neale

(1999). Expertise in forecasting performance of security analysts. Journal of Accounting and Economics, 28, 51–82.

51.

Kezdi

(2004). Robust standard error estimation in fixed-effects panel models. Hungarian Statistical Review Special, 9, 96–116.

52.

Kothari

S. P.

Short

J. E.

(2009). The effect of disclosures by management, analysts, and business press on cost of capital, return volatility, and analyst forecasts: A study using content analysis. The Accounting Review, 84(5), 1639–1670.

53.

La Porta

Lopez-De-Silanes

Shleifer

. (2006). What works in securities laws? The Journal of Finance, 61, 1–32.

54.

Lacy

S.T. F.

Simon

T. F.

(1993). The economics and regulation of United States newspapers. Ablex Publishing Corporation.

55.

Lang

Lins

K. V.

Maffett

(2012). Transparency, liquidity, and valuation: International evidence on when transparency matters most. Journal of Accounting Research, 50, 729–774.

56.

Lang

Lundholm

(1993). Cross-sectional determinants of analyst ratings of corporate disclosures. Journal of Accounting Research, 31, 246–271.

57.

Lau

S. T.

Zhang

(2012). Information environment and equity risk premium volatility around the world. Management Science, 58(7), 1322–1340.

58.

E. X.

Ramesh

Shen

(2011). The role of newswires in screening and disseminating value-relevant information in periodic SEC reports. The Accounting Review, 86(2), 669–701.

59.

Ljungqvist

Marston

Starks

L. T.

Wei

K. D.

Yan

(2007). Conflicts of interest in sell-side research and the moderating role of institutional investors. Journal of Financial Economics, 85, 420–456.

60.

Mazza

Pedersen

J. S.

(2004). From press to e-media? The transformation of an organizational field. Organization Studies, 25, 875–895.

61.

Mullainathan

Shleifer

(2005). The market for news. The American Economic Review, 95, 1031–1053.

62.

Peress

(2014). The media and the diffusion of information in financial markets: Evidence from newspaper strikes. The Journal of Finance, 69(5), 2007–2043.

63.

Petersen

M. A.

(2009). Estimating standard errors in finance panel data sets: Comparing approaches. Review of Financial Studies, 22, 435–480.

64.

Roth

Wald

J. K.

(2010). Political rights and the cost of debt. Journal of Financial Economics, 95, 202–226.

65.

Rees

Sharp

Twedt

(2015). Who’s heard on the street? Determinants and consequences of financial analyst coverage in the business press. Review of Accounting Studies, 20(1), 173–209.

66.

Rennhoff

A. D.

Wilbur

K. C.

(2012). Local media ownership and media quality. Information Economics and Policy, 24, 231–242.

67.

Sonney

(2009). Financial analysts’ performance: Sector versus country-specialization. Review of Financial Studies, 22, 2081–2131.

68.

Twedt

(2016). Spreading the word: Price discovery and newswire dissemination of management earnings guidance. The Accounting Review, 91(1), 317–346.

69.

World Association of Newspapers and News Publications. (2010). World press trends. https://www.wan-ifra.org/

70.

World Association of Newspapers and News Publications. (2014). World press trends. https://www.wan-ifra.org/

71.

The World Bank Institute. (2002). The right to tell: The role of mass media in economic development.

72.

Yildirim

Gal-Or

Geylani

(2013). User-generated content and bias in news media. Management Science, 59(12), 2655–2666.

73.

(2008). Analyst coverage and earnings management. Journal of Financial Economics, 88, 245–271.