Page 1

Electronic copy available at: http://ssrn.com/abstract=1787237

The Asset Growth Effect:

Insights from International Equity Markets∗

Akiko Watanabe†

Yan Xu‡

Tong Yao§

Tong Yu¶

First draft: Nov. 2009

This version: Feb. 2011

∗We thank conference participants at the Financial Management Association 2010 Meeting, the 2010 China Interna-

tional Finance Conference, and the Nippon Finance Association 2010 Meeting. We also thank seminar participants at

the University of New South Wales and Lingnan University. The comments from Gil Aharoni (FMA discussant), Zhi

Da, Shingo Goto, Kewei Hou, Po-Hsuan Hsu, Andrew Karolyi, Laura Xiaolei Liu (CICF discussant), Ronald Masulis,

Michael Schill, Neal Stoughton, Gloria Tian, Toshifumi Tokunaga (NFA discussant), Masahrio Watanabe, Bohui Zhang,

Lu Zhang, and Yuzhao Zhang are greatly appreciated. All errors are our own.

†Department of Finance and Statistical Analysis, University of Alberta School of Business; Edmonton, Alberta, Canada

T6G 2R6; phone: +1 780 492 0385; fax: +1 780 492 3325; email: akiko.watanabe@ualbert.ca.

‡College of Business Administration, University of Rhode Island; 7 Lippitt Road, Kingston, RI 02881-0802, USA;

phone: +1 401 874 4190; fax: +1 401 874 4312; email: yan xu@mail.uri.edu.

§Department of Finance, Henry B. Tippie College of Business, University of Iowa, Iowa city, IA 52242, USA; phone:

+1 319 335 3924; fax: +1 319 335 3690; email: tongyao@uiowa.edu.

¶College of Business Administration, University of Rhode Island; 7 Lippitt Road, Kingston, RI 02881-0802, USA;

phone: +1 401 874 7415; fax: +1 401 874 4312; email: tongyu@uri.edu.

Page 2

Electronic copy available at: http://ssrn.com/abstract=1787237

The Asset Growth Effect:

Insights from International Stock Markets

Abstract

Stocks with higher asset growth rates experience lower future returns in 40 international equity

markets, consistent with the U.S. evidence documented by Cooper et al. (2008). This negative effect

of asset growth on stock return is stronger in developed markets and in markets where stocks are more

efficiently priced. For each country, we estimate a parsimonious model to quantify the cash flow channel

and the discount rate channel of the investment-return relationship proposed by the q-theory model

of Li, Livdan, and Zhang (2008). The estimated cash flow beta and discount rate beta successfully

explain the cross-country variation in the asset growth effect on future stock returns. In contrast,

country characteristics related to corporate governance and investor protection, and measures of limits

to arbitrage, do not explain such effect. This evidence suggests that the q-theory model does better

than mispricing-based hypotheses in explaining the asset growth effect internationally.

Page 3

Electronic copy available at: http://ssrn.com/abstract=1787237

1 Introduction

It has been documented that firms experiencing rapid growth by raising external financing and making

real investments subsequently have low stock returns, whereas firms experiencing contraction via dives-

ture, share repurchase, and debt retirement consequently enjoy high stock returns. Recently, Cooper,

Gulen, and Schill (2008) summarize the synergistic effect of firms’ investment and financing activities

by constructing a simple measure of total asset growth rates and studying its effect on future returns.

They show that in the United States, stocks with higher asset growth rates significantly underperform

stocks with lower asset growth rates and such cross-sectional return difference cannot be explained by

standard asset pricing models.

One of the most actively debated issues in current finance literature is whether the negative effect

of investment and financing on stock returns – as epitomized by the asset growth effect – is evi-

dence of market inefficiency or can be viewed as a rational asset pricing result. From the behavioral

camp, several mispricing-based explanations have been proposed. They include 1) over-investment and

empire-building tendency of corporate managers (e.g., Titman, Wei, and Xie, 2004), 2) capital structure

market timing when raising and retiring external financing (e.g., Baker and Wurgler, 2002), 3) earn-

ings management prior to financing activities or acquisitions (e.g., Teoh, Welch, Wong, 1998a; 1998b),

and 4) excessive extrapolation on past growth by investors when they value firms (e.g., Lakonishok,

Shleifer, and Vishny, 1994). From the rational asset pricing camp, the explanations center around

the association between investment and expected return, albeit with some variations. For example, in

Cochrane (1991, 1996) and Liu, Whited, and Zhang (2009), firms making large investments are likely

to be those with low discount rates. In Lyandres, Sun, and Zhang (2008) and Li, Livdan, and Zhang

(2009), higher investments are associated with lower expected returns via both decreasing return to

scale and the discount rate effect (that lower discount-rate firms make larger investments). Addition-

ally, in Berk, Green and Naik (1999) and Carlson, Fisher, and Giammarino (2004), firms have reduced

risk and expected return after their growth options are exercised through real investments.1

In this paper, we study the existence and potential causes of the asset growth effect in international

1A number of empirical papers, such as Agrawal, Jaffee, and Mandelker (1992), Ikenberry, Lakonishok, and Vermaelen

(1995), Loughran and Ritter (1995), Rau and Vermaelen (1998), and Richardson and Sloan (2003), have subscribed to

one or multiple mispricing-based explanations. On the other hand, a few other studies such as Anderson and Garcia-

Feijoo (2006), Fama and French (2006), and Xing (2008), have provided empirical evidence consistent with the rational

investment effect.

1

Page 4

equity markets. We document the existence of the negative relation between asset growth and stock

returns in a large number of stock markets outside the U.S., as well as a large variation of this relation

across countries. We further show that the two rational investment effects based on the q-theory

model (i.e., the cash flow channel and the discount rate channel) successfully explain the cross-country

variation in the way asset growth affects stock returns. Importantly, the results are robust to the control

of various proxies for corporate governance and investor protection or proxies for limits to arbitrage

(which have little power to explain the cross-country difference in the asset growth effect).

Specifically, using the Datastream-WorldScope data for the period from 1982 to 2006, we find that,

on aggregate, the asset growth effect exists in 40 international equity markets outside the U.S. When

stocks are pooled together across all the countries and sorted on asset growth into decile portfolios,

the return spread between the top and bottom decile portfolios is -2.42% per year for equal-weighted

portfolios and -2.46% for value-weighted portfolios, both significantly negative. Results based on Fama-

MacBeth regressions further show that the asset growth effect is not explained away by other return-

predictive stock characteristics such as size, book-to-market, and momentum. Across countries, the

magnitude of the asset growth effect varies largely. For instance, the return spread between the top and

bottom asset growth decile portfolios ranges from -9.66% to 7.11% across countries. They are negative

in 27 countries but positive in the remaining 13 countries. Notably, the asset growth effect tends to be

stronger in developed markets and in markets where stocks are more efficiently priced.

The q-theory model suggests two potential channels through which investments affect investment

returns as well as expected stock returns – a “cash flow” channel and a “discount rate” channel.

We investigate how well these channels account for the cross-country variation of the asset growth

effect. Since asset growth is a comprehensive measure of firm investment and disinvestment (Cooper,

Gulen, and Schill, 2008), and the value of corporate assets is related to the underlying quantity of

capital (Abel, Mankiw, Summers, and Zeckhauser, 1989), we use return on total assets as a proxy

for investment return, following the approach of Larrain and Yogo (2008). We further follow their

approach to derive a decomposition of the observed relation between investment and investment return,

based on loglinearized present value relation and budget constraints. This decomposition enables us

to quantify three components of the investment-investment return relation predicted by “structural”

q-theory models (e.g., Li, Livdan, and Zhang, 2009) – one driven by innovations in cash flows (referred

to as the “ cash flow beta”), one driven by innovations in discount rate (the “discount rate beta”), and

2

Page 5

a third term due to the interaction between cash flow news and discount rate news (the “interaction

beta”). For a given country, the decomposition results in a system of VARs across a large panel of

stocks, which is estimated by dynamic panel GMM procedure.

Empirically, we find that the (negative) effect of asset growth on stock return is stronger in countries

where the (negative) relation between investment and investment return is stronger. Relative to the

discount rate beta and the interaction beta, the cash flow beta has a much larger magnitude of variation

across countries. But generally, both the cash flow beta and the discount rate beta have significant

power to explain the relation between asset growth and stock returns across countries.2On the other

hand, the explanatory power of the interaction beta is insignificant.

If the betas capture the q-theory model based rational investment effects and can explain the asset

growth-stock return relation, the explanatory power of the betas should be stronger among countries

where stocks are more efficiently priced. We test this prediction using three proxies for market efficiency.

The first is stock return synchronicity, following the argument of Morck, Yeung, and Yu (2000) that

this measure is negatively related to the extent to which firm-specific information is incorporated into

individual stock prices. The second is the ratio of a country’s stock market capitalization to its GDP,

an indicator of stock market development. The third is the ratio of a country’s domestic credit volume

over its stock market capitalization, following the argument that the dominance of banking financing

increases financial market opacity (e.g., Morck and Nakamura 1999; La Porta, Lopez-de-Silanes, and

Shleifer, 2002). Our findings are consistent with the prediction tested – the explanatory power of

the cash flow beta and the discount rate beta is stronger among countries with lower stock return

synchronicity, higher market-cap to GDP ratio, and lower bank credit to market-cap ratio.

We also evaluate the explanatory power of the betas against that of the variables motivated by

mispricing-based hypotheses. Under the mispricing-based hypotheses, the asset growth effect on stock

returns should be weaker among markets with stronger investor protection and corporate governance,

and stronger when there is a higher degree of limit to arbitrage. We find that the explanatory power

of the betas remains significant after controlling for proxies for governance and investor protection,

including a country’s legal origin, the anti-director rights index, the anti-self dealing index, the ac-

2We find that only the cash flow beta, but not the discount rate beta, has the explanatory power for the value-weighted

spread portfolio returns across countries, suggesting that the cash flow channel is more important for the asset growth

effect among larger firms.

3

Page 6

counting quality index, and proxies related to the limits to arbitrage such as the short-sale constraints

and stock trading liquidity. Indeed, these control variables generally do not exhibit significant power

to explain the asset growth effect across countries.

