Aggregate Earnings Informativeness and Economic Shocks:

International Evidence

Yuto Yoshinagaa* and Makoto Nakanob

a
Graduate School of Economics and Management, Tohoku University, Sendai, Japan;

b
Graduate School of Business Administration, Hitotsubashi University, Kunitachi, Japan

* Corresponding author; (e-mail address): [email protected], (ORCID iD):

https://orcid.org/0000-0001-7509-787X

Acknowledgments: We are grateful to Hee-Yeon Sunwoo and Bryan Howieson for useful comments.

This work was supported by the Japan Society for the Promotion of Science under Grant [number

19K20877; 20K02033; 20K13635]; and the Management Innovation Research Center at Hitotsubashi

University. No potential conflict of interest exists. Any errors are our own.

Electronic copy available at: https://ssrn.com/abstract=3461082

Aggregate Earnings Informativeness and Economic Shocks:

International Evidence

Our study aims to clarify the use of aggregate earnings to forecast future GDP growth.
Using empirical analyses with global quarterly data, we investigate whether aggregate-
level profitability drivers, which are components of aggregate earnings, are relevant for
forecasting GDP growth. We confirm that aggregate-level profitability drivers are
useful for forecasting future GDP growth worldwide. We also show that considering the
effects of crises improves the forecast model of GDP growth. In addition, aggregate-
level profitability drivers are observed to affect future market returns through the
prediction of GDP growth in developed countries, but not in emerging countries.

Keywords: aggregate earnings; macro-accounting; GDP forecast; DuPont analysis;

economic shock.

Subject classification codes: E01; M41

1. Introduction

The aim of our study is to clarify the proper use of aggregate-level earnings information to

forecast future gross domestic product (GDP) growth of countries. A large body of empirical

accounting research has investigated the usefulness of financial statement ratios. Financial

statement analysis at the firm level has been a recurring topic of academic research in the

accounting field (e.g., Ball and Brown 1968, 2019; Ohlson 1995; Kothari 2001; Courteau et

al. 2006; Penman 2012). These studies aim to investigate the link between earnings and stock

return, to forecast future earnings, or to estimate the intrinsic value of equity at the firm level.

The aim of firm-level research is to contribute to efficient macro-level resource allocation

through investors’ decision making.

Although the usefulness of accounting information is mainly its informativeness for

investors, it has a wider role in practice. For example, economic newspapers and the

government often aggregate accounting information to capture the macroeconomic condition.

Electronic copy available at: https://ssrn.com/abstract=3461082

However, there is little investigation of the link between accounting information and the

macroeconomy in traditional accounting research. We aim to fill this gap between practical

and academic interests.

Recently, a new accounting research area, which Konchitchki (2016) calls ‘macro-

accounting,’ has started to investigate the link between aggregate earnings and

macroeconomic indicators. Konchitchki (2016, 27) states: ‘This new research area focuses on

addressing real-life world problems using the value added that accounting can bring to

various macro-level topics that are at the forefront of the academic and professional

discussions.’ 1 There are two streams in the macro-accounting literature. One focuses on the

earnings–return relationship at the aggregate level (e.g., Kothari, Lewellen, and Warner 2006;

Sadka and Sadka 2009; Kang 2019). The other investigates the link between aggregate

earnings and macroeconomy. This study focuses on the latter.

Prior studies in this area have shown that aggregate earnings contain information

about the macroeconomy, such as future inflation, GDP growth, and economic policy.

Shivakumar (2007) finds that aggregate earnings growth is positively associated with future

inflation levels. Konchitchki and Patatoukas (2014a, 2014b) show that aggregate earnings

changes improve predictions of future GDP growth by macroeconomists. Gallo, Hann, and Li

(2016) find that aggregate earnings can predict future monetary policy.

Konchitchki and Patatoukas (2014b) thoroughly investigate the information contents

of aggregate earnings based on the DuPont analysis framework. They decompose return on

net operating assets (RNOA) at the aggregate level into asset turnover (ATO) and return on

1
Ball and Sadka (2015, 56) emphasize the growing importance of this research stream: ‘In sum, we

view the aggregate earnings literature as a comparatively recent but growing body of work with a

promising future.’

Electronic copy available at: https://ssrn.com/abstract=3461082

sales (ROS). Prior studies provide evidence that RNOA and its drivers are useful for

forecasting economic activity and stock return at the firm level (Soliman 2008; Penman

2012). Konchitchki and Patatoukas (2014b) apply the DuPont analysis framework to the

macro level. They report that aggregate-level profitability drivers are useful for improving the

forecast accuracy of GDP growth in the United States (US). In addition, Konchitchki and

Patatoukas (2014b) investigate how aggregate-level accounting information affects investors’

activity. They report that aggregate-level profitability drivers affect future market return only

through the prediction of GDP growth. This result suggests that investors interpret future

economic growth from aggregate-level earnings and make an investment decision in the US

market.

However, prior studies have two limitations. First, they do not consider the effects of

financial crises. Although several crises have strongly damaged the world economy, prior

studies do not consider the potential effects of the crises on the informativeness of aggregate

earnings. Recent studies show that accounting quality strengthens during crisis periods (Filip

and Raffournier 2014; Arthur, Tang, and Lin 2015). Doukakis et al. (2020) suggest that

financial crises can change the relationship between the macroeconomy and the profitability

of listed firms. Based on these studies, financial crises alter the effects of aggregate-level

fundamentals and GDP growth. Considering the effects of crises improves the forecast model

of GDP growth with aggregate-level profitability drivers. We mainly investigate whether the

usefulness of the forecast model increases after including the effects of crises.

The second limitation is external validity, which arises from the use of only US data in

the empirical analyses. The findings of analyses using the data of one specific country can

yield implications for other countries. However, we cannot extend the findings to other

countries without checking the suitability of doing so. In fact, many non-US accounting

studies report different results from those of US research (e.g., Gabrielsen, Gramlich, and

Electronic copy available at: https://ssrn.com/abstract=3461082

Plenborg 2002; Rahman, Yammeesri, and Perera 2010; Gajewski and Quéré 2013).

Furthermore, the effects of aggregate earnings can differ between developed and emerging

markets. The information environment is less sophisticated in emerging countries (Lopes and

de Alencar 2010). On the one hand, since less information is available in emerging countries,

the relative impacts of aggregate earnings could be stronger in emerging markets. On the

other hand, developed countries have more sophisticated markets, which reflect aggregate-

level profitability timeously. Thus, our sample covers both developed and emerging countries.

We obtain three main findings. First, we confirm that aggregate-level profitability

drivers are useful for forecasting future GDP growth worldwide. We show that containing the

drivers improves the forecast accuracy of the model globally. Second, we show that

considering the effects of crises, in addition to the aggregate-level profitability drivers,

improves the forecast model of GDP growth. Third, we find that the route of impacts of

aggregate-level profitability drivers on future market returns through the prediction of GDP

growth is observed only in developed countries.

Our study contributes to macro-accounting research in three ways. First, prior studies

rely solely on US data. Our study covers 21 countries, including both developed and emerging

countries, which improves the external validity of our findings. In particular, our research has

implications for economists of emerging countries. Prior research focuses on the

informativeness of aggregate earnings for economists of the US, which is the typical example

of developed countries (Shivakumar 2007; Konchitchki and Patatoukas 2014a, 2014b; Gallo,

Hann, and Li 2016). In contrast, our sample covers emerging countries in addition to

developed countries. We show that including both aggregate-level profitability drivers and

crisis effects raises the accuracy of the forecast model in emerging countries, not only

developed countries. Our findings suggest that economists of emerging countries should

Electronic copy available at: https://ssrn.com/abstract=3461082

incorporate aggregate earnings and crisis effects into their GDP forecasts to obtain more

accurate predictions.

Second, we successfully improve the forecast model of GDP growth by considering

the effects of financial crises. The pioneering works of Konchitchki and Patatoukas (2014a,

2014b) document that aggregate earnings growth is a significant leading indicator of GDP

growth. We upgrade their model by including a crisis dummy and interaction terms. At this

point in time, the real economy is suffering from the recessionary impact of the COVID-19

pandemic. Our results suggest that considering this ongoing recession as a crisis period

improve the forecast accuracy of the model. Therefore, our evidence may make a timely

contribution to current macroeconomic analysis.

Third, we suggest that how aggregate-level profitability drivers affects market return

depends on the development level of the country. Konchitchki and Patatoukas (2014b) report

that aggregate-level profitability drivers affect future market return through the prediction of

GDP growth in the US market. We show that this route of impacts is observed only in

developed countries, not in emerging countries.

