International evidence on the relative importance of the determinants of earnings forecast accuracy

Abstract

We analyze earnings forecasting errors made by financial analysts for 18 developed countries over the 1990–2006 period. We use the Heston–Rouwenhorst approach to unravel country-, industry-, and firm-specific effects as a source of variation in financial analysts’ earnings forecast errors. We first estimate each effect with a dummy variable regression, and then decompose the variance of forecast errors into different effects. We provide evidence that the differences between countries, industrial sectors, and analyst-following offer a weak explanation for differences in forecast errors. Country effects however largely dominate industry and analyst-following effects. By contrast, the type of earnings (profits or losses)—and variations in earnings (increases or decreases) play a significant role in the forecast accuracy of financial analysts.

Introduction

As reported by Kothari (2001), Easton, Taylor, Shroff, and Sougiannis (2002), and more recently by Frankel, Kothari, and Weber (2006), earnings forecasts are more than ever a crucial topic for capital markets researchers and investors. Four motivations justify the important role of financial analysts’ forecasts (hereafter referred to as FAFs).¹ First, “almost all models of valuation either directly or indirectly use earnings forecast”.² Second, the research on financial statement information and security returns often requires a model of expected earnings. Third, the predictability of stock returns would be associated with FAFs. Fourth, as a source of information in the capital markets, FAFs tend to influence the level and variability of security prices. Frankel et al. (2006) have recently analyzed the determinants of the magnitude of stock price reaction to analyst reports. Therefore, there is a large literature acknowledging that FAFs are an important input to investment decisions.³ With the tremendous increase of international investment activities during the past two decades, there is an urgent demand to improve our understanding of FAFs, especially in an international context.⁴

Forecast accuracy is defined as the absolute earnings forecast error, and forecast bias as the relative earnings forecast error. Much work has been dedicated to the accuracy and bias, with research in the field focusing largely on the U.S. market. Among the most documented determinants of the accuracy of FAFs are earnings type – profits vs. losses, increases vs. decreases – (Ciccone, 2005, Dowen, 1996); the firm's business activities (Dunn & Nathan, 1998); the economic situation (Chopra, 1998); the forecast horizon (Richardson, Teoh, & Wysocki, 1999); the industrial sector (Brown, 1997); and the analysts’ competence (Mikhail, Walther, & Willis, 1997). Most of these studies provide U.S. evidence for the accuracy and bias of FAFs, and generally each study focuses on a single determinant. They do not allow for a proper evaluation of the accuracy and bias of FAFs in different environments. Recently, some articles that have taken an interest in FAFs around the world, have shown significant differences in their respective accuracy levels (Ang and Ciccone, 2001, Capstaff et al., 1998, Chang et al., 2000, Hope, 2003). These studies try to explain the reasons for the differences unmasked, underscoring worldwide determinants of the accuracy of FAFs. Studies led by Allen, Cho, and Jung (1999), Chang et al. (2000), Ang and Ciccone (2001), and Black and Carnes (2006), among others, document that accounting, legal and economic systems tend to have an important impact on the accuracy of forecasts. The accounting, legal and institutional environments are the most obvious country-related determinants of FAF accuracy. Further, as pointed out by Bhattacharya, Daouk, and Welker (2003) and Leuz, Nanda, and Wysocki (2003), there are systematic differences in the way earnings are managed in different countries around the world. Beyond the type of earnings effect largely documented in the U.S., these studies highlight the importance of country and industry effects. Although Hope (2003) shows that firm-specific factors (profits vs. losses or increases vs. decreases) are the most important in explaining the characteristics of FAFs, international studies on the determinants of forecast errors focus almost exclusively on the different aspects of the country effect.⁵

To our knowledge, no study has analyzed the relative importance of country-, industry- and firm-specific effects in explaining the cross-sectional variance in FAF errors.⁶ The question is nonetheless a fundamental one for financial analysts, international investors, and capital markets researchers. Cavaglia, Brightman, and Aked (2000) provide evidence that sector factors became more important determinants of the stock returns of developed countries in the late 1990s and early 2000s. In the near future, the international portfolios of developed stock markets could be structured along sector lines rather than along country lines. As recently reported by Eun, Huang, and Lai (2008), this view is not unanimously held; the relative importance of country vs. industry factors is a significant unsettled issue.⁷ While Brooks and Del Negro (2004) argue that the rising importance of industry factors relative to country factors is a temporary phenomenon associated with the stock market fluctuations, Moerman (2008), focusing on Euro area stock markets, finds strong evidence that “diversification over industries yields more efficient portfolios than diversification over countries”. Nevertheless, we observe with Ferreira and Ferreira (2006) that the dominance of the country effect has diminished, while the industry effect has increased, more specifically on European stock markets.

The variation of FAF accuracy and bias across markets and industries motivate the three questions we attempt to answer in this paper. (1) Is the pure country effect more important than the pure industry effect in the explanation of FAF accuracy and bias? (2) Are firm-specific effects more relevant to explain FAF accuracy and bias around the world? (3) What is the evolution of the relative importance of country-, industry- and firm-specific effects in explaining the cross-sectional variance in FAF errors over the period 1990–2006?

Our contribution to the debate on the determinants of the accuracy and bias of FAFs (and indirectly on their impact on international stock returns) is threefold. First, we use a more powerful methodology to separate the relative importance of each class of determinants. This approach differs in several respects from previous studies carried out at the international level. The few previous studies that analyze country effects on forecast accuracy and bias compare the moments and distribution of errors.⁸ This conventional and traditional approach is open to criticism insofar as it is unable to disentangle country-, industry-, and firm-specific effects and to estimate their relative importance. Second we concentrate on a sample of 18 developed countries (excluding the U.S.)⁹ over the 1990–2006 period. Our sample includes (1) countries from Europe, North America and Australasia, which have experienced significant international harmonisation over the last 17 years, and (2) countries with sharply contrasted sectors (Energy in Canada, Finance and Banking in Singapore, Hong Kong and Switzerland). Third, all these regions have implemented significant financial and legal reforms in order to restore trust among investors.¹⁰ This evolving financial context offers the opportunity to analyze the evolution of the factors influencing FAF accuracy and bias.

Section 2 presents and justifies our conceptual framework for analyzing FAF accuracy and bias during the period. Section 3 describes the data source and the forecast error measures used in the analysis. We describe the methodology employed in Section 4, and present results in Section 5. In Section 6, we summarise our main results.