Overall, we find that the q-theory model does better than mispricing based hypotheses in explaining

the asset growth effect across international equity markets. To our knowledge, this is the first study

that explicitly examines the rational investment effects using international data. Our findings provide

a useful perspective to understand the debate on the relation between corporate investment activities

and stock returns. To differentiate the rational asset pricing hypotheses from the mispricing-based

hypotheses of the investment-return relation, recent studies have mainly focused on the U.S. data

and have examined the effect of investments on stock returns during subperiods or in subsamples of

stocks where the mispricing (or rational) effect is likely to be strong. For example, Wei, Xie, and

Titman (2004) find that the negative investment-return relation is stronger among firms with greater

managerial investment discretion (e.g., higher cash flows and lower debt ratios), and is only significant

during time periods when external corporate governance is weak (e.g., when hostile takeovers are not

prevalent). Cooper, Gulen, and Schill (2008) show that the asset growth effect on stock return is

weaker during sample periods when external corporate oversight becomes stronger, and is stronger

when investor sentiment (measured by past market return) is stronger. Further, both Lipson, Mortal,

and Schill (2009) and Lam and Wei (2009) report a stronger asset growth anomaly among stocks facing

more severe limits to arbitrage as measured by arbitrage risk, information cost, and liquidity. While

these studies favor mispricing-based interpretations, Li and Zhang (2009) point out that in a rational

investment model, the investment-return relation should also be stronger among firms facing higher

investment and financing frictions. They argue that a stronger asset growth effect when limits to

arbitrage are more severe cannot be automatically viewed as exclusive evidence of the mispricing-based

hypothesis, because proxies for limits to arbitrage may also be proxies for frictions to investments and

financing.

A paper closely related to ours is McLean, Pontiff, and Watanabe (2009), who analyze the negative

cross-sectional relation between net share issuance and stock returns in international markets, but they

do not directly examine the rational investment effects as a possible cause of their findings. Although

firms’ equity financing and asset growth are correlated phenomena, it is known that they do not

subsume each other in predicting returns (e.g., Cooper et al. 2008). In addition, the asset growth

4

Page 7

effect and the net share issuance effect vary in a different way with certain country characteristics. To

highlight, McLean et al. find that the share issuance effect is stronger in countries with greater issuance

activity, stronger investor protection and less earnings management. We do not find such patterns for

the asset growth effect.3

In a study contemporaneous to ours, Titman, Wei, and Xie (2010) also examine the asset growth

effect in international markets. Consistent with ours, they show that the asset growth effect exists

internationally, and is stronger in developed markets than in emerging markets. Different from ours,

they focus on the effect of external financing and show that the asset growth effect is stronger in

countries with easier access to external financing. While they explicitly consider an over-investment

based hypothesis, their finding can be potentially reconciled with ours in the following way. Arguably,

the effect of decreasing return to scale, which is measured by the cash flow beta, should be more visible

for firms with easy access to financing relative to firms facing financing constraints. In addition, easy

access to the capital markets is often associated with capital market development, hence stock price

efficiency.

The remainder of the paper is organized as follows. Section 2 describes the data and provides

evidence on the existence of the asset growth effect in international equity markets. Section 3 describes

the decomposition of the cash flow channel and the discount rate channel, and then empirically inves-

tigates the cross-country relation between the two betas and the asset growth effect. Finally, Section

4 concludes.

2 International Evidence on Asset Growth Effect

2.1Data

The data on stock market and accounting variables for international firms are obtained from Thomson-

Reuter Datastream and Worldscope. We select non-financial common stocks that are listed on each

country’s major stock exchange(s) from both active and defunct research files of Datastream in order

3Another related paper is Yao, Yu, Zhang, and Chen (2011). They document a relatively weak asset growth effect in

nine Asian markets, and attribute this to the relative homogeneity of asset growth and the reliance of bank financing for

growth in this region.

5

Page 8

to avoid the survivorship bias.4

We perform initial data screenings for basic coding errors via the

methods outlined in Ince and Porter (2006). To ensure meaningful cross-sectional sample size for each

country, we require that a country has at least a monthly average of 30 firms with valid observations

on asset growth, market capitalization, and monthly stock returns. These screenings leave us with 40

stock markets outside of the U.S. during a period between July of 1982 and June of 2006.5

For a comparison purpose, we also include the U.S. market in our analysis. The U.S. sample

consists of common stocks from non-financial industries (excluding firms with SIC codes between 6000

and 6999) listed on NYSE, AMEX, and NASDAQ. Data on the U.S. market are obtained from CRSP

and Compustat.

Table 1 provides the summary statistics for the sample. The starting date of inclusion into our

sample varies across countries, depending on each country’s data availability. The sample consists of

2,822,534 firm-month observations when the U.S. is included and 1,864,036 observations when the U.S.

is excluded. As expected, the U.S. represents the largest part of the overall sample, accounting for 34%

of the total firm-month observations and 42% of the total market capitalization on average. Japan is

the second largest, accounting for 13% of the total observations and 20% of the total market value.

While the remaining countries typically account for less than 5% of the total observations and market

value, our sample covers a variety of countries from different regions.

The main variable of interest is the firm-level proxy of investment, the asset growth rate (AG).

Following Cooper et al. (2008), a firm’s total asset growth rate observed at the end of June of year t is

given by the percentage change in total assets from the end of fiscal year t−2 to the end of fiscal year

t − 1 (fiscal year t is defined here as the fiscal year ending in calendar year t). The measure of total

assets in local currency is the Worldscope field 02999 for international firms and the Compustat item

AT for the U.S. firms. To compute AG, we require that a firm has a positive value for total assets at

the end of both fiscal years t−2 and t−1. We treat firm-year observations with absolute values of AG

exceeding 1000% as coding errors and exclude them from analysis. To further alleviate the effect of

outliers, we winsorize the remaining observations at the top and bottom 1% of the distribution within

each country.

4Datastream industry codes are used to exclude SIC-defined financial firms (four-digit SIC codes between 6000 and

6999).

5For convenience, “country” and “market” are used interchangeably in this paper, with the acknowledgement that

some markets are not sovereign countries.

6

Page 9

The last two columns of Table 1 provide the median and standard deviation of the asset growth

rates (AG) for each country, averaged across sample years. There is noticeable cross-country dispersion

in these statistics, with the median AG ranging between 3.886% (Japan) and 64.527% (Turkey) and

the standard deviation between 14.583% (Japan) and 48.647% (Turkey). The median asset growth

rate is slightly lower for the international firms (6.792%) than for the U.S. firms (8.136%). We also

find that the cross-firm variation in the asset growth rate is smaller outside the U.S. (34.632%) than

in the U.S. (42.184%). The greater homogeneity of asset growth rates relative to the U.S. has been

previously documented by Yao et al. (2011) for nine Asian markets.

2.2 Asset Growth and Stock Returns

We now turn to one of the main themes of our paper and examine whether the significant return-

predictive power of AG evidenced in the U.S. is also present in the international markets. We first

quantify the country-level asset growth effect using sorted portfolios. Specifically, we sort stocks in

each country at the end of every June into deciles based on AG measured at the end of June. We

then compute a one-year holding-period return for the AG-sorted decile portfolios and calculate the

return spread between the top and the bottom deciles (SPREAD). This return spread captures the

magnitude of the asset growth effect within each country. To construct the portfolio return series,

we use Datastream’s monthly local-currency return index (RI) for international firms and the CRSP

monthly return data for the U.S. firms.6Following McLean et al. (2009), we trim monthly returns

for the non-U.S. firms at the top and bottom 1% of the distribution within each country to account

for coding errors, but impose no such restriction on the U.S. firms’ returns. The monthly returns are

then equally weighted across stocks belonging to the same AG decile group, and the resulting returns

are compounded from July of this year to June of the next year to obtain annual portfolio returns.

In addition, we consider a value-weighted return spread (VWSPRD), which is the return difference

between the top and bottom value-weighted decile portfolios sorted on AG.7Relative to SPREAD,

VWSPRD captures more of the asset growth effect among large firms.

6All of our results remain similar when we repeat the analysis using the U.S.-dollar converted variables for all countries.

7We also compute the value weighted return spread scaled by value weighted AG spread. It turns out that this measure

is highly correlated with STDSPRD and thus are not reported for the purpose of brevity.

7

Page 10

Recall from Table 1 that the cross-firm dispersion in the asset growth rates differs substantially

across countries. The cross-country difference in the magnitude of the asset growth effect measured by

SPREAD is at least partially due to the difference in the dispersion of asset growth rates. To control

for the effect of AG dispersion, we further construct two measures that capture the return sensitivity

to a per-unit increment in AG. The first is the standardized return spread (STDSPRD), computed as

SPREAD divided AGSPREAD, the difference in the average values of AG between the top and the

bottom AG deciles. The second measure is derived from univariate predictive regressions. Within each

country, we regress annual stock returns (obtained by compounding monthly returns from July of year

t to June of year t + 1) onto AG measured at end of year t. The estimated slope coefficient on AG,

denoted as SLOPE, serves as an alternative measure of the per-unit asset growth effect.

Table 2 reports AGSPREAD and the four measures of the asset growth effect (SPREAD, STD-

SPRD, SLOPE, and VWSPRD) for all countries. The results are sorted by regions and by emerging-

and developed-market status (provided by the International Financial Corporation of the World Bank

Group). The pattern of low stock returns following high asset growth rates is evident in international

markets, as seen by the prevalence of negative return spreads between the top and the bottom AG

decile portfolios. Out of the 41 countries, 27, 25, 28 and 28 countries have negative values of SPREAD,

STDSPRD, SLOPE and VWSPRD respectively. We also find that there are large dispersions in the

magnitude of the AG effect across countries, which will hence be the focus of our analysis in the follow-

ing sections. Outside of the U.S., the average values of SPREAD, STDSPRD, SLOPE and VWSPRD

range respectively from -9.66% (U.K.) to 7.11% (Taiwan), -10.55% (China) to 13.33% (Portugal),

-17.37% (Peru) to 12.00% (Taiwan) and -16.60% (Canada) to 20.00% (Philippines).

Table 2 also provides the corresponding statistics for country subgroups. The mean values of

SPREAD, STDSPRD, SLOPE and VWSPRD averaged across all non-U.S. countries are -1.72%, -

1.32%, -2.21% and -1.08%, respectively. SPREAD and SLOPE are both significant at the 5% level.