We organize the remainder of this paper as follows. Section 2 discusses our

motivation and research question. Then, we present our main research design and sample

selection in Section 3. The empirical results follow in Section 4. We investigate an additional

analysis in Section 5. Section 6 shows the robustness of our results. Finally, Section 7

summarizes our study and provides concluding remarks.

2. Literature Review and Research Question

Empirical accounting research traditionally focuses on firm-level accounting information. For

example, Beaver (1968), Ball and Brown (1968), and their subsequent studies investigate the

Electronic copy available at: https://ssrn.com/abstract=3461082

usefulness of firm-level earnings information for investors. 2 In contrast, some pioneering

macro-accounting studies have inspected accounting information from a different perspective

more recently. They calculate the cross-sectional average or sum of firm-level accounting

information of listed firms in a country and investigate such aggregate-level accounting

information (e.g., Ball, Sadka, and Sadka 2009; Sadka and Sadka 2009).

In macro-accounting research, two factors support the usefulness of aggregate

earnings for GDP forecast. First, we expect that aggregate earnings have the potential to

forecast future GDP growth. Corporate profits are one component of GDP and they are

correlated with other components. In addition, they are assumed to be a main driver of

economic growth (Konchitchki and Patatoukas 2014a). Second, accounting earnings are

timely information. Listed companies must report quarterly financial statements in some

countries, and those reporting dates are usually earlier than those of macroeconomic

indicators. For instance, Japanese listed companies have to submit quarterly financial reports

within 45 days of the quarter-end date. By contrast, the Quarterly Financial Statements

Statistics of Corporations, which releases aggregate-level accounting data of Japanese

companies, is issued around 60 days after each quarter-end by the Ministry of Finance of

Japan. In addition, prior US studies point out that companies release quarterly accounting

information before some macroeconomic indicators (e.g., Konchitchki and Patatoukas 2014a,

2014b; Nallareddy and Ogneva 2017). Thus, accounting data collected and analyzed from the

quarterly reports of listed firms provide more timely information than do government

documents.

Prior studies empirically support the usefulness of aggregate earnings for forecasting

macroeconomic indicators. The pioneering works of this research stream are Konchitchki and

2
See Kothari (2001) for a detailed review of the related literature.

Electronic copy available at: https://ssrn.com/abstract=3461082

Patatoukas (2014a, 2014b). They investigate the usefulness of aggregate earnings for GDP

forecasts. Konchitchki and Patatoukas (2014a) find that aggregate earnings changes have a

significantly positive relationship with future GDP growth. Konchitchki and Patatoukas

(2014b) report that drivers of aggregate earnings changes, decomposed based on the DuPont

analysis method, are incrementally useful for forecasting GDP growth. However, these

studies cover only the US and do not investigate the external validity of their results around

the world. Prior studies report that empirical results can be different from US evidence when

they use non-US data (e.g., Gabrielsen, Gramlich, and Plenborg 2002; Rahman, Yammeesri,

and Perera 2010; Gajewski and Quéré 2013). Therefore, we first investigate whether

aggregate-level profitability drivers are useful based on the global data of 21 countries.

Furthermore, considering the state of the economy will improve the forecast model of

GDP growth using aggregate-level profitability drivers. Filip and Raffournier (2014) report

that earnings management behavior significantly decreased during the global financial crisis.

Arthur, Tang, and Lin (2015) show that earnings quality improved during the crisis. These

studies show that earnings quality improves in crisis periods. Related to the state of the

macroeconomy, Loh and Stulz (2018) focus on the role of sell-side analysts. They find that

analysts are more valuable in bad times. The stock-price impact of earnings forecast revisions

is greater in bad times. Accounting numbers may convey important information during

economic shocks. In addition, Doukakis et al. (2020) analyze data from 16 European

countries and find that macroeconomic expectations are useful in predicting firm-level future

profitability only during non-crisis periods. Doukakis et al. (2020) suggest that the

relationship between the macroeconomy and the profitability of individual firms can differ

Electronic copy available at: https://ssrn.com/abstract=3461082

depending on the state of the economy. 3 Based on these studies, the effects of aggregate-level

profitability drivers on future GDP growth can differ between crisis periods and non-crisis

periods. Considering the effects of crises improves the forecast model of GDP growth.

Therefore, we hypothesize that containing the crisis effects can upgrade the forecast model of

GDP growth proposed by Konchitchki and Patatoukas (2014b). 4

3. Main Research Design and Sample Selection

3.1. Regression Models

Before examining the effects of financial crisis, we first check whether aggregate-level

profitability drivers improve forecasting future GDP growth using global data. Our models

are Equations (1) and (2), which are based on Konchitchki and Patatoukas (2014b). 5

3
It should be emphasized that our aim is to predict not firm-level profitability but future GDP growth.

In that sense, our study is quite different from Doukakis et al. (2020).
4
Another reason for the usefulness of considering crisis effects is that the forecasting power of other

predictors, market returns, can deteriorate during crisis periods. Although market returns are

usually used to forecast future macroeconomy (Konchitchki and Patatoukas 2014a, 2014b), they

can lose ties with the real economy owing to investor sentiment reflecting severe crisis (cf.,

Veronesi 1999; Williams 2015). Meanwhile, investor sentiment affects accounting information less

than market returns. Owing to the difference in the effect of investor sentiment between stock

return and accounting information, considering economic shock can improve the forecast model

with aggregate earnings for predicting future GDP growth.

5
Konchitchki and Patatoukas (2014b) additionally decompose ROS into the ratio of operating income

before depreciation to sales and the ratio of depreciation to sales. However, unlike sales and

operating profit, the depreciation is not timeously released by quarter in many countries. When we

exclude firm/quarter observations missing from the depreciation data, the number of firm/quarter

Electronic copy available at: https://ssrn.com/abstract=3461082

 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+ℎ = 𝛼𝛼 + 𝛽𝛽1 ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 + 𝛽𝛽2 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞 + 𝛽𝛽3 𝑀𝑀𝑀𝑀𝑐𝑐,𝑞𝑞 + 𝜀𝜀 (1)

∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+ℎ = 𝛼𝛼 + 𝛽𝛽1 ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 + 𝛽𝛽2 ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑞𝑞 + 𝛽𝛽3 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞 + 𝛽𝛽4 𝑀𝑀𝑀𝑀𝑐𝑐,𝑞𝑞 + 𝜀𝜀 (2)

∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+ℎ denotes future GDP growth for country c in quarter q+h (h=1, 2, 3, 4).

∆𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 represents the changes in RNOA. In Equation (2), we decompose ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 into

∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 and ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑞𝑞 , which represent the changes in ROS and those in ATO, respectively.

𝑀𝑀𝑀𝑀𝑐𝑐,𝑞𝑞 is market return. Since ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞 and 𝑀𝑀𝑀𝑀𝑐𝑐,𝑞𝑞 are control variables in prior studies

(Konchitchki and Patatoukas 2014a, 2014b), we include both variables in our models. We

focus on the incremental usefulness of aggregate fundamentals for macroeconomic forecasts.

Thus, we compare the forecast power of our models to the following base model:

∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+ℎ = 𝛼𝛼 + 𝛽𝛽1 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞 + 𝛽𝛽2 𝑀𝑀𝑀𝑀𝑐𝑐,𝑞𝑞 + 𝜀𝜀 (3)

We check adjusted R squares of these models and adopt likelihood-ratio tests to check

whether aggregate-level profit drivers significantly improve the forecasting accuracy.

Equations (1) and (2) add aggregate-level fundamentals to Equation (3). If aggregate

fundamentals are incrementally useful, the adjusted R squares of Equations (1) and (2) should

be higher and the goodness of fit should be significantly higher than those of Equation (3). In

addition, if aggregate-level profitability drivers have significant effects on future GDP growth

in most countries, the estimated coefficients of the variable of interest (∆𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 , ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 ,

and ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑞𝑞 ) should be significant in Equations (1) and (2).

observations becomes less than one-third of the original. If we calculate aggregate-level variables

with these reduced observations, they do not reflect the general performance of listed firms in the

country. Therefore, we do not use depreciation in our empirical analyses.

Electronic copy available at: https://ssrn.com/abstract=3461082

 Second, we check whether considering the effects of crisis improves the forecasting

power. We use the following interaction model with a crisis dummy, Equation (4), and

compare the forecasting power of this model to that of Equation (2). We use Equation (2) for

comparison and do not adopt Equation (1), because Konchitchki and Patatoukas (2014b)

show that the adjusted R square of the decomposed model is higher than that of the model

using changes in RNOA.

∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+ℎ = 𝛼𝛼 + 𝛽𝛽1 ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 + 𝛽𝛽2 ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑞𝑞 + 𝛽𝛽3 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞 + 𝛽𝛽4 𝑀𝑀𝑀𝑀𝑐𝑐,𝑞𝑞

+ 𝛽𝛽5 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 + 𝛽𝛽6 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 ∗ ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑡𝑡 + 𝛽𝛽7 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 ∗ ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 (4)

+ 𝛽𝛽8 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 ∗ ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑡𝑡 + 𝛽𝛽9 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 ∗ 𝑀𝑀𝑀𝑀𝑐𝑐,𝑡𝑡 + 𝜀𝜀

𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 is a dummy variable that takes the value of one when the country/quarter is

during the crisis period, and otherwise zero. If considering the effects of financial crisis

increases the forecast accuracy, then the adjusted R squares of Equation (4) should be higher

than those of Equation (2) and the difference in goodness of fit should be significant.

Even if the forecast accuracy of Equation (4) is higher than that of Equation (2) in the

second test, one concern remains. Aggregate-level profitability drivers might not contribute to

the results. In other words, only the crisis dummy and the interaction term on GDP growth or

market return can improve the accuracy in Equation (4). To tackle this concern, we engage in

the third analysis. To test whether aggregate-level variables raise accuracy in Equation (4), we

compare the following Equation (5) to Equation (4).

∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+ℎ = 𝛼𝛼 + 𝛽𝛽1 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞 + 𝛽𝛽2 𝑀𝑀𝑀𝑀𝑐𝑐,𝑞𝑞 + 𝛽𝛽3 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 + 𝛽𝛽4 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 ∗ ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑡𝑡
(5)
+ 𝛽𝛽5 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 ∗ 𝑀𝑀𝑀𝑀𝑐𝑐,𝑡𝑡 + 𝜀𝜀

Equation (5) excludes accounting variables from Equation (4). If aggregate-level

profitability drivers increase the forecast accuracy in Equation (4), the adjusted R squares of

Electronic copy available at: https://ssrn.com/abstract=3461082

Equation (4) should be higher than those of Equation (5). We should also observe a

significant difference in the goodness of fit between the two models.

We employ heteroskedasticity-consistent standard errors proposed by White (1980)

and control country fixed effects when running these regressions. Including interaction terms

raises the possibility of the problem of multicollinearity biasing the regression results in

general. We check that the maximum of the variance inflation factor (VIF) in each regression

is lower than 10, indicating that multicollinearity does not bias our results.

3.2. Variable Definitions

We collect quarterly data and aggregate firm-level variables to calculate aggregate-level

accounting fundamentals. We define aggregate-level RNOA (𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 ) as the sum of

operating income divided by the sum of the average net operating assets. Following

Konchitchki and Patatoukas (2014b), we define net operating assets as operating assets

(which are total assets minus cash and short-term investments) minus operating liabilities

(which are total liabilities minus long- and short-term debt). We decompose 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 into

ROS (𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 ) and net operating ATO (𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑞𝑞 ) as Equation (6). 𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 denotes the sum of

operating income divided by the sum of sales. 𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑞𝑞 represents the sum of sales divided by

the sum of average net operating assets.

𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 =
𝑁𝑁𝑁𝑁𝑁𝑁 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴

𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 (6)

= ×
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑁𝑁𝑁𝑁𝑁𝑁 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴

= 𝑅𝑅𝑅𝑅𝑅𝑅 × 𝐴𝐴𝐴𝐴𝐴𝐴

Electronic copy available at: https://ssrn.com/abstract=3461082

 We calculate the seasonal (year-over-year) differences of these fundamentals for the

regression models to reduce the seasonal effects (∆𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 , ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 , ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑞𝑞 ). 𝑀𝑀𝑀𝑀𝑐𝑐,𝑞𝑞 is the

log return of the stock market index in country c. We measure market return over the 12

months leading to the end of quarter q, because Konchitchki and Patatoukas (2014b, Table 3)

show that a 12-month return is the most useful. ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞 is seasonal (year-over-year) log

growth of real GDP.

𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 is a crisis dummy. It takes the value of one when the country/quarter occurs

during a crisis period, and is zero otherwise. Our sample covers three crises: the burst of the

dot-com bubble in 2001, the global financial crisis from 2007 to 2009, and the ongoing

COVID-19 recession in 2020. Thus, if we calculate at least one of the variables using data

during these periods, we classify the country/quarter as the observation in the ‘crisis period.’

We regard the others as ‘non-crisis periods.’

3.3. Sample Selection

Our main data source is the S&P Capital IQ. We obtain the financial data of listed firms,

stock market index, and real GDP from the database. We convert financial data into the

dominant currency at the historical rate. We determine the dominant currency by country; it is

the currency that most firm/quarter observations adopt as the financial reporting currency in

the country. After collecting financial data, we impose the following requirements on the

firm/quarter observations.

(1) Quarterly accounting information is available within 3 months after the end of the

fiscal quarter.

(2) The main SIC codes are not 6000–6499 or 6700–6999.

(3) The fiscal year end is at the end of March, June, September, or December.

(4) Missing variables do not exist.

Electronic copy available at: https://ssrn.com/abstract=3461082

 (5) Total revenues and net operating assets are positive.

We impose Requirement (1), because we can use accounting information collected

within 3 months after the fiscal quarter-end to forecast GDP growth one to four quarters

ahead. We impose Requirement (2) to exclude financial firms, Requirement (3) to match

fiscal quarters, and Requirement (4) to remove observations with missing variables.

Requirement (5) dispels concerns that our fundamentals of interest can have negative

denominators.

After imposing these data requirements, we select country/quarters that contain 50 or

more firm/quarters, because aggregating only a few firms cannot diversify firm-specific

information. We delete observations without all the necessary macroeconomic variables

(𝑀𝑀𝑀𝑀𝑐𝑐,𝑞𝑞 , and ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞 ). Finally, we choose countries that have at least 20 country/quarter

observations.

Our final sample contains 1,239 country/quarters covering 21 countries (Argentina,

Brazil, Canada, Denmark, Finland, Germany, Greece, Hong Kong, Italy, Japan, Mexico, Peru,

the Philippines, Singapore, South Korea, Spain, Sweden, Taiwan, Thailand, Turkey, and the

US). Table 1 presents the descriptive statistics (Panel A) and correlation matrix for the

variables of interest (Panel B). Table 2 indicates some statistics for each country.

(Insert Tables 1 and 2 around here)

4. Empirical Results

4.1. Improvement of Forecast Accuracy of GDP Growth with Aggregate-level

Profitability Drivers
(Insert Tables 3 and 4 around here)

First, we consider the usefulness of aggregate-level profitability drivers to forecast future

GDP growth in the global context. Table 3 shows the regression results of Equations (1) and

Electronic copy available at: https://ssrn.com/abstract=3461082

(3), and Table 4 indicates those of Equations (2) and (3). Our focus is whether aggregate-level

fundamentals are useful for predicting GDP growth. Therefore, we check their incremental

usefulness. We compare the adjusted R squares of the model with and without aggregate-level

accounting variables. We also investigate whether the difference in the fitness of the model is

significant using chi squares estimated by likelihood-ratio tests.

In Table 3, some of the adjusted R squares do not change that much when adding

∆𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 and some chi squares are not significant. Meanwhile, all the adjusted R squares

increase after containing aggregate-level decomposed variables in Table 4. When predicting

GDP growth h quarters ahead, the inclusion of ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 and ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑞𝑞 in the forecast models

raises the adjusted R squares by 0.3% (h=1), 0.8% (h=2), 1.4% (h=3), and 0.8% (h=4). The

chi squares in Table 4 are all significant, which means that aggregate-level profitability

drivers significantly improve the goodness of fit. These results are consistent with

Konchitchki and Patatoukas (2014b), because they show that the adjusted R squares of the

decomposed model are higher. In addition, we find significantly positive coefficients on

∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 regressing GDP growth one quarter ahead, consistent with those of Konchitchki and

Patatoukas (2014a, 2014b). Overall, we observe that aggregate-level profitability drivers are

useful for forecasting future GDP growth in many countries.

4.2. Improvement of Forecast Accuracy by Considering Crisis Effects

(Insert Table 5 around here)

In this subsection, we focus on the effects of the crises. Table 5 indicates the results

comparing Equations (2) and (4). This table shows that considering the crisis effects improves

the forecast accuracy of the model. Adjusted R squares largely increase after the inclusion of

the crisis dummy and its interaction terms. When predicting GDP growth h quarters ahead,

inclusion of these variables in the forecast models increases the adjusted R squares by 1.7%

Electronic copy available at: https://ssrn.com/abstract=3461082

(h=1), 3.8% (h=2), 3.3% (h=3), and 1.7% (h=4). In addition, all the chi squares are significant

in Table 5. 6

Table 5 shows that considering the effects of a crisis raises the forecast accuracy of

GDP growth. However, this result could be caused by the crisis dummy (𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 ) and its

interaction terms with non-accounting variables (𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 ∗ ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑡𝑡 , and 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑠𝑠𝑡𝑡 ∗ 𝑀𝑀𝑀𝑀𝑐𝑐,𝑡𝑡 ).

Therefore, we compare the results of Equations (4) and (5). Equation (5) excludes aggregate-

level profitability drivers and their interaction terms. If we do not observe significant

differences between these two models, then aggregate-level profitability drivers do not

contribute to the forecast accuracy.

(Insert Table 6 around here)

Table 6 eliminates this concern. We observe that the increases in adjusted R squares

are all significant, after including aggregate-level fundamentals. Predicting GDP growth h

quarters ahead, including aggregate-level profitability, in the models raises the adjusted R

squares by 0.7% (h=1), 1.0% (h=2), 1.6% (h=3), and 0.5% (h=4). These results show that

aggregate-level fundamentals improve the forecast accuracy of Equation (4).

In summary, we show how to improve the model to forecast GDP growth.

Considering crisis effects raises the accuracy of the forecast model proposed by Konchitchki

and Patatoukas (2014a, 2014b) with aggregate-level profitability.

6
We find that 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 ∗ ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑡𝑡 has significantly positive coefficients but ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑡𝑡 has significantly

negative coefficients, although this is not our main focus. The finding suggests that the positive effects

of aggregate-level changes in ROS on GDP growth reported in Konchitchki and Patatoukas (2014a,

2014b) can be limited during crisis periods.

Electronic copy available at: https://ssrn.com/abstract=3461082

5. Further analysis

5.1. Purpose and Research Design

We next investigate how aggregate-level profitability drivers are reflected in market return to

obtain implications for stock valuation as an additional analysis. Konchitchki and Patatoukas

(2014b, Table 7) show that aggregate-level fundamentals affect market return only through

GDP growth. Their results suggest that investors understand future GDP growth using

aggregate-level profitability drivers, which reflects the investment activity. Since we find that

containing crisis effects and aggregate-level profitability drivers improve the accuracy of

GDP growth forecast globally, we investigate how aggregate-level profitability drivers affect

stock markets using our modified model and global data.

In this analysis, Konchitchki and Patatoukas (2014b) first estimate predictable

information on subsequent GDP growth with aggregate-level profitability drivers. They

extract the information already expected by investors from the predictable information to test

the impacts of this prediction on investors’ decisions. Thus, Konchitchki and Patatoukas

(2014b) adopt two-stage regression in their estimation. In the first regression, they regress

GDP growth one quarter ahead on aggregate-level profitability drivers to obtain fitted values.

They regress these fitted values on the recent market return to obtain residuals. They treat the

residuals as the predicted contents by aggregate-level profitability drivers on GDP growth but

not expected by investors. Konchitchki and Patatoukas (2014b) regress future market return

on these predicted contents and aggregate-level profitability drivers. As a result, only the

predicted contents have significant coefficients, which means that aggregate-level profitability

drivers impact market returns only through the prediction on GDP growth.

We investigate whether this implication can be observed when using global data and

an improved forecast model of GDP growth. We use Equation (8) and regress future market

return on the predictable contents of GDP growth, aggregate-level profitability, the crisis

Electronic copy available at: https://ssrn.com/abstract=3461082

dummy, and its interaction terms. We also use Equation (7), since Konchitchki and

Patatoukas (2014b) report the results only using the predictable contents of GDP growth as an

explanatory variable.

3𝑚𝑚 𝑝𝑝 𝑝𝑝
𝑀𝑀𝑀𝑀𝑐𝑐,𝑡𝑡+1 = 𝛼𝛼 + 𝛽𝛽1 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑡𝑡+1 + 𝛽𝛽2 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 + 𝛽𝛽3 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 ∗ ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑡𝑡 + 𝜀𝜀 (7)

3𝑚𝑚 𝑝𝑝 𝑟𝑟𝑟𝑟𝑟𝑟 𝑟𝑟𝑟𝑟𝑟𝑟

𝑀𝑀𝑀𝑀𝑐𝑐,𝑡𝑡+1 = 𝛼𝛼 + 𝛽𝛽1 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑡𝑡+1 + 𝛽𝛽2 ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑡𝑡 + 𝛽𝛽3 ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 + 𝛽𝛽4 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡

𝑝𝑝 𝑟𝑟𝑟𝑟𝑟𝑟
+ 𝛽𝛽5 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 ∗ ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑡𝑡+1 + 𝛽𝛽6 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 ∗ ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑡𝑡 (8)

𝑟𝑟𝑟𝑟𝑟𝑟
+ 𝛽𝛽6 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 ∗ ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 + 𝜀𝜀

𝑝𝑝
∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑡𝑡+1 is estimated by two-stage regression. In the first regression, we use

Equation (9) to estimate the predictable contents of GDP growth by aggregate-level

profitability drivers. We add the crisis dummy and its interaction terms, because our main

results show the importance of considering the crisis effects.

∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+1 = 𝛼𝛼 + 𝛽𝛽1 ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 + 𝛽𝛽2 ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑞𝑞 + 𝛽𝛽3 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 + 𝛽𝛽4 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 ∗ ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑡𝑡
(9)
+ 𝛽𝛽5 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 ∗ ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 + 𝜀𝜀

� 𝑐𝑐,𝑞𝑞+1 ) with Equation (9), we extract the

After obtaining the fitted values (∆𝐺𝐺𝐺𝐺𝐺𝐺

� 𝑐𝑐,𝑞𝑞+1 on
expected contents by investors from them by the second regression. We regress ∆𝐺𝐺𝐺𝐺𝐺𝐺

3𝑚𝑚 𝑝𝑝
the recent market return (𝑀𝑀𝑀𝑀𝑐𝑐,𝑡𝑡 ) and estimate the residuals (∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑡𝑡+1 ) 7. 𝑀𝑀𝑀𝑀𝑐𝑐,𝑡𝑡+1 is the 3-

7
When we include the crisis dummy and its interaction terms in the second regression, the VIF of

Equations (7) and (8) becomes over 200 and its estimated coefficients are not stable. To address the

multicollinearity concern, we exclude these crisis effects in the second regression model. Instead,

we assume that there is not a huge difference in investors’ predictions of future GDP growth

between during the crisis and non-crisis periods. Because of this treatment, the VIF of Equations

Electronic copy available at: https://ssrn.com/abstract=3461082

month buy-and-hold market return beginning in the periods when the necessary accounting

information is available. Due to our Requirement (1), we assume that quarterly accounting
3𝑚𝑚
information is available within 3 months of the end of the fiscal quarter. Therefore, 𝑀𝑀𝑀𝑀𝑐𝑐,𝑡𝑡+1 is

𝑟𝑟𝑟𝑟𝑟𝑟 𝑟𝑟𝑟𝑟𝑟𝑟
measured beginning 3 months after the end of the fiscal quarter. ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑡𝑡 and ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 are
𝑝𝑝
the residuals estimated by the regression of ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 and ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑞𝑞 on ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑡𝑡+1 , respectively,

following Konchitchki and Patatoukas (2014b). Using these residuals reduces the problem of
𝑝𝑝
multicollinearity, since ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑡𝑡+1 represents the predicted contents by ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 and ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑞𝑞 .

All the regressions in this additional analysis control country fixed effects, including the
𝑝𝑝
� 𝑐𝑐,𝑞𝑞+1, ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑡𝑡+1 𝑟𝑟𝑟𝑟𝑟𝑟 𝑟𝑟𝑟𝑟𝑟𝑟
regression to estimate the variables (∆𝐺𝐺𝐺𝐺𝐺𝐺 , ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑡𝑡 , and ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 ).

Our focus is how aggregate-level profitability drivers affect subsequent market return.

Konchitchki and Patatoukas (2014b) report that aggregate-level fundamentals affect

subsequent 3-month market return only through GDP growth. If their results are observed
𝑝𝑝
outside the US market, ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑡𝑡+1 should have significantly positive coefficients. In addition,

𝑟𝑟𝑟𝑟𝑟𝑟 𝑟𝑟𝑟𝑟𝑟𝑟
∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑡𝑡 and ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 should not have significant coefficients.

5.2. Results
(Insert Table 7 around here)

Table 7 shows the result on the relationship between subsequent market return and the

GDP growth predicted by aggregate-level profitability drivers. In the full sample test, we
𝑝𝑝 𝑟𝑟𝑟𝑟𝑟𝑟
cannot observe significantly positive coefficients on ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑡𝑡+1 . Instead, ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑡𝑡 has

significantly positive coefficients. These findings are not consistent with Konchitchki and

Patatoukas (2014b).

(7) and (8) becomes less than 10, so that multicollinearity does not bias the results shown in Table

7.