Alternatively, if we pool non-U.S. firms together to form the AG-sorted decile portfolios or run the

cross-sectional regressions, the average values of SPREAD, STDSPRD, SLOPE and VWSPRD are

-2.42% (significant at 1% level), -1.80%, -2.10% (significant at 5% level) and -2.46% (significant at 10%

level), respectively. These results, along with the country-level evidence, indicate the pervasiveness

of the asset growth effect in international markets. We notice, however, that the effect is weaker

internationally than in the U.S. The mean values of SPREAD, STDSPRD, SLOPE and VWSPRD for

8

Page 11

the U.S. are -9.65%, -5.68%, -4.51% and -6.30%, respectively, two to five times the magnitude of the

corresponding statistics for the non-U.S. sample (pooled or unpooled).

The table further shows that the AG effect is stronger in developed countries and weaker in emerg-

ing markets. The average values of SPREAD, STDSPRD, SLOPE and VWSPRD, averaged across all

developed markets, are -3.22%, -2.46%, -2.50%, and -4.64%. The corresponding numbers, averaged

across all emerging markets, are -0.54%, -0.34%, -2.01%, and 2.41%. Since stock markets in devel-

oped countries tend to be more efficient, the result appears to suggest a positive relation between the

magnitude of the asset growth effect and stock market efficiency.

The relation between the asset growth effect and stock market efficiency is more explicitly illustrated

in Figure 1. We plot the measures of the asset growth effect across subgroups of countries classified

by three indicators of financial market efficiency. The asset growth effect is quantified by the average

STDSPRD, SLOPE, and VWSPRD across countries within a subgroup. We consider the following three

indicators of market efficiency. The first indicator, stock return synchronicity, measures the extent to

which variations in individual stock returns are driven by aggregate market returns. According to

Roll (1988), Morck, Yeung, and Yu (2000), and Durnev, Yeung, and Zarowin (2003), stock return

synchronicity is inversely related to the pricing efficiency of a stock market. Following the existing

literature, we measure stock return synchronicity by the R-square (R2) of a market model for weekly

stock returns. The second market efficiency indicator is the ratio of equity market capitalization (MKT)

of listed companies of a country to its GDP, based on the observation that stock prices are more efficient

in a more developed stock market. The third indicator is the the ratio of domestic bank credit to the

equity market capitalization (C/M), based on the observation that the stock market tends to be less

developed and less efficient when economy growth is primarily financed by bank credits (e.g., Morck and

Nakamura, 1999; La Porta, Lopez-de-Silanes, and Shleifer, 2002). The details of these three indicators

are provided in Appendix A. Data for domestic bank credit, equity market capitalization and GDP are

all from the World Bank Development Index database.

The figure reveals that the asset growth effect tends to be stronger in countries with more efficient

financial markets, relative to those with less efficient markets. Specifically, the asset growth effect is

stronger (indicated by more negative values of STDSPRD, SLOPE, VWSPRD) in countries with lower

stock return synchronicity, higher market capitalization to GDP ratio, and lower bank credit to equity

9

Page 12

market capitalization ratio.8

This result is immediately interesting, as one normally expects stock

market mispricing to be more likely in less efficient markets, while the rational asset pricing effect to

be more visible when stocks are more efficiently priced.

To shed light on the persistence and the robustness of the asset growth effect, we also conduct

Fama and MacBeth (1973) cross-sectional regression analysis. We regress individual stocks’ holding

period returns, measured over the first month, first year, second year, and third year, on AG and other

standard return-predictive firm attributes every month. Specifically, for regressions from July of year

t to June of year t + 1, our independent variables include AG measured over the previous fiscal year

t − 1, the natural logarithm of the market capitalization measured at the end of June in year t (ME),

the natural logarithm of the book value of common equity at the end of fiscal year t − 1 divided by

the market value of common equity at the end of December in year t − 1 (BM), and the five-month

cumulative return from January to May in year t (MOM), all of which are measured in local currency.9

For non-U.S. firms, ME is constructed using Datastream’s market capitalization variable MV, BM is

calculated using MV and the book value of common equity provided by Worldscope item 03501, and

MOM is given by compounding the percentage change in Datastream return index RI over the relevant

months. For the U.S. firms, the market capitalization is given by the price times the number of shares

outstanding provided by CRSP, BM is constructed following Davis, Fama, and French (2000), and

MOM is constructed by compounding CRSP monthly returns.

We report time-series averages of the intercepts, slope coefficients, and adjusted R2s obtained

from the cross-sectional regressions. We follow Pontiff’s (1996) procedure to calculate t-statistics with

autocorrelation-consistent standard errors. To quantify the average magnitude of the within-country

AG effect in international markets, we pool firm observations from non-U.S. countries into one sample

and estimate monthly regressions with country dummy variables. Following McLean et al. (2009), we

run our regressions by both equal-weighting and scaled-weighting each observation. The latter assigns

each firm-month observation the weight that equals the firm’s market value divided by the average

market value of the firm’s country, both measured at the beginning of the month. Since the scaled-

8An exception to this pattern is that SLOPE is more negative in countries with low MKT. In addition, the difference

is not apparent for STDSPRD between low and high MKT countries.

9We include size, book-to-market ratio, and past returns as control variables in our regressions since prior studies have

shown that their effects are important in many countries. See, for example, Heston, Rouwenhorst, and Wessels (1995) for

the size effect, Fama and French (1998) for the value effect, and Rouwenhorst (1998) and Griffin, Martin, and Ji (2003)

for the momentum effects in international markets.

10

Page 13

weighting is equivalent to a within-country value-weight, the results from the two regressions show how

the AG effect varies between small and large firms within each country.

Panel A of Table 3 shows that the AG effect is robust to the inclusion of various control variables.

The AG coefficients are significantly negative for one-month and one-year holding periods and turn

insignificant thereafter with both equal- and scaled-weights. The slope coefficients on AG and associated

t-statistics are larger for the equal-weighted regressions than those for the scaled-weighted regressions,

indicating the stronger association between asset growth rates and future stock returns for smaller

firms. As a comparison, we also estimate the same cross-sectional relationship for the U.S. sample

by both equal- and value-weighting firm-month observations. Panel B of Table 3 shows a significant

asset growth effect in the U.S. after controlling for the size, value, and momentum effects. Further,

the AG effect is stronger for smaller U.S. firms since the AG slope coefficients and t-statistics for the

equal-weighted regressions are larger than those for the value-weighted regressions. The AG effect is

also more persistent for smaller firms in the U.S. in that its effect remains significant for up to two years

in equal-weighted regressions whereas it remains significant for only up to one year in value-weighted

regressions. These results are consistent with the U.S. evidence documented by Cooper et al. (2008).

A comparison of the U.S. and international evidence further reveals that the economic magnitude,

statistical significance, and persistence of the asset growth effect are all stronger in the U.S. than in

international markets. For example, the equal-weighted one-month holding period regression yields an

AG coefficient estimate of -0.90 (with t-statistics = -8.85) for the U.S. and -0.30 (with t-statistics =

-4.93) for the non-U.S. sample. This implies that one standard deviation increase in the asset growth

rate (42.184% for the U.S. and 34.632% for non-U.S.) leads to a -0.38% and -0.10% decline in the cross

section of monthly stock returns in the U.S. and in international markets, respectively. In addition, the

t-statistics for the AG slopes are generally higher than those for other return-predictive characteristics

in the U.S., whereas the corresponding statistics are lower than those for BM and MOM coefficients

outside of the U.S. The relatively high R2s we observe for the non-U.S. sample compared to those for

the U.S. is due to the inclusion of country dummies in their regressions. Overall, both our portfolio

sort and regression analysis provide evidence of significant negative effect of asset growth rates on

future cross-sectional returns in international markets, albeit weaker than that in the U.S., and the

substantial dispersion in the magnitude of the AG effect across these countries.

11

Page 14

3 The Asset Growth Effect: The Cash Flow Channel and the Dis-

count Rate Channel

3.1The Cash flow and Discount Rate Channel of Investment Effect: A Decompo-

sition

The empirical results reported in the previous section show that the asset growth effect is stronger in

developed markets and in markets where stocks are more efficiently priced. This suggests, although

indirectly, that the asset growth effect may be rational investment phenomenon. In this section, more

directly, we examine to what extent the two rational investment channels predicted by the q-theory

model explain the cross-country variation of the AG effect.

The investment-based asset pricing model has been used to explain the cross section of stock returns

by several existing studies. One of the rational investment effects, known as the discount rate channel

in the literature, is first introduced by Cochrane (1991). Specifically, under the q theory, all else equal,

a low discount rate implies high marginal q and high investment, and a high discount rate implies low

marginal q and low investment. This theory thus implies that firms increase their investments when

discount rates are low, and cut their investments when discount rates are high. Cochrane (1996) and

Liu, Whited, and Zhang (2009) empirically analyze the discount rate channel. The second rational

investment effect is known as the cash flow channel. While the intuition for this channel has previously

existed, its formal development is relatively recent; e.g., Li, Livdan and Zhang (2009), and Lyandres,

Sun and Zhang (2007). This channel is based on decreasing return to scale. With decreasing return to

scale, more investments lead to lower investment returns, which, combined with a monotonic relation

between investment return and stock return, means lower stock returns.

Cochrane (1996) and Liu, Whited, and Zhang (2009) empirically examine the discount rate channel

of investment. Several other studies have linked patterns of cross-sectional stock returns to the effects

of both channels. For example, Li, Livdan and Zhang (2009) shows that optimal investment is an

important driving force of external financing anomalies. Lyandres, Sun and Zhang (2007) shows that

investment factor helps explain the new issue puzzle. Wu, Zhang and Zhang (2010) also finds that

q-theory model is useful in understanding the accrual anomaly. Most recently, Chen, Novy-Marx and

12

Page 15

Zhang (2010) develop an alternative three-factor model to explain a series of anomalies, including the

asset growth effect, and provide evidence that their new model outperform the existing asset pricing

models.

The two channels are illustrated in Li, Livdan and Zhang (2009) using a simple and intuitive two-

period model of investment with adjustment cost to capital. In such a two period model, investment

return equals stock return. There are two periods, 1 and 2. A firm’s production function is given by kα

t

where k is capital, α < 1 with t = 1,2. k1depreciates at the rate of δ, hence k2= i + (1 − δ)k1, where

i1is investment in period 1. There are quadratic adjustment costs (a/2)(i1/k1)2k1, with a > 0. The

firm faces a gross discount rate (expected return) of r, which is known at the beginning of period 2.