Electronic copy available at: https://ssrn.com/abstract=3461082

 To establish why our results differ from Konchitchki and Patatoukas (2014b), we

focus on the difference in the sample countries. Konchitchki and Patatoukas (2014b) use data

of the US, a representative developed country. By contrast, our sample contains not only

developed countries (Canada, Denmark, Finland, Germany, Hong Kong, Italy, Japan,

Singapore, Spain, Sweden, and the US) but also emerging countries (Argentina, Brazil,

Greece, Mexico, Peru, the Philippines, South Korea, Taiwan, Thailand, and Turkey) based on

the MSCI website definition. 8 This difference in the sample coverage could affect our results.

In fact, many non-US accounting studies report different results from US research

(e.g., Gabrielsen, Gramlich, and Plenborg 2002; Rahman, Yammeesri, and Perera 2010;

Gajewski and Quéré 2013). In addition, the effects of aggregate earnings could differ between

developed and emerging markets. The information environment is less sophisticated in

emerging countries (Lopes and de Alencar 2010). Since less information is available in

emerging countries, the relative impacts of aggregate earnings could be stronger in emerging

markets. Meanwhile, developed countries have more sophisticated markets, which reflect

aggregate-level profitability timeously. We split our sample into developed and emerging

countries and run the same regression with Equations (7) and (8) to investigate whether the

results change depending on the development level of countries. 9

8
https://www.msci.com/market-classification (accessed on August 10, 2020).
9
We do not create a dummy variable for whether the country is regarded as a developed country, since

we already have another dummy variable (𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 ) and its interaction effects in the model. If we

were to adopt a developed country dummy, there would be triple interaction terms in our model,

which would cause a multicollinearity concern and create difficulties in the interpretation of the

results. Thus, we split our sample in this additional analysis.

Electronic copy available at: https://ssrn.com/abstract=3461082

 The results using developed countries and emerging countries are displayed in the

middle two columns and right two columns in Table 7, respectively. For developed countries,
𝑝𝑝 𝑟𝑟𝑟𝑟𝑟𝑟
we observe significantly positive coefficients on ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑡𝑡+1 and the coefficients on ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑡𝑡

𝑟𝑟𝑟𝑟𝑟𝑟
and ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 are not significant. These results are consistent with the US results of
𝑝𝑝
Konchitchki and Patatoukas (2014b). Meanwhile, the coefficients on ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑡𝑡+1 are not

𝑟𝑟𝑟𝑟𝑟𝑟
significant in emerging countries. The coefficients of ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 are significant instead. These

results suggest that the channel of aggregate-level profitability drivers on market returns could

differ depending on the countries’ development level. One of the reasons could be related to

market efficiency. Developed countries have more efficient stock markets, and thus, the

predicted future economic growth by the aggregate-level profitability drivers could be

reflected timeously in market return.

6. Robustness Checks

We conduct four robustness checks (untabulated). First, we confirm that using nominal GDP

instead of real GDP does not largely affect our results. In addition, we find that the adjusted R

squares of the regression using nominal GDP are generally higher, since neither accounting

figures nor stock prices are adjusted by the movement of prices.

Second, we check whether the significance of the coefficients is stable, using other

standard errors. In the abovementioned tests, we use heteroskedasticity-consistent standard

errors proposed by White (1980). We confirm that the significance does not change that much

when using heteroskedasticity- and autocorrelation-consistent standard errors proposed by

Newey and West (1987) with lag length varying from one to four.

Third, we investigate whether the usefulness of aggregate-level profitability drivers is

limited to developed or emerging countries. We run the regressions with Equations (4) and

Electronic copy available at: https://ssrn.com/abstract=3461082

(5), using observations of either developed countries or emerging countries. We confirm that

in both sub-samples, the usefulness of aggregate-level profitability drivers is observed.

Fourth, we test whether our results are robust to the definition of the crisis periods. In

the abovementioned tests, we define 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 based on the three crises: the burst of the dot-

com bubble in 2001, the global financial crisis from 2007 to 2009, and the COVID-19

recession in 2020. We examine whether these crises have similar effects by using three

additional crisis dummy variables (𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡01, 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡07−09, and 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡20) instead of 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 .

𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡01 is an additional crisis dummy that takes the value one when the observation is

during the dot-com bubble. We define 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑠𝑠𝑡𝑡07−09 and 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡20 as additional dummy

variables for the global financial crisis and the COVID-19 recession, respectively. We run the

same regressions replacing 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 with each of these additional dummies. In these

regressions, we exclude country/quarter observations during the other crises to reduce their

effects.

The results with 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡07−09 are all similar to those with 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 at first. Using

𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡20 yields two slightly different results. When forecasting GDP growth three quarters

ahead with Equations (4) and (5), the goodness of fit is not significantly different between

these two models. However, the adjusted R square increases (by 0.2%) after including

aggregate-level profitability drivers, like the main results. In addition, when we engage in the
𝑟𝑟𝑟𝑟𝑟𝑟
additional tests written in Section 5 (Equation 8), the coefficient of ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 is significant in

the sub-sample of developed countries. The VIF of this regression is higher than 10, and thus,

this result should be caused by multicollinearity. Overall, although the number of

observations about the ongoing COVID-19 recession is very limited due to data availability,

differences between the results using 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡20 and 𝐶𝐶𝐶𝐶𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑡𝑡 are trivial. Our empirical

implications can be applied to the ongoing COVID-19 recession.

Electronic copy available at: https://ssrn.com/abstract=3461082

 Replacing 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 with 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡01 causes several differences. Although adjusted R

squares increase after containing aggregate-level profitability drivers, some of them are not
𝑝𝑝
significant based on chi squares. In the regression with Equations (7) and (8), ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑡𝑡+1 has

significantly positive coefficients in both the full-sample tests and the sub-sample tests.
𝑟𝑟𝑟𝑟𝑟𝑟
However, the coefficients of ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 are significantly negative in the developed country

sub-sample. The reason that 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡01 causes these differences should be the impacts of the

2001 crisis on the real economy. The burst of the dot-com bubble mainly affects stock market

and its impacts on the real economy are limited, compared to the global financial crisis and

the COVID-19 recession. When we investigate the difference in the average GDP growth

between each of the crisis periods and non-crisis periods through unequal variance t-tests,

average GDP growth is observed to be significantly lower during the global financial crisis

and the COVID-19 recession than in non-crisis periods at the 1% level. However, the

difference in average GDP growth is not significant between the 2001 crisis periods and non-

crisis periods. Therefore, we probably should treat only a crisis that severely damages the real

economy as an abnormal condition in constructing the forecast model of GDP growth.

7. Conclusion

Accounting information has the potential to support the work of economists. Macro-

accounting research tackles this topic. Prior studies in this field report that aggregate earnings

are incrementally useful for forecasting future GDP growth. However, prior studies do not

consider the effects of the global financial crisis on the informativeness of aggregate earnings.

In addition, since prior studies investigate the informativeness of aggregate earnings using

only US data, we cannot conclude that aggregate earnings are useful for other countries

without checking the suitability for them. To examine these points, in this study, we use

global data and investigate whether considering crisis effects improves the forecast model of

Electronic copy available at: https://ssrn.com/abstract=3461082

GDP growth with aggregate earnings worldwide.

The results yield three main pieces of evidence. First, aggregate-level profitability

drivers possess incremental information on future GDP growth in both developed countries

and emerging countries. Second, we show that considering crisis effects improves the forecast

model of future GDP growth with aggregate-level profitability drivers. Third, we suggest that

the route of the impacts of aggregate-level profitability drivers on market return differs

between developed and emerging countries. In developed countries, aggregate-level

profitability drivers affect future GDP growth through the prediction of future GDP growth,

which is consistent with prior research (Konchitchki and Patatoukas 2014b). However, this

impact route is not observed in emerging countries.

We contribute to the macro-accounting research in three ways. First, our results

expand the external validity of US research by using global data of 21 countries. To the best

of our knowledge, our study is the first to show the usefulness of aggregate earnings in

emerging countries. Our findings suggest that economists of emerging countries could

incorporate aggregate earnings into their GDP forecasts. Second, we improve the forecast

model of future GDP growth with aggregate-level profitability drivers by considering the

crisis effects. We extend the research of Konchitchki and Patatoukas (2014a, 2014b) by

proposing the better use of aggregate earnings information for GDP forecasting. The world

economy is currently suffering from the COVID-19 recession. Since we suggest that

regarding this ongoing recession as a crisis period raises the forecast accuracy of the model,

our evidence may make a timely contribution to current macroeconomic analysis. Third, we

find that the route of the impacts of aggregate-level profitability drivers on investors’ activity

differs between developed and emerging countries. Konchitchki and Patatoukas (2014b)

report that investors interpret future economic growth from aggregate-level profitability

Electronic copy available at: https://ssrn.com/abstract=3461082

drivers and make investment decisions. Our findings suggest that this activity can be observed

only in developed countries, not emerging countries.

However, the current study has the following limitation. We do not empirically

investigate the mechanisms underlying the links between aggregate profitability drivers and

future GDP growth. Recently, Shivakumar and Urcan (2017) propose two explanations for the

relationship between aggregate earnings growth and future inflation: firms’ investment

channel and consumers’ consumption channel. Future research could investigate the

mechanisms through which aggregate earnings growth causally affect future GDP growth

following Shivakumar and Urcan (2017). Such research would broaden our knowledge of the

link between accounting and the macroeconomy.

Nonetheless, we provide evidence that accounting earnings information is of interest

to macroeconomists and macro policy setters. Accounting standard-setting bodies, such as the

International Accounting Standards Board and the Financial Accounting Standards Board,

have traditionally overlooked these groups as users. Our findings explicitly extend the

potential users of accounting information. Accounting can contribute to macroeconomic

discussion.

Electronic copy available at: https://ssrn.com/abstract=3461082

References

Arthur, N., Q. Tang, and Z. S. Lin. 2015. ‘Corporate Accruals Quality during the 2008–2010
Global Financial Crisis.’ Journal of International Accounting, Auditing and Taxation
25: 1–15. doi: 10.1016/j.intaccaudtax.2015.10.004
Ball, R., and P. Brown. 1968, ‘An Empirical Evaluation of Accounting Income Numbers.’
Journal of Accounting Research 6 (2): 159–178. doi: 10.2307/2490232
Ball, R., and P. Brown. 2019. ‘Ball and Brown (1968) after Fifty Years.’ Pacific-Basin
Finance Journal 53: 410–431. doi: 10.1016/j.pacfin.2018.12.008
Ball, R., and G. Sadka. 2015. ‘Aggregate Earnings and Why They Matter.’ Journal of
Accounting Literature 34: 39–57. doi: 10.1016/j.acclit.2015.01.001
Ball, R., G. Sadka, and R. Sadka. 2009. ‘Aggregate Earnings and Asset Prices.’ Journal of
Accounting Research 47 (5): 1097–1133. doi: 10.1111/j.1475-679X.2009.00351.x
Beaver, W. H. 1968. ‘The Information Content of Annual Earnings Announcements.’ Journal
of Accounting Research 6: 67–92. doi: 10.2307/2490070
Courteau, L., J. L. Kao, T. O’Keefe, and G. D. Richardson. 2006. ‘Relative Accuracy and
Predictive Ability of Direct Valuation Methods, Price to Aggregate Earnings Method
and a Hybrid Approach.’ Accounting and Finance 46: 553–575. doi: 10.1111/j.1467-
629X.2006.00183.x
Doukakis, L., D. C. Ghicas, G. Siougle, and T. Sougiannis. 2020. ‘The Informativeness of
Micro and Macro Information During Economic Crisis and Non-Crisis Periods:
Evidence from Europe.’ European Accounting Review 29 (3): 467–492. doi:
10.1080/09638180.2019.1642221
Filip, A., and B. Raffournier. 2014. ‘Financial Crisis and Earnings Management: The
European Evidence.’ The International Journal of Accounting 49 (4): 455–478. doi:
10.1016/j.intacc.2014.10.004
Gabrielsen, G., J. D. Gramlich, and T. Plenborg. 2002. ‘Managerial Ownership, Information
Content of Earnings, and Discretionary Accruals in a Non-US Setting.’ Journal of
Business Finance and Accounting 29 (7–8): 967–988. doi: /10.1111/1468-5957.00457
Gajewski, J. F., and B. P. Quéré. 2013. ‘A Comparison of the Effects of Earnings Disclosures
on Information Asymmetry: Evidence from France and the US.’ The International
Journal of Accounting 48(1): 1–25. doi: 10.1016/j.intacc.2013.01.004

Electronic copy available at: https://ssrn.com/abstract=3461082

Gallo, L. A., R. N. Hann, and C. Li. 2016. ‘Aggregate Earnings Surprises, Monetary Policy,
and Stock Returns.’ Journal of Accounting and Economics 62 (1): 103–120. doi:
10.1016/j.jacceco.2016.04.003
Kang, W. 2019. ‘The Asymmetric Responses of Aggregate Earnings and Stock Returns to Oil
Shocks and Policy Uncertainty.’ Asia-Pacific Journal of Accounting & Economics 1–
13. doi: /10.1080/16081625.2019.1654395
Konchitchki, Y. 2016. ‘Accounting Valuation and Cost of Capital Dynamics: Theoretical and
Empirical Macroeconomic Aspects. Discussion of Callen.’ Abacus 52 (1): 26–34. doi:
/10.1111/abac.12071
Konchitchki, Y., and P. N. Patatoukas. 2014a. ‘Accounting Earnings and Gross Domestic
Product.’ Journal of Accounting and Economics 57 (1): 76–88. doi:
10.1016/j.jacceco.2013.10.001
Konchitchki, Y., and P. N. Patatoukas. 2014b. ‘Taking the Pulse of the Real Economy Using
Financial Statement Analysis: Implications for Macro Forecasting and Stock
Valuation.’ The Accounting Review 89 (2): 669–694. doi: 10.2308/accr-50632
Kothari, S. P. 2001. ‘Capital Markets Research in Accounting.’ Journal of Accounting and
Economics 31 (1): 105–231. doi: 10.1016/S0165-4101(01)00030-1
Kothari, S. P., J. Lewellen, and J. B. Warner. 2006. ‘Stock Returns, Aggregate Earnings
Surprises, and Behavioral Finance.’ Journal of Financial Economics 79 (3): 537–568.
doi: 10.1016/j.jfineco.2004.06.016
Loh, R. K., and R. M. Stulz. 2018. ‘Is sell-side research more valuable in bad times?’ The
Journal of Finance 73 (3): 959–1013. doi: 10.1111/jofi.12611
Lopes, A. B., and R. C. de Alencar. 2010. ‘Disclosure and Cost of Equity Capital in Emerging
Markets: The Brazilian Case.’ The International Journal of Accounting 45 (4): 443–
464. doi: 10.1016/j.intacc.2010.09.003
Nallareddy, S., and M. Ogneva. 2017, ‘Predicting Restatements in Macroeconomic Indicators
Using Accounting Information.’ The Accounting Review 92 (2): 151–182. doi:
10.2308/accr-51528
Newey, W. K., and K. D. West. 1987. ‘A Simple, Positive Semi-definite, Heteroskedasticity
and Autocorrelation Consistent Covariance Matrix.’ Econometrica 55 (3): 703–708.
doi: 10.2307/1913610
Ohlson, J. A. 1995. ‘Earnings, Book Values, and Dividends in Equity Valuation.’
Contemporary Accounting Research 11 (2): 661–687. doi: /10.1111/j.1911-
3846.1995.tb00461.x

Electronic copy available at: https://ssrn.com/abstract=3461082

Penman, S. H. 2012. Financial Statement Analysis and Security Valuation. 5th ed. McGraw-
Hill Education.
Rahman, A., J. Yammeesri, and H. Perera. 2010. ‘Financial Reporting Quality in International
Settings: A Comparative Study of the USA, Japan, Thailand, France and Germany.’
The International Journal of Accounting 45 (1): 1–34. doi:
10.1016/j.intacc.2010.01.001
Sadka, G., and R. Sadka. 2009. ‘Predictability and the Earnings–Returns Relation.’ Journal of
Financial Economics 94 (1): 87–106. doi: 10.1016/j.jfineco.2008.10.005
Shivakumar, L. 2007. ‘Aggregate Earnings, Stock Market Returns and Macroeconomic
Activity: A Discussion of ‘Does Earnings Guidance Affect Market Returns? The
Nature and Information Content of Aggregate Earnings Guidance.’ Journal of
Accounting and Economics 44 (1): 64–73. doi: 10.1016/j.jacceco.2006.12.001
Shivakumar, L., and O. Urcan. 2017. ‘Why Does Aggregate Earnings Growth Reflect
Information about Future Inflation?’ The Accounting Review 92 (6): 247–276. doi:
10.2308/accr-51714
Soliman, M. T. 2008. ‘The Use of DuPont Analysis by Market Participants.’ The Accounting
Review 83 (3): 823–853. doi: 10.2308/accr.2008.83.3.823
Veronesi, P. 1999. ‘Stock Market Overreactions to Bad News in Good Times: A Rational
Expectations Equilibrium Model.’ Review of Financial Studies 12: 975–1007. doi:
10.1093/rfs/12.5.975
White, H. 1980. ‘A heteroskedasticity-consistent covariance matrix estimator and a direct test
for heteroskedasticity.’ Econometrica 48 (4): 817–838. doi: 10.2307/1912934
Williams, C. D. 2015. ‘Asymmetric Responses to Earnings News: A Case for Ambiguity.’
The Accounting Review 90 (2): 785–817. doi: 10.2308/accr-50866