Then the cash flow channel works through decreasing returns to scale, and the discount rate channel

works through capital adjustment costs:

∂r

∂i=α(α − 1)kα−2

1 + a(i/k1)

?

2

???

cash flow channel

−

?αkα−1

[1 + a(i/k1)]2k1

?

2

+ 1 − δ?a

???

discount rate channel

< 0(1)

This simple yet powerful relationship can actually be taken to empirical tests, which motivates our

analysis in this section. We illustrate the basic idea of our approach in achieving that goal. Consider

the following regression to capture the ∂r/∂i in the above equation:

rt+1= α + β∆at+ εt+1

where r is investment return at year t + 1, ∆at is the asset growth at year t, a proxy for gross

investment. Then the regression coefficient β corresponds to the partial derivative ∂r/∂i as shown

above. In addition, since equation (1) suggests that β is driven by both a cash flow channel and a

discount rate channel, it is also imperative that we study the role of the two different channels separately.

Further we decompose this slope coefficient β to reflect these separate effects, by identifying the nature

of shocks, i.e, those due to cash flow variations or discount rate variations.

Unlike Li, Livdan and Zhang (2009), we do not assume specific functions of production or adjust-

ment costs. Rather, our approach follows Cochrane (2008) and is based on the present value relations.

We use log-linearization to derive a closed-form expression for the above regression coefficient and its

13

Page 16

components. An advantage of this approach is its generality, which enables us to empirically quantify

the rational investment channels from the data. However, this also comes with an apparent downside:

absent of stronger rational asset pricing restrictions, one cannot completely rule out that the derived

empirical measures of the cash flow channel and the discount rate channel could be influenced by

mispricing. We take this into account when performing empirical analysis.

3.1.1 Intertemporal Budget and Log-linearization

Our set-up is quite similar to Larrain and Yogo (2008), and we refer readers to their paper for many

of the details. Larrain and Yogo (2008) use the following accounting identity to study the changes in

firms’ total asset values due to the variations both in cash flow and discount rates. Let Yt+1be earnings,

It+1investment, At+1total asset, Et+1net payout (the net cash outflow from the firm, composed of

dividends, interest, equity repurchase net of issuance, and debt repurchase net of issuance), Rt+1the

return on assets, and Et+1/At+1net payout yield.10Then by definition the firm’s total return on assets

is this period’s earnings over last period’s asset value, plus 1:

Rt+1= 1 +Yt+1

At

Then the firm’s intertemporal budget is

At+1= At+ It+1= At+ Yt+1− Et+1

Then we can rewrite

Rt+1=At+1+ Et+1

Et+1

Et+1

Et

Et

At

10An advantage of the Larrain and Yogo approach is that investors’ cash flow is measured by the net payout, rather

than by dividends as in the traditional approach. This difference is important for the separation of the cash flow effect

and the discount rate effect for our purpose.

14

Page 17

With lower case letters as log of upper case letters, and leaving out constants in expressions, log-

linearization of the above yields the approximate identity

rt+1

=ρ∆at+1+ (1 − ρ)(et+1− at) (2)

ρ=1/(1 + exp[E (e − a)])

∞

?

j=1

et− at

=Et

ρj−1(rt+j− ∆et+j) (3)

where ρ is a constant slightly below 1, related to the typical level of the payout yield. The last equation

3 simply means net payout yield et−atis a cointegration error that must forecast long run cash outflow

growth or return on assets, or some combination of both.

3.1.2 A simple VAR

Following the similar setup in the literature for stock return predictability (e.g., Cochrane 2008; Stam-

baugh 1999), we study return on assets with a simple vector autoregrssion (VAR) where the predictive

power of net payout yield is justified as above:

rt+1

=ar+ br(et− at) + εr

t+1

∆et+1

=ae+ be(et− at) + εe

t+1

et+1− at+1

=aea+ φ(et− at) + εea

t+1

rt+1

=ρ(at+1− et+1) + ∆et+1− (at− et)

where in return (payout growth) equation the predictive coefficient is br(be), and the autocorrelation

coefficient of payout yield is φ. The last equation is just a replication of equation (2), which dictates

the cross equation restriction such that

br

=1 − ρφ + be

εr

r+1

=εe

t+1− ρεea

t+1

15

Page 18

therefore the error terms follow

εr

r+1

εe

t+1

εea

t+1

˜

0

0

0

,

σ2

r

σr,e

1

ρ

?σr,e− σ2

?σ2

?σ2

r

?

?

σr,e

σ2

e

1

ρ

e− σr,e

1

ρ

?σr,e− σ2

r

?

1

ρ

?σ2

e− σr,e

?

1

ρ2

r− 2σr,e+ σ2

e

?

3.1.3 Cash Flow channel and Discount Rate Channel

We now derive the theoretical value of “discount rate β” and “cash flow β”. It can be shown that, in

the simple VAR system described above, when regressing the future return on assets rt+1onto asset

growth ∆at, the coefficient β has the following expression:

β =Cov(rt+1,∆at)

V ar(∆at)

= βr+ βcf+ βr,fc

(4)

where

βr

=

?

1

ρb2

r+

?

1 −1

ρ

?

(brbe+ br)

?

φ1

ρ2

1

1−φ2−br

ρ2

V ar(∆at)

σ2

r

?

?

???

ρ−1

ρ2

discount rate channel

1 −1

ρ

(brbe+ br)

V ar(∆at)

(5)

βcf

=

1

ρb2

r+

???

φ1

ρ2

1

1−φ2+ br

??

σ2

e

?

−

???

cash flow channel

(6)

βr,cf

=

?

1

ρb2

r+

?

1 −1

ρ

?

(brbe+ br)

?

φ1

ρ2

2

1−φ2+ br

?

ρ−1

ρ2

?

V ar(∆at)

σr,e

? ???

interaction

(7)

where

V ar(∆at+1)= V ar

?1

?

ρrt+1+

σ2

ea

1 − φ2+ σ2

?

?

1 −1

ρ

?

(∆et+1+ et− at)

?

=

1

ρ2

+2ρ − 1

b2

rr

?

+

?

σ2

1 −1

ρ

?2?

(be+ 1)2

σ2

ea

1 − φ2+ σ2

e

?

ρ2

br(be+ 1)

ea

1 − φ2+ σe,r

?

We identify these two channels by the source of the shocks. We collect all the terms involving σ2

r,

16

Page 19

the variation of return on asset forecast equation shocks and denote it “discount rate β” as it purely

reflects the variation of ROA shocks εr

t+1= rt+1− Etrt+1. Similarly, we collect all the terms involving

σ2

e, the variation of payout growth forecast equation shocks and denote it “cash flow β” as it purely

reflects the variation of cash flow shocks εe

t+1= ∆et+1− Et∆et+1. The last term βr,cfcaptures the

interaction of these two shocks σr,eand can not be easily decomposed.

3.2 Estimation of the Cash Flow Beta and the Discount Rate Beta

We use the VAR described in Section 3.1.2 to estimate the betas – β, βcf, βr, and βcf,r– for each

country. To implement the VAR, we first compute the average ρ = 1/[1 + exp(e − a)] for each country.

Then we conduct a panel regression according to the equations in the VAR system with the cross

equation restriction imposed. We use GMM to estimate the panel VAR for each country. Then, we

plug the estimated coefficients into the expressions (4), (5), and (6), and (7) to obtain the betas. Thus,

for each country we have one point estimate for each beta.

The variables in the VAR are defined following Larrain and Yogo (2008). The net payout Etis

the net cash outflow from the firm (composed of dividends, interest expense, equity repurchase net of

issuance, and debt repurchase net of issuance). We use the book value of debt and market value of

equity to calculate total value of assets At. Then return on assets is the value of total assets plus net

payout over the last period total asset value. The payout yield is calculated as the net payout over the

value of total assets in the same period.

One difference between our empirical implementation and that of Larrain and Yogo (2008) is that

they estimate the model using the time series observations at the market level, while we deal with

the firm-level data. Therefore we use the panel VAR approach and assume that the firms within each

country follow the same dynamics, a cross-sectional assumption that has been adopted by a few existing

studies, such as Vuolteenaho (2002), and Cohen, Gompers, and Vuolteenaho (2002). To deal with the

dynamic-panel nature of our sample, we use the Arellano-Bond (1991) GMM procedure to estimate

the panel VAR. Details of this procedure are provided in Appendix B of this paper.

Table 4 reports the betas estimated for the 41 countries (including the U.S.). First, the beta, which

measures the overall investment-investment return relation, varies from -0.90 (China) to 0.83 (Israel).

17

Page 20

Recall that a positive β suggests that investment is positively associated with investment return and

a negative β suggests a negative relationship between investment and investment return. Across all

markets, the mean of β is -0.007 and the standard deviation is 0.308. Similarly, there is a large cross-

sectional variation in the cash flow beta βcf, which varies from -0.99 (China) to 0.55 (Norway), with

a cross-sectional standard deviation of 0.286. By comparison, although the average magnitude of the

discount rate beta βris comparable to that of the cash flow beta, the cross-sectional variation of βr

are much smaller, which varies from -0.18 (China) to 0.17 (Israel), with a cross-sectional standard

deviation of 0.069. The cross-country variation of the beta interaction effect, βcf,r, is also relatively

small, with a cross-sectional standard deviation of 0.059. On the face of these variations, the lion’s

share of negative investment return-investment effect captured by the overall β is due to the cash flow

channel, as most of the cross-country variation in β is driven by βcf.

Table 4 also provides a number of country characteristics that will be used in junction with the

betas in our subsequent analysis. They include indicators for market efficiency and economic growth,

indicators for a country’s legal origin, variables related to investor protection and corporate governance,

and variables related to limits to arbitrage. These variables will be discussed when they are formally

introduced in analysis.

Further, in Table 5, we report the cross-sectional correlations among the betas and the country

characteristics. The correlation among β, βr, βcfare high, all above 0.85. Meanwhile, the correlations

of βcf,rwith the other three betas are highly negative. The high magnitude of correlations across

the betas suggests that we may encounter the multi-collinearity problem when these betas are jointly

used as explanatory variables. In addition, the betas tend to have relatively low correlations with

various country characteristics, except that they exhibit modest-level correlations with the stock return

synchronicity, the Anti-Director rights index, and the short-sale restriction indicator.