Electronic copy available at: https://ssrn.com/abstract=3461082

Tables

Table 1. Descriptive statistics and correlation matrix

Panel A Descriptive statistics

Mean SD Minimum 25% Median 75% Maximum N

∆𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 0.000 0.010 -0.083 -0.005 0.000 0.004 0.062 1,239
∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 0.000 0.028 -0.234 -0.012 0.000 0.012 0.219 1,239
∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑞𝑞 -0.003 0.056 -0.663 -0.022 -0.002 0.016 0.677 1,239
∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞 0.025 0.033 -0.155 0.011 0.026 0.043 0.168 1,239
𝑀𝑀𝑀𝑀𝑐𝑐,𝑞𝑞 0.061 0.251 -1.064 -0.062 0.085 0.204 1.064 1,239

Panel B Correlation matrix

∆𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑞𝑞 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞 𝑀𝑀𝑀𝑀𝑐𝑐,𝑞𝑞
∆𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 0.858 0.574 0.312 0.425
∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 0.889 0.203 0.195 0.376
∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑞𝑞 0.404 0.095 0.372 0.302
∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞 0.350 0.239 0.274 0.337
𝑀𝑀𝑀𝑀𝑐𝑐,𝑞𝑞 0.471 0.424 0.182 0.422
Notes: In Panel A, SD is the standard deviation of the variables. N is the number of observations. In Panel B,
Pearson (Spearman) correlations are below (above) the diagonal.

Electronic copy available at: https://ssrn.com/abstract=3461082

Table 2. Descriptive statistics for each country
N Start End ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑞𝑞 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞 𝑀𝑀𝑀𝑀𝑐𝑐,𝑞𝑞

Argentina 29 2006q4 2017q4 -0.002 -0.006 -0.001 0.009 0.197

Brazil 81 1998q4 2020q1 0.000 0.001 0.002 0.023 0.116

Canada 88 1997q1 2020q1 0.000 0.001 -0.005 0.023 0.043

Denmark 46 2007q2 2020q1 0.000 0.002 -0.003 0.012 0.084

Finland 71 2001q2 2020q1 0.000 0.001 -0.003 0.013 0.011

Germany 78 1999q3 2020q1 -0.001 0.000 -0.014 0.013 0.032

Greece 46 2002q2 2016q4 -0.002 -0.005 -0.009 -0.015 -0.113

Hong Kong 22 2006q3 2016q4 -0.002 -0.004 0.028 0.027 -0.013

Italy 25 2001q2 2010q1 -0.002 -0.001 -0.014 0.004 -0.018

Japan 69 2001q4 2020q1 0.000 0.002 -0.007 0.009 0.043

Mexico 61 2003q2 2020q1 -0.001 -0.002 -0.002 0.021 0.098

Peru 73 2000q4 2020q1 0.000 -0.002 0.003 0.050 0.153

Philippines 37 2002q4 2020q1 0.001 0.004 0.005 0.058 0.128

Singapore 64 2003q2 2020q2 0.000 -0.001 -0.003 0.050 0.037

South Korea 71 2001q2 2020q1 0.000 0.000 -0.003 0.037 0.067

Spain 29 2009q3 2020q1 0.000 0.000 0.001 0.016 0.012

Sweden 79 1999q2 2020q1 0.000 0.002 -0.010 0.022 0.040

Taiwan 47 2007q2 2020q1 -0.002 -0.008 0.005 0.027 0.019

Thailand 73 2000q4 2020q1 0.000 -0.003 0.001 0.039 0.095

Turkey 55 2004q2 2020q1 -0.001 0.001 -0.015 0.046 0.094

US 95 1995q2 2020q1 0.000 0.001 -0.005 0.024 0.071

Notes: The table reports the number of country/quarter observations (N), the beginning and end of the quarter

with the data to calculate one of the variables (Start, and End), and the mean of the listed variables for each

country.

Electronic copy available at: https://ssrn.com/abstract=3461082

Table 3. Effects of containing the changes in RNOA on forecast accuracy
Equation (3) Equation (1)
∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+1 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+2 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+3 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+4 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+1 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+2 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+3 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+4

𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 0.005*** 0.012*** 0.018*** 0.025*** 0.005*** 0.011*** 0.018*** 0.024***

(5.613) (9.778) (13.225) (16.419) (5.562) (8.867) (12.077) (15.168)

∆𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 0.078 -0.071 -0.203** -0.263***

(1.248) (-0.837) (-2.020) (-2.621)

∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞 0.683*** 0.372*** 0.099** -0.135*** 0.677*** 0.377*** 0.115*** -0.114**

(26.263) (10.730) (2.439) (-2.993) (25.032) (10.370) (2.755) (-2.464)

𝑀𝑀𝑀𝑀𝑐𝑐,𝑞𝑞 0.036*** 0.050*** 0.049*** 0.036*** 0.035*** 0.051*** 0.052*** 0.040***

(12.487) (13.565) (12.591) (9.125) (11.795) (12.758) (11.707) (9.150)

𝐴𝐴𝑅𝑅2 0.759 0.547 0.378 0.285 0.759 0.547 0.380 0.289

𝐿𝐿𝐿𝐿 𝑐𝑐ℎ𝑖𝑖 2 2.073 0.937 5.570** 8.306***

𝑁𝑁 1,239 1,223 1,206 1,189 1,239 1,223 1,206 1,189

Notes: The table reports the results from country fixed-effect regressions of Equations (1) and (3). We report

t-statistics using robust standard errors in parentheses proposed by White (1980). ***, **, and * denote

statistical significance at the 1%, 5%, and 10% level, respectively, using two-tailed tests. 𝐴𝐴𝐴𝐴2 is the adjusted

R square of the regressions. 𝐿𝐿𝐿𝐿 𝑐𝑐ℎ𝑖𝑖 2 represents chi-squared statistics in the likelihood-ratio test between two

models of the same dependent variable. 𝑁𝑁 is the number of observations.

Electronic copy available at: https://ssrn.com/abstract=3461082

Table 4. Effects of containing the changes in ROS and ATO on forecast accuracy
Equation (3) Equation (2)
∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+1 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+2 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+3 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+4 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+1 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+2 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+3 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+4

𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 0.005*** 0.012*** 0.018*** 0.025*** 0.005*** 0.011*** 0.017*** 0.024***

(5.613) (9.778) (13.225) (16.419) (5.152) (8.488) (11.633) (14.890)

∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 0.052** 0.029 0.006 -0.027

(2.496) (1.049) (0.183) (-0.793)

∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑞𝑞 -0.030*** -0.058*** -0.078*** -0.059***

(-2.898) (-3.245) (-3.294) (-3.178)

∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞 0.683*** 0.372*** 0.099** -0.135*** 0.693*** 0.397*** 0.137*** -0.103**

(26.263) (10.730) (2.439) (-2.993) (25.850) (11.079) (3.284) (-2.200)

𝑀𝑀𝑀𝑀𝑐𝑐,𝑞𝑞 0.036*** 0.050*** 0.049*** 0.036*** 0.034*** 0.050*** 0.050*** 0.038***

(12.487) (13.565) (12.591) (9.125) (11.943) (12.837) (11.602) (8.736)

𝐴𝐴𝑅𝑅2 0.759 0.547 0.378 0.285 0.762 0.555 0.392 0.293

𝐿𝐿𝐿𝐿 𝑐𝑐ℎ𝑖𝑖 2 18.741*** 23.789*** 30.404*** 16.192***

𝑁𝑁 1,239 1,223 1,206 1,189 1,239 1,223 1,206 1,189

Notes: The table reports the results from country fixed-effect regressions of Equations (2) and (3). We report

t-statistics using robust standard errors in parentheses proposed by White (1980). ***, **, and * denote

statistical significance at the 1%, 5%, and 10% level, respectively, using two-tailed tests. 𝐴𝐴𝑅𝑅2 is the adjusted

R square of the regressions. 𝐿𝐿𝐿𝐿 𝑐𝑐ℎ𝑖𝑖 2 represents chi-squared statistics in the likelihood-ratio test between two

models of the same dependent variable. 𝑁𝑁 is the number of observations.

Electronic copy available at: https://ssrn.com/abstract=3461082

Table 5. Results of considering the effects of crisis on forecast accuracy
Equation (2) Equation (4)
∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+1 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+2 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+3 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+4 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+1 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+2 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+3 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+4

𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 0.005*** 0.011*** 0.017*** 0.024*** 0.006*** 0.010*** 0.016*** 0.023***

(5.152) (8.488) (11.633) (14.890) (6.105) (8.058) (10.845) (12.771)

∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 0.052** 0.029 0.006 -0.027 -0.051** -0.057** -0.039 -0.009

(2.496) (1.049) (0.183) (-0.793) (-2.558) (-2.232) (-1.384) (-0.290)

∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑞𝑞 -0.030*** -0.058*** -0.078*** -0.059*** -0.009 -0.011 -0.020 -0.038**

(-2.898) (-3.245) (-3.294) (-3.178) (-1.132) (-0.932) (-1.355) (-2.195)

∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞 0.693*** 0.397*** 0.137*** -0.103** 0.703*** 0.516*** 0.286*** 0.045

(25.850) (11.079) (3.284) (-2.200) (21.289) (11.778) (5.808) (0.772)

𝑀𝑀𝑀𝑀𝑐𝑐,𝑞𝑞 0.034*** 0.050*** 0.050*** 0.038*** 0.020*** 0.024*** 0.027*** 0.023***

(11.943) (12.837) (11.602) (8.736) (6.306) (6.739) (6.270) (4.755)

𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 0.000 0.003 0.003 0.002

(-0.130) (1.173) (1.005) (0.683)

𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 0.193*** 0.148*** 0.076 -0.045
∗ ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑡𝑡
(4.307) (2.750) (1.228) (-0.649)
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 -0.040*** -0.079** -0.098** -0.022
∗ ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑞𝑞
(-2.776) (-2.089) (-2.097) (-0.589)
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 -0.050 -0.253*** -0.278*** -0.272***
∗ ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞
(-1.057) (-3.921) (-3.821) (-3.190)
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 0.026*** 0.050*** 0.043*** 0.032***
∗ 𝑀𝑀𝑀𝑀𝑐𝑐,𝑞𝑞
(4.078) (6.240) (5.051) (3.585)

𝐴𝐴𝑅𝑅2 0.762 0.555 0.392 0.293 0.779 0.593 0.425 0.310

𝐿𝐿𝐿𝐿 𝑐𝑐ℎ𝑖𝑖 2 97.204*** 116.23*** 72.017*** 34.045***

𝑁𝑁 1,239 1,223 1,206 1,189 1,239 1,223 1,206 1,189

Notes: The table reports the results from country fixed-effect regressions of Equations (2) and (4). We report

t-statistics using robust standard errors in parentheses proposed by White (1980). ***, **, and * denote

statistical significance at the 1%, 5%, and 10% level, respectively, using two-tailed tests. 𝐴𝐴𝐴𝐴2 is the adjusted

R square of the regressions. 𝐿𝐿𝐿𝐿 𝑐𝑐ℎ𝑖𝑖 2 represents chi-squared statistics in the likelihood-ratio test between two

models of the same dependent variable. 𝑁𝑁 is the number of observations.

Electronic copy available at: https://ssrn.com/abstract=3461082

Table 6. Results of the effects of crisis and changes in ROS and ATO on forecast accuracy
Equation (5) Equation (4)
∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+1 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+2 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+3 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+4 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+1 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+2 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+3 ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞+4

𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 0.006*** 0.011*** 0.017*** 0.023*** 0.006*** 0.010*** 0.016*** 0.023***

(6.519) (8.638) (11.465) (13.262) (6.105) (8.058) (10.845) (12.771)

∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑞𝑞 -0.051** -0.057** -0.039 -0.009

(-2.558) (-2.232) (-1.384) (-0.290)

∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑞𝑞 -0.009 -0.011 -0.020 -0.038**

(-1.132) (-0.932) (-1.355) (-2.195)

∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞 0.699*** 0.509*** 0.274*** 0.032 0.703*** 0.516*** 0.286*** 0.045

(21.099) (11.743) (5.614) (0.557) (21.289) (11.778) (5.808) (0.772)

𝑀𝑀𝑀𝑀𝑐𝑐,𝑞𝑞 0.018*** 0.022*** 0.026*** 0.022*** 0.020*** 0.024*** 0.027*** 0.023***

(5.910) (6.537) (6.182) (4.758) (6.306) (6.739) (6.270) (4.755)

𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 0.000 0.004* 0.004 0.003 0.000 0.003 0.003 0.002

(0.008) (1.653) (1.630) (0.919) (-0.130) (1.173) (1.005) (0.683)

𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 0.193*** 0.148*** 0.076 -0.045
∗ ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑡𝑡
(4.307) (2.750) (1.228) (-0.649)
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 -0.040*** -0.079** -0.098** -0.022
∗ ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑞𝑞
(-2.776) (-2.089) (-2.097) (-0.589)
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 -0.071 -0.299*** -0.341*** -0.302*** -0.050 -0.253*** -0.278*** -0.272***
∗ ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑞𝑞
(-1.568) (-4.819) (-4.878) (-3.687) (-1.057) (-3.921) (-3.821) (-3.190)
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 0.036*** 0.057*** 0.047*** 0.030*** 0.026*** 0.050*** 0.043*** 0.032***
∗ 𝑀𝑀𝑀𝑀𝑐𝑐,𝑞𝑞
(5.779) (7.937) (6.382) (3.849) (4.078) (6.240) (5.051) (3.585)

𝐴𝐴𝑅𝑅2 0.772 0.583 0.409 0.305 0.779 0.593 0.425 0.310

𝐿𝐿𝐿𝐿 𝑐𝑐ℎ𝑖𝑖 2 44.134*** 36.097*** 37.461*** 13.131**

𝑁𝑁 1,239 1,223 1,206 1,189 1,239 1,223 1,206 1,189

Notes: The table reports the results from country fixed-effect regressions of Equations (4) and (5). We report

t-statistics using robust standard errors in parentheses proposed by White (1980). ***, **, and * denote

statistical significance at the 1%, 5%, and 10% level, respectively, using two-tailed tests. 𝐴𝐴𝐴𝐴2 is the adjusted

R square of the regressions. 𝐿𝐿𝐿𝐿 𝑐𝑐ℎ𝑖𝑖 2 represents chi-squared statistics in the likelihood-ratio test between two

models of the same dependent variable. 𝑁𝑁 is the number of observations.

Electronic copy available at: https://ssrn.com/abstract=3461082

Table 7. Market returns and predicted contents of changes in aggregate-level profitability
drivers on subsequent GDP growth
Full Sample Developed Emerging
𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 0.018*** 0.018*** 0.018*** 0.014*** 0.013*** 0.021***
(5.038) (5.042) (5.038) (3.360) (3.019) (3.643)
𝑝𝑝
∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑡𝑡+1 0.482 0.763 1.721** 2.185** -0.441 0.299
(0.639) (0.997) (2.191) (2.406) (-0.375) (0.252)
𝑟𝑟𝑟𝑟𝑟𝑟
∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑡𝑡 0.281* 0.257 0.194
(1.772) (1.388) (0.850)
𝑟𝑟𝑟𝑟𝑟𝑟
∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 0.100 -0.040 0.332**
(1.535) (-0.574) (2.079)
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 -0.017** -0.011 -0.022** -0.019* -0.011 0.003
(-1.981) (-1.126) (-2.091) (-1.731) (-0.769) (0.181)
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 -1.275 -1.330 -1.585 -1.888* -0.911 -1.526
𝑝𝑝
∗ ∆𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐,𝑡𝑡+1
(-1.488) (-1.530) (-1.569) (-1.795) (-0.723) (-1.154)
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 -1.658** -1.220 -1.572*
𝑟𝑟𝑟𝑟𝑟𝑟
∗ ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑐𝑐,𝑡𝑡
(-2.571) (-1.403) (-1.683)
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 -0.436** -0.257 -0.891***
𝑟𝑟𝑟𝑟𝑟𝑟
∗ ∆𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡
(-2.422) (-1.131) (-3.208)
𝐴𝐴𝑅𝑅2 0.012 0.022 0.002 0.005 0.025 0.041
𝑁𝑁 1,236 1,236 663 663 573 573
Notes: The table reports the results from country fixed-effect regressions of Equations (7). The left two

columns are full sample results. The middle two columns are the results for developed countries. The right

two columns are the results for developed countries. We report t-statistics using robust standard errors in

parentheses proposed by White (1980). ***, **, and * denote statistical significance at the 1%, 5%, and 10%

level, respectively, using two-tailed tests. 𝐴𝐴𝐴𝐴2 is the adjusted R square of the regressions. 𝑁𝑁 is the number of

observations.

Electronic copy available at: https://ssrn.com/abstract=3461082