3.3 The Baseline Regression

It is worth clarifying that the betas are estimated from the observed relation between investment and

investment returns, not from the relation between asset growth and stock returns. Therefore, the

mapping between the betas and the asset growth effect on stock return is not a tautology. Rather, such

18

Page 21

a mapping is predicted by the q-theory model. Under the q-theory model, a firm’s investment reduces

expected investment returns (i.e., return to per unit capital investment) via both the cash flow channel

and the discount rate channel. Further, the theory suggests that there is a positive mapping between

the investment return and stock return.11Therefore, the investment based asset pricing theories predict

a positive correlation between the betas (i.e. the investment-investment return relation) and the asset

growth effect (i.e. the investment-stock return relation). We now perform cross-country analysis to

examine this prediction.

In the baseline cross-country regressions we examine whether the variation of the betas matches the

variation of the asset growth effect. The dependent variables are STDSPRD, SLOPE, and VWSPRD,

respectively. The first two represent the per-unit asset growth effect, and the third one captures the

total asset growth effect but is less subject to the influence of small firms (relative to SPREAD). The

explanatory variable is one of the four betas. As noted earlier, the correlations among these betas are

high, and therefore we do not consider a regression where the betas are joint regressors. We follow

Karolyi et al. (2007) and conduct a series of ordinary least squares (OLS) estimations. When drawing

statistical inferences we use bootstrapped standard errors (with 1,000 bootstraps) to take into account

potential heteroscedasticity of the observations.12

Table 6 reports the baseline regression results. We find that the β and the first two beta components

βcfand βrare statistically significant in explaining the AG effect, no matter which of the three

dependent variables we use. Specifically, when the dependent variable is STDSPRD, the coefficients

of β, βcfand βrare 0.074, 0.072,and 0.326 with t-stats of 2.58, 2.51 and 2.42 respectively. When the

dependent variable is SLOPE, the coefficients of β, βcf, βrare 0.063, 0.065, and 0.319, with t-stats of

2.25, 2.04 and 2.44 respectively. Finally, when the dependent variable is VWSPRD, the coefficients of

β, βcf, βrare 0.102, 0.107, and 0.328, with t-stats of 3.46, 2.96 and 1.86 respectively. These results

suggests that indeed the asset growth effect is related to the impact of investment on investment return,

and that both the cash flow channel (the decreasing return to scale effect) and the discount rate channel

(the capital adjustment cost effect) can explain the asset growth effect across countries.

11Investment return equals stock return in a two period model or in a model with constant return to scale and constant

financial leverage. See, e.g., Li, Livdan and Zhang, 2009, and Liu, Whited and Zhang, 2009.

12Essentially this is the between-group estimation. Since our explanatory variables are time-invariant, this regression

is equivalent to the Fama-MacBeth regression in a balanced panel. We have also obtained jackknife standard errors. We

further conduct panel regressions with clustered standard errors both with time and country. All these additional results

remain similar and are available upon request.

19

Page 22

In contrast, the explanatory power of the beta-interaction component βcf,ris non-existing, as the

coefficients for βcf,rare never significant. We note that the literature on the rational investment

channels does not have any specific prediction related to the interaction effect that βcf,rrepresents.

Therefore in subsequent analysis, we drop βcf,rand focus on β, βcf, and βr.

3.4Subsample Analysis

As noted earlier, the betas are derived under the present value relation, without imposing structural

assumptions on the production function or adjustment cost function. The upside of this approach is its

generality. On the other hand, the downside is that the relation between the resulting betas and the

asset growth effect is not uniquely predicted by the investment-based asset pricing models. In q-theory

based asset pricing framework the betas capture the cash flow channel and the discount rate channel of

the investment effect. But the betas can still exist even when asset prices are not rationally determined.

For example, a negative relation between investment and subsequent cash flow profitability can exist

when firms over-invest or when firms manage their earnings upward prior to financing or acquisition

activities. Therefore, a sharper test is desirable in order to fully differentiate the rational and behavioral

effects on the betas.

The substantial cross-country variation in the financial market efficiency offers a way to test a

sharper, conditional prediction. In a more efficient market, security mispricing is less rampant, and

therefore the opportunistic corporate external financing behavior, as well as investors’ extrapolation

bias, should be less prevalent. Arguably, more efficient and more informative stock prices can also

enable shareholders better monitor corporate managers and thus more effectively curb over-investments.

Therefore, if the relation between the betas and the asset growth effect is driven by over-investment,

investors’ extrapolation bias, or firms’ opportunistic external financing, this relation should be less

prevalent in more efficient financial markets. On the other hand, if the asset growth effect is driven

by the rational investment effect, the ability of the betas to explain the asset growth effect should be

stronger in financial markets where stock prices more efficiently reflect fundamental information.

We test this prediction using subsample regression analysis. We categorize countries into two groups

based on measures of their financial market efficiency. We then examine the relation between the asset

20

Page 23

growth effect and the betas separately in the two groups, using the regressions as those in the baseline

analysis. That is, the dependent variables are STDSPRD, SLOPE, and VWSPRD respectively, and

the explanatory variable is one of the three betas: β, βcf, and βr.

The three market efficiency indicators are those considered in Section 2.2: stock return synchronicity

(R2), equity market capitalization to GDP ratio (MKT), bank credit to equity market capitalization

ratio (C/M), with detailed definition provided in Appendix A. Since the indicator values can change

over time for a given country, we first compute their time series averages for a country, and then

compute the cross-country means. A market is defined as efficient if its (time-averaged) R2 is below,

or MKT is above, or C/M is below, the respective cross-country means. The values of these indicators

for each country are reported in Table 4, with the correlations of these indicators with other variables

used in the study reported in Table 5.13

Table 7 reports the results, with panel A, B and C separately for each of the three dependent

variables STDSPRD, SLOPE and VWSPRD. The overall pattern is that the relation between the

betas and the asset growth effect is stronger in more efficient financial markets, i.e., countries with low

stock return synchronicity (R2), high market cap to GDP ratio (MKT), and low bank credit to market

cap ratio (C/M). This is the opposite of what is predicted by the mispricing-based hypothesis, and

consistent with the hypothesis that the asset grow effect, and the effects captured by the betas, are

rational asset pricing effects.

For example, consider in Panel A, where the dependent variable is STDSPRD. When the explana-

tory variable is β, its slope coefficient is 0.115, with a t-stat of 3.661 in the low R2 group. In contrast,

in the high R2 group, not only is the magnitude of the slope coefficient for β ten times smaller (only

0.011), but the t-stat is just 0.268. We further note that this result is not driven by a small sample for

any of the two groups, as the low R2 and high R2 groups have about the same numbers of countries

(22 vs. 19). Similarly, for the high MKT group, the slope coefficient for β is 0.102, with an significant

t-stat of 2.214, while for the low MKT group, the slope is 0.058 with an insignificant t-stat of 1.108.

13It is important to note that our inference does not rely solely on a single efficiency measure, but rather is reinforced

by the consistent results from multiple efficiency measures. Any individual efficiency measure is likely to be noisy. For

example, despite supportive evidence provided by Morck, Yeung, and Wu (2000) and Dunev, Morck, and Yeung (2006),

a few recent studies have questioned the validity of R2 as a negative indicator of stock price efficiency; see, e.g., Chan

and Hameed (2006), Kelly (2005), Ashbaugh-Skaife, Gassen, and LaFond (2006), Griffin, Kelly, and Nadari (2006), Hou,

Peng, and Xiong (2007), and Teoh, Yang, and Zhang (2008).

21

Page 24

Finally, for the low C/M group, the coefficient for β is 0.082, with a significant t-stat of 2.218, while

for high C/M countries, the slope is 0.069 with a t-stat of 0.429. This pattern for the coefficients on

the other two betas, βcfand βr, appear to be similar.

The same pattern exists in Panel B, where the dependent variable is SLOPE, although the explana-

tory power of βs varies. β has significantly positive coefficients in low-R2 and low-C/M countries, while

insignificant coefficients in high-R2 and high-C/M countries. The beta coefficients are not significant

in either of the groups when the countries are classified by MKT. The coefficients for βcfare consistent

with the rational asset pricing prediction when the efficiency indicator is based on R2, but the coeffi-

cients are both insignificant when countries are classified by MKT or C/M. The coefficients for βcfare

consistent with the rational prediction in all three ways of classifying the countries.

Finally, in Panel C, where the dependent variable is VWSPRD, the pattern is also largely consistent

with the rational prediction, yet with somewhat richer variations. The coefficients for the betas are

significant in efficient markets and insignificant in inefficient markets for the following pairs of beta and

market efficiency indicator: β and C/M, βcfand C/M, βrand MKT. The coefficients are significant

in both low-efficiency and high-efficiency groups for the following pairs of beta and market efficiency

indicator: β and R2, β and MKT, and βcfand MKT. However, even in these three cases, the beta

coefficients are higher for the high-efficiency group than for the low-efficiency group, consistent with

the rational prediction. There are two cases where the coefficients for βrare not significant for either

groups – when the efficiency indicator is R2 or C/M. Finally, there is only one case that the pattern

is explicitly inconsistent with the rational prediction – the coefficient for βcfis insignificant for the

high-MKT group and significant for the low-MKT group.

3.5 Controlling for Country Characteristics of Governance and Limits to Arbitrage

We now turn the attention more directly to the predictions under the mispricing-based hypotheses.

The mispricing hypotheses attribute the asset growth effect to corporate managers’ over-investment

tendency, market timing behavior in corporate financing activities, or investor over-reaction to past

firm growth. Under these explanations, the asset growth effect should be weaker among countries

with better corporate governance and better investor protection, and among countries with less severe

22

Page 25

limits to arbitrage. In this part of the analysis, we test the cross-country explanatory power of country

characteristic variables that are proxy for governance and limits to arbitrage in junction with the betas.

We consider the following country-specific variables that might be relevant to explain the asset

growth effect under the mispricing hypothesis. First, a country’s legal origin is often considered an

indicator of the country’s investor protection and corporate governance effectiveness. Accordingly, we

create dummies for major legal traditions – UK (United Kindom), FR (France), GE (Germany), and

SC (Scandinivian). As documented by La Porta, Lopez-deSilanes, Shleifer and Vishny (2000, hereafter

LLSV), countries with English legal origin typically are more effective in corporate governance. We

therefore should expect a negative relation between the UK legal origin and the magnitude of the AG

effect under the mispricing hypothesis. Second, we use the Revised Anti-Directed rights index (AD)

and Anti-Self dealing index (AS) by Djankov, La Porta, Lopez-Silanes, and Shleifer (DLLS, 2008) as

proxies for corporate governance quality. The value of the Revised AD index varies between 0 and 5,

and the value of the AS index varies between 0 and 1. A higher value indicates a higher level of investor

protection in that country. Third, we follow LLSV (2000) to consider a measure of accounting quality

(Account), which is an index based on the reporting or omission of 90 items from annual reports. A

higher value of Account indicates better accounting reporting quality, and we expect a negative relation

between Account and the magnitude of the asset growth effect under the mispricing hypothesis. Finally,

we consider two measures associated with limits to arbitrage – stock trading liquidity and short sale

restrictions. Stock liquidity is measured by Turn, the average trading volume share as a percentage of

total shares outstanding during the previous year averaged across all stocks. Short is a dummy variable

equal to 1 if short selling is allowed and zero if it is not allowed. We obtained this measure from Bris,

Goetzmann, and Zhu (2007). The values of these variables for each country are reported in Table 4,

with the correlations of these indicators with other variables used in the study reported in Table 5.

To summarize the above, under the mispricing hypothesis, we expect countries with the UK legal

origin, better governance and investor protection (lower values of anti-director rights index and higher

values of anti-self dealing index), higher accounting quality, higher stock liquidity, and lower short-

sale constraints to have lower magnitude of the asset growth effect. However, we note a caveat here

– these country characteristics may also be associated with stock market development and market

efficiency. For example, countries with English legal origin tend to develop their stock market earlier.

Good accounting quality, stock liquidity and lack of short-sale restriction are also indicators of market

23

Page 26

maturity. Therefore, even when we indeed document a significant relation between any of the country

characteristics and the asset growth effect, mispricing may not be the exclusive reason for the link.

We perform cross-country regressions. The dependent variable is one of the three measures of the

asset growth effect – STDSPRD, SLOPE, and VWSPRD. The explanatory variables include one of

the three betas, and combinations of the country characteristics discussed above. In implementing the

regression analysis we encounter a technical issue that requires an additional note. As shown in Table

5, the correlations among some of the control variables are quite high. In particular, the correlation

between the legal origin indicator UK and the anti-self dealing index (AS) is 0.76. As a consequence,

when we include all the country characteristic variables as joint regressors, we find a multi-collinearity

problem. Therefore we choose to include all the legal origin variables in the regressions, but only

to include the one of the remaining characteristics in each regression. We have also experimented

with other regression specifications that do not suffer from multi-colinearity, and have obtained results

consistent with those reported in the table. For brevity these additional regression results are not

tabulated.

Table 8 reports the results with STDSPRD as the dependent variable. Panel A, B and C separately

focus on results with β, βcfand βras the main independent variable. Again, we do not put the

betas as joint regressors due to the high correlations among them due to the multi-colinearity concern.

Common to each regression are the country characteristics discussed above. Notably, the explanatory

power of βs remains intact when we add different combinations of country characteristic variables.

The dummies for legal origins, UK, FR, and GE tend to be significant and with the opposite

sign to that of the regression intercept. The magnitudes of these fixed effect estimates are, however,

similar and tend to offset that of the common intercept.

14This is inconsistent with the mispricing

hypothesis which predicts countries with weaker legal system such as FR will have stronger AG effects

than those with stronger legal system such as UK. Other than these legal origin dummies, none of the

additional control variables, such as Revised AD rights index, the AS index, accounting quality, short-

sale constraint, and stock liquidity, has a significant impact on STDSPRD. The results are qualitatively

similar when we use βcf(Panel B) and βr(Panel C) in the regression.

14Given the relatively large standard errors of the intercept and fixed effects, formal tests show that each group’s

separate intercept estimate has small t-stats.

24

Page 27

Table 9 reports the results with SLOPE as the dependent variable. The explanatory power of the βs

remains intact after controlling for country characteristics developed under the mispricing hypothesis.

On the other hand, these country characteristics themselves (other than the legal origin dummies)

appear to have little explanatory power on SLOPE, inconsistent with the mispricing hypothesis. The

only exceptions are the coefficients for the short-sale restriction, which are significantly positive.

Finally, Table 10 reports the results with VWSPRD as the dependent variable. Similar to the

previous two tables, the explanatory power of β and βcfremains strong after controlling for country

characteristics developed under the mispricing hypothesis. However, an exception is the result in Panel

C, where the coefficients for the discount rate beta βrtends to be insignificant. Comparing with the

results in Panel A and B, one can perhaps infer that the across-country variation in the asset growth

effect, when measured by the value weighted asset growth portfolio return spread, is mainly driven

by the cash flow channel, and the effect of the discount rate channel is rather weak. Moreover, given

that there is no clear distinction in the magnitude of the two channels when the AG effect is measured

by STDSPRD and SLOPE (both giving equal weights to small and large firms), we can infer that

the cash flow channel is the main working channel for the AG effect in large firms (evidenced by the

VWSPRD result in Panel C here), while both channels work for the AG effect in small firms. We see

this evidence as more consistent with the q-theory model. As larger firms are naturally more important

for a corporate sector, the fact that the stock return spreads due to asset growth for these firms solely

reflect the changes in rational cash flows expectations will be more supportive for a rational argument

(for a similar argument see Cohen, Polk, and Vuolteenaho, 2003).

4 Conclusion

The contribution of this paper can be summarized as follows. First, we document the existence of

a negative relation between asset growth and future stock returns in 40 international equity markets

outside the U.S. Thus, we show that the asset growth effect initially documented by Cooper et al.

(2008) for the U.S. market is a pervasive global asset pricing phenomenon. Interestingly, we also find

that the asset growth effect is stronger in developed markets and in markets where stocks tend to

be more efficiently priced. We take advantage of the large cross-country variations in the magnitude

25

Page 28

of the asset growth effect and in various country characteristics related to market efficiency, investor

protection, and market frictions to evaluate hypotheses on the cause of the asset growth effect.

Second, we explicitly examine the two rational investment effects predicted by the q-theory model

– the cash flow channel and the discount rate channel. We quantify these two channels by decomposing

the the investment-investment return relation for each country. The resulting measures, termed the

cash flow beta and the discount rate beta, are empirically successful in explaining the asset growth

effect across countries. Their explanatory power tends to be stronger among markets with more efficient

stock prices, and remains significant after controlling for various country characteristics motivated by

the mispricing-based hypotheses. On the other hand, these country characteristics do not have strong

power to explain the cross-country difference in the asset growth effect.

We therefore conclude that the rational asset pricing effects of corporate investments perform better

than mispricing-based hypotheses in explaining the asset growth effect across countries.

26

Page 29

References

Abel, A. B., N. G. Mankiw, L. H. Summers, R. J. Zeckhauser, 1989. Assessing Dynamic Efficiency:

Theory and Evidence. The Review of Economic Studies 56(1), pp. 1–19.

Agrawal, A., J. F. Jaffe, G. N. Mandelker, 1992. The Post-Merger Performance of Acquiring Firms: A

Re-Examination of an Anomaly. The Journal of Finance 47(4), 1605–1621.

Anderson, C. W., L. Garcia-Feijoo, 2006. Empirical Evidence on Capital Investment, Growth Options,

and Security Returns. Journal of Finance.

Arellano, M., S. Bond, 1991. Some Tests of Specification for Panel Data: Monte Carlo Evidence and

an Application to Employment Equations. Review of Economic Studies 58(2), 277–297.

Baker, M., J. Wurgler, 2002. Market Timing and Capital Structure. The Journal of Finance 57(1),

1–32.

Berk, J. B., R. C. Green, V. Naik, 1999. Optimal Investment, Growth Options, and Security Returns.

The Journal of Finance 54(5), 1553–1607.

Bris, A., W. Goetzmann, N. Zhu, 2007. Efficiency and the bear: short sales and markets around the

world. Journal of Finance.

Carlson, M., A. Fisher, R. Giammarino, 2004. Corporate Investment and Asset Price Dynamics: Im-

plications for the Cross-Section of Returns. The Journal of Finance 59(6), 2577–2603.

Chan, K., A. Hameed, 2006. Stock price synchronicity and analyst coverage in emerging markets.

Journal of Financial Economics 80(1), 115 – 147.

Chen, L., Z. Da, L. Borja, 2010. What Moves Investment?. Working paper, Washington University.

Chen, L., R. Novy-Marx, L. Zhang, 2010. An Alternative Three-Factor Model. SSRN eLibrary.

Cochrane, J. H., 1991. Production-Based Asset Pricing and the Link Between Stock Returns and

Economic Fluctuations. The Journal of Finance 46(1), 209–237.

, 1996. A Cross-Sectional Test of an Investment-Based Asset Pricing Model. The Journal of

Political Economy 104(3), 572–621.

, 2008. The Dog That Did Not Bark: A Defense of Return Predictability. Review of Financial

Studies 21(4), 1533–1575.

Cohen, R. B., P. A. Gompers, T. Vuolteenaho, 2002. Who underreacts to cash-flow news? evidence

from trading between individuals and institutions. Journal of Financial Economics 66(2-3), 409 –

462.

Cohen, R. B., C. Polk, T. Vuolteenaho, 2003. The Value Spread. The Journal of Finance 58(2), pp.

609–641.

Cooper, M., H. Gulen, M. Schill, 2008. Asset Growth and the Cross-Section of Stock Returns. Journal

of Finance 63(4).

27

Page 30

Davis, J. L., E. F. Fama, K. R. French, 2000. Characteristics, Covariances, and Average Returns: 1929

to 1997. The Journal of Finance 55(1), pp. 389–406.

Dimson, E., 1979. Risk measurement when shares are subject to infrequent trading. Journal of Financial

Economics 7(2), 197 – 226.

Djankov, S., R. L. Porta, F. L. de Silanes, A. Shleifer, 2008. The law and economics of self-dealing.

Journal of Financial Economics 88(3), 430 – 465.

Durnev, A., R. Morck, B. Yeung, P. Zarowin, 2003. Does Greater Firm-Specific Return Variation Mean

More or Less Informed Stock Pricing?. Journal of Accounting Research 41(5), 797–836.

Fama, E. F., K. R. French, 1998. Value versus Growth: The International Evidence. The Journal of

Finance 53(6), pp. 1975–1999.

, 2006. Profitability, investment and average returns. Journal of Financial Economics 82(3), 491

– 518.

Fama, E. F., J. D. MacBeth, 1973. Risk, Return, and Equilibrium: Empirical Tests. The Journal of

Political Economy 81(3), 607–636.

Griffin, J. M., X. Ji, J. S. Martin, 2003. Momentum Investing and Business Cycle Risk: Evidence from

Pole to Pole. The Journal of Finance 58(6), 2515–2547.

Griffin, J. M., P. Kelly, F. Nardari, 2006. Are Emerging Markets More Profitable? Implications for

Comparing Weak and Semi-Strong Form Efficiency. Working paper, University of South Florida.

Heston, S. L., K. G. Rouwenhorst, R. E. Wessels, 1995. The structure of international stock returns

and the integration of capital markets. Journal of Empirical Finance 2(3), 173 – 197.

Hou, K., L. Peng, W. Xiong, 2006. A Tale of Two Anomalies: The Implication of Investor Attention

for Price and Earnings Momentum. Working paper, Ohio State University.

Ikenberry, D., J. Lakonishok, T. Vermaelen, 1995. Market underreaction to open market share repur-

chases. Journal of Financial Economics 39(2-3), 181 – 208.

Ince, O., R. Porter, 2006. Individual equity return data from Thomson Datastream: Handle with care!.

Journal of Financial Research 29, 463–479.

Jin, L., S. C. Myers, 2006. R2 around the world: New theory and new tests. Journal of Financial

Economics 79(2), 257 – 292.

Karolyi, A., K.-H. Lee, M. V. Dijk, 2007. Commonality in Returns, Liquidity and Trading Volume

around the World. Working paper, Ohio State University.

Kelly, P., 2005. Information Efficiency and Firm-Specific Return Variation. Working paper, University

of South Florida.

Lakonishok, J., A. Shleifer, R. W. Vishny, 1994. Contrarian Investment, Extrapolation, and Risk. The

Journal of Finance 49(5), 1541–1578.

28

Page 31

Lam, F. E., K. C. J. Wei, 2009. The Asset Growth Anomaly and the Role of Limits to Arbitrage. Hong

Kong University of Science and Technology working paper.

LaPorta, R., F. Lopez-de Silanes, A. Shleifer, R. Vishny, 2002. Investor Protection and Corporate

Valuation. Journal of Finance 57(3), 1147–1170.

LaPorta, R., F. Lopez-de Silanes, A. Shleifer, R. W. Vishny, 1998. Law and Finance. Journal of Political

Economy 106(6), 1113–1155.

, 2000. Agency Problems and Dividend Policies around the World. Journal of Finance 55(1),

1–33.

Larrain, B., M. Yogo, 2008. Does firm value move too much to be justified by subsequent changes in

cash flow. Journal of Financial Economics 87(1), 200 – 226.

Li, D., L. Zhang, 2009. Costly external finance: Implications for capital markets anomalies. Working

paper, University of Michigan.

Li, E. X. N., D. Livdan, L. Zhang, 2009. Anomalies. Review of Financial Studies 22(11), 4301–4334.

Lipson, M. L., S. Mortal, M. J. Schill, 2009. What Explains the Asset Growth Effect in Stock Returns?.

Darden Business School Working Paper.

Liu, L. X., T. M. Whited, L. Zhang, 2009. Investment-based expected stock returns. Journal of Political

Economy 117, 1105–1139.

Loughran, T., J. R. Ritter, 1995. The New Issues Puzzle. The Journal of Finance 50(1), 23–51.

Lyandres, E., L. Sun, L. Zhang, 2008. The New Issues Puzzle: Testing the Investment-Based Explana-

tion. Review of Financial Studies 21(6), 2825–2855.

McLean, R. D., J. Pontiff, A. Watanabe, 2009. Share issuance and cross-sectional returns: International

evidence. Journal of Financial Economics 94(1), 1 – 17.

Morck, R., M. Nakamura, 1999. Banks and Corporate Control in Japan. The Journal of Finance 54(1),

pp. 319–339.

Morck, R., B. Yeung, W. Yu, 2000. The information content of stock markets: why do emerging markets

have synchronous stock price movements?. Journal of Financial Economics 58(1-2), 215 – 260.

Pontiff, J., 1996. Costly Arbitrage: Evidence from Closed-End Funds. The Quarterly Journal of Eco-

nomics 111(4), 1135–1151.

Rau, P. R., T. Vermaelen, 1998. Glamour, value and the post-acquisition performance of acquiring

firms. Journal of Financial Economics 49(2), 223 – 253.

Richardson, S., R. Sloan, 2003. External Financing and Future Stock Returns. Working Paper, Wharton

School.

Roll, R., 1988. R2. The Journal of Finance 43(3), 541–566.

Rouwenhorst, K. G., 1998. International Momentum Strategies. The Journal of Finance 53(1), 267–284.

29

Page 32

Shumway, T., 1997. The Delisting Bias in CRSP Data. The Journal of Finance 52(1), 327–340.

Skaife, H. A., J. Gassen, R. LaFond, 2005. Does Stock Price Synchronicity Represent Firm-Specific

Information? The International Evidence. Working paper, University of Wisconsin Madison.

Stambaugh, R. F., 1999. Predictive regressions. Journal of Financial Economics 54(3), 375 – 421.

Teoh, S. H., I. Welch, T. J. Wong, 1998a. Earnings Management and the Long-Run Market Performance

of Initial Public Offerings. The Journal of Finance 53(6), 1935–1974.

, 1998b. Earnings management and the underperformance of seasoned equity offerings. Journal

of Financial Economics 50(1), 63 – 99.

Teoh, S. H., Y. G. Yang, Y. Zhang, 2009. R-Square and Market Efficiency. Working paper, UC Irvine.

Titman, S., K. C. J. Wei, F. Xie, 2004. Capital Investments and Stock Returns. The Journal of Financial

and Quantitative Analysis 39(4), 677–700.

Vuolteenaho, T., 2002. What Drives Firm-Level Stock Returns?. The Journal of Finance 57(1), pp.

233–264.

Wu, J. G., L. Zhang, X. F. Zhang, 2010. The q-Theory Approach to Understanding the Accrual

Anomaly. Journal of Accounting Research 48(1), pp. 177–223.

Xing, Y., 2008. Interpreting the Value Effect Through the Q-Theory: An Empirical Investigation.

Review of Financial Studies 21(4), 1767–1795.

Yao, T., T. Yu, T. Zhang, S. Chen, 2008. Asset Growth and Stock Returns: Evidence from the Pacific-

Basin Stock Markets. Pacific-Basin Finance Journal 19, 115–139.

30

Page 33

Appendix A: Country Characteristic Variables

• Stock return synchronicity (R2): the R-square of the regressing weekly returns of individual

stocks on market returns in a country. Following Jin and Myers (2006), Karolyi et al (2007), we

use the Dimson (1979) procedure to control for the non-synchronous trading effect by including

the lead and lag of market return in the market model. We run regression for each stock for every

year. Then we take the average of each firm’s R2 every year for a time series of an aggregate

measure for a country.

• Market capitalization to GDP ratio (MKT): the ratio of the total market capitalization of listed

companies of a country to its GDP. Data are from World Bank development index database from

1980 to 2006.

• Bank credit to market capitalization ratio (C/M): the ratio of domestic bank credit volume to

private sector to the total market capitalization of listed companies. Data are from World Bank

development index database from 1980 to 2006.

• UK: an indicator on a country’s English legal origin, equaling 1 for a country with English legal

origin. We obtain the data from Andrei Shleifer’s website.15English origin countries typically

are more effective in corporate governance (LLSV, 2000).

• FR: An indicator on a country’s French legal origin, equaling 1 for a country with French legal

origin. Relative to English origin countries, French origin countries are less effective in corporate

governance (LLSV, 2000).

• GE: an indicator on a country’s German legal origin, equaling 1 for a country with German legal

origin. Relative to English origin countries, German origin countries are less effective in corporate

governance (LLSV, 2000).

• SC: an indicator on a country’s Scandinavian legal origin, equaling 1 for a country with Scandi-

navian legal origin. Relative to English origin countries, Scandinavian origin countries are less

effective in corporate governance (LLSV, 2000).

• Revised Anti-Directed rights index (AD): Sum of 6 indices: Vote by mail, Shares not deposited,

Cumulative voting, Oppressed minority, Preemptive rights, and Capital to call meeting (Djankov,

La Porta, Lopez-Silanes, and Shleifer (DLLS), 2008)

• Anti-Self dealing index (AS): average of ex-ante and ex-post private control of self-dealing.

(Djankov, La Porta, Lopez-Silanes, and Shleifer (DLLS), 2008)

• Accounting standards (ACCOUNT): LLSV’s (1998) accounting index of accounting standards

where a higher value represents better accounting standards. The index is based on the reporting

or omission of 90 items from annual reports.

• Short-sale restriction (SHORT): a dummy variable equal to 1 if short selling is allowed and zero

if it is not allowed. We obtained this measure from Bris, Goetzmann, and Zhu (2007). Following

Mclean, Pontiff and Watanabe (2009), if short selling was legal prior to 1990, we assume that

15Andrei Sheifer’s website: http://www.economics.harvard.edu/faculty/shleifer/dataset.

31

Page 34

short selling was allowed in each of the years prior to 1990. Short sale constraints reduce the

ability of sophisticated traders to induce corrective price pressure.

• Stock trading turnover (TURN): measured by each firm’s average monthly share turnover during

the previous fiscal year. Turnover is the total dollar value of stocks traded, scaled by the value

of shares outstanding, for the period 1996–2000.

Appendix B: The Arellano-Bond GMM Procedure

Dynamic panel regression model can be written as

yi,t= αyi,t−1+ β⊤xi,t−1+ λt+ ηi+ εi,t

where yi,tis dependent variable, β is the slope coefficient, λtand ηiare time and individual specific

effects respectively, and εi,tdenotes the regression errors.

It is well known that the conventional Least Square Dummy Variable (LSDV) method is biased in

the above panel vector autoregressive model with firm-level individual effects. To see this, denote the

time mean of error terms εi,tas ¯ εi=?T

the strict exogeneity condition is violated when regressors include lagged dependent variables:

t=1εi,t. Simple within-group transformation would show that

?T

t=1E[yi,t−1(εi,t− ¯ εi)] ?= 0.

When the time dimension of the panel data T is small the biases will be very large regardless of the

size of the cross-section.

We use the Arellano-Bond GMM estimator to address this issue. Since common panel data estima-

tion procedure usually involves removing the individual specific effects first, Arellano-Bond estimator

corrects the bias by using a set of instruments in the regressions after this data transform. Arellano

and Bond (1991) show that for each period t = 3,...T, the all lagged variables yi,t−τ where τ ≥ 2

are good instruments and orthogonal to the regression error terms ε∗

example, first difference) is taken to remove individual specific effects. We take past three available

lagged endogenous variables as instruments in the differenced regressions. Specifically, our moment

conditions are

E[yi,t−τ⊗ ε∗

i,t)] = 0 for τ = 2,3,4; t = 3,...,T

i,twhere a data transform (for

where ⊗ denotes the Kronecker product and ε∗

taking first difference.

i,tis the residuals from the regressions on variables after

32

Page 35

Table 1: Sample Descriptive Statistics

CountryStart

Date

End

Date

Firm-

Month

Obs

% of

Total

Obs

No. of

Firms

% of

Total

Firm

Total % of Total Asset

Growth

Median

Asset

Growth

Stdev

Mkt Value

(US$M)

Total

Per MonthMkt Value

Argentina

Brazil

Chile

Mexico

Peru

South Africa

Israel

Greece

Poland

Portugal

Turkey

China

India

Indonesia

Malaysia

Pakistan

Philippines

South Korea

Taiwan

Thailand

Australia

New Zealand

Canada

Hong Kong

199807

199607

199307

199201

199507

198705

199907

199011

199807

199009

199507

199607

199407

199307

198607

199507

199307

199007

199507

199307

198207

199607

198207

198707

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

4,659

40,930

14,796

16,165

6,110

33,419

3,815

28,157

6,301

9,656

12,478

71,247

35,058

21,683

55,806

7,281

14,083

65,238

40,763

32,839

72,759

6,291

110,610

54,296

0.17

1.45

0.52

0.57

0.22

1.18

0.14

1.00

0.22

0.34

0.44

2.52

1.24

0.77

1.98

0.26

0.50

2.31

1.44

1.16

2.58

0.22

3.92

1.92

49

341

95

93

60

145

50

150

66

51

95

604

243

139

240

55

90

340

309

211

253

57

384

238

0.50

3.48

0.97

0.95

0.61

1.48

0.51

1.53

0.67

0.52

0.96

6.16

2.48

1.42

2.44

0.56

0.92

3.47

3.15

2.15

2.58

0.58

3.92

2.43

32,060

142,339

45,448

92,404

6,872

76,066

19,345

28,977

14,841

26,360

25,112

200,647

98,002

24,534

59,714

7,095

17,173

85,498

176,425

36,180

97,250

18,717

229,966

122,985

0.38

1.69

0.54

1.10

0.08

0.90

0.23

0.34

0.18

0.31

0.30

2.38

1.16

0.29

0.71

0.08

0.20

1.01

2.09

0.43

1.15

0.22

2.73

1.46

13.163

9.767

12.145

20.169

5.009

10.672

7.808

16.018

9.005

8.291

64.527

8.797

12.693

13.674

7.996

8.755

9.024

11.056

9.737

7.245

8.704

4.124

8.131

11.189

31.302

31.51

26.862

24.358

15.595

37.594

27.012

37.15

41.024

27.799

48.647

33.023

30.999

36.274

34.684

23.924

31.986

35.032

27.17

29.938

41.91

39.299

44.608

43.985

Continued on next page

33

Page 36

Table 1 – continued from previous page

CountryStart

Date

End

Date

Firm-

Month

Obs

% of

Total

Obs

No. of

Firms

% of

Total

Firm

Total % of TotalAsset

Growth

Median

Asset

Growth

Stdev

Mkt Value

(US$M)

Total

Per Month Mkt Value

Japan

Singapore

Austria

Belgium

Denmark

Finland

France

Germany

Ireland

Italy

Netherlands

Norway

Spain

Sweden

Switzerland

United Kingdom

United States

198207

198907

198910

198907

198907

198807

198207

198207

199107

198602

198207

198907

198907

198402

198207

198207

198207

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

200606

364,104

28,457

11,082

15,466

24,547

17,871

102,637

87,011

7,426

44,166

29,428

22,788

16,861

41,147

38,760

247,845

958,498

12.90

1.01

0.39

0.55

0.87

0.63

3.64

3.08

0.26

1.56

1.04

0.81

0.60

1.46

1.37

8.78

33.96

1,264

139

12.90

1.42

0.57

0.77

1.23

0.89

3.64

3.08

0.42

1.84

1.04

1.14

0.84

1.56

1.37

8.78

33.96

1,648,770

66,397

20,790

48,585

47,475

53,978

407,059

317,247

25,432

179,266

156,845

44,069

159,823

89,896

55,100

681,972

3,529,426

19.54

0.79

0.25

0.58

0.56

0.64

4.82

3.76

0.30

2.12

1.86

0.52

1.89

1.07

0.65

8.08

41.84

3.886

8.926

4.295

6.711

4.995

5.783

8.517

4.837

9.181

7.765

5.356

8.383

8.294

8.662

4.776

8.521

8.136

14.583

31.301

37.722

38.677

32.59

28.584

33.108

35.863

44.945

36.535

32.217

45.082

34.324

37.559

28.208

43.543

42.184

55

76

120

87

356

302

41

180

102

112

83

153

135

861

3,328

All

All excluding U.S.

2,822,534

1,864,036

100.00

66.04

9,800

6,472

100.00

66.04

8,436,488

4,907,063

100.00

58.16

7.144

6.792

37.647

34.632

This table provides summary statistics for the 41 countries in our study. We report the beginning and ending dates (Columns 2 and

3) on which each country is included into the sample, the total number of firm-month observations (Column 4), the average number

of firms per month (Column 6), and the average monthly total market capitalization (Column 8, in millions of U.S. dollars). Each

country’s firm-month observations, average number of firms and average market capitalization as percentages of the total sample

are also reported (Columns 5, 7, and 9). The last two columns (10 and 11) report the medians and standard deviations of the asset

growth rates for each country.

34

Page 37

Table 2: Asset Growth and Stock Returns: by Countries and Regions

Country

Africa - Emerging

South Africa

Region average

Asia-Developed

Hong Kong

Japan

Singapore

Region average

Asia-Emerging

China

Indonesia

India

South Korea

Malaysia

Philippines

Pakistan

Taiwan

Thailand

Region average

Australiasia-Developed

Australia

New Zealand

Region average

Europe-Developed

Germany

Belgium

Denmark

Spain

Finland

France

Ireland

Italy

Netherlands

Norway

Austria

Sweden

Switzerland

United Kingdom

AGSPREAD

t-stat SPREAD

t-stat STDSPRD

t-statSLOPE

t-stat VWSPRD

t-stat

118.31%

118.31%

5.78-4.95%

-4.95%

-1.21-4.47%

-4.47%

-0.79 -1.10%

-1.10%

-0.20-8.80%

-8.80%

-1.59

121.78%

45.53%

91.73%

86.35%

10.59

23.68

8.57

-1.05%

-0.03%

3.01%

0.64%

-0.32

-0.01

1.04

-1.61%

2.15%

3.39%

1.31%

-0.56

0.30

0.54

-1.35%

0.48%

4.68%

1.27%

-0.52

0.07

1.00

0.50%

0.40%

-2.10%

-0.40%

0.07

0.10

-0.44

101.80%

107.90%

102.51%

110.02%

96.31%

88.02%

54.16%

87.43%

97.86%

94.00%

6.59

12.00

16.36

12.66

10.22

10.83

13.74

10.09

10.12

-5.46%

-5.30%

-9.23%

-5.71%

3.50%

-4.54%

2.48%

7.11%

-5.61%

-2.53%

-1.18

-0.71

-1.93

-1.20

0.76

-0.64

0.60

0.86

-0.83

-10.55%

-3.26%

-8.73%

-5.36%

1.99%

-3.61%

2.12%

11.57%

-7.37%

-2.58%

-1.55

-0.41

-1.79

-1.16

0.30

-0.52

0.24

1.43

-0.93

-13.92%

-10.62%

-8.49%

-4.39%

-1.78%

-4.81%

-13.76%

12.00%

-10.91%

-6.30%

-1.77

-1.67

-1.38

-1.41

-0.41

-0.40

-1.55

1.00

-1.63

-4.90%

-4.20%

-3.20%

-4.50%

5.00%

20.00%

0.40%

14.30%

9.20%

3.60%

-1.13

-0.41

-0.32

-0.53

0.84

2.36

0.06

1.48

0.94

131.21%

78.32%

104.77%

7.45

6.37

-6.15%

5.07%

-0.54%

-1.53

1.50

-2.37%

8.37%

3.00%

-0.58

1.42

-2.04%

-0.24%

-1.14%

-0.62

-0.05

-10.10%

9.60%

-0.30%

-2.35

1.38

117.92%

84.31%

100.86%

76.81%

78.31%

102.25%

77.75%

109.98%

93.41%

136.60%

69.19%

116.62%

84.55%

147.26%

5.12

9.90

9.00

9.98

5.44

8.18

9.76

8.61

6.85

7.15

6.55

6.70

9.95

9.69

-5.39%

-2.37%

-7.87%

-0.55%

-5.35%

-5.34%

1.24%

-3.44%

-2.26%

-3.46%

0.96%

-5.38%

-5.40%

-9.66%

-1.88

-0.49

-1.79

-0.14

-0.83

-1.09

0.24

-0.86

-0.56

-0.89

0.19

-1.23

-1.14

-3.84

-10.24%

-0.75%

-11.76%

1.08%

-10.41%

-8.27%

1.07%

-1.95%

-4.32%

-0.15%

4.42%

-9.75%

7.79%

-9.39%

-1.93

-0.12

-1.88

0.18

-0.70

-1.37

0.14

-0.36

-0.84

-0.03

0.55

-1.85

0.60

-3.43

-3.60%

-0.78%

-13.68%

-1.61%

-13.59%

-5.52%

-2.51%

-2.27%

-0.08%

-1.05%

4.58%

-3.00%

4.48%

-9.25%

-0.82

-0.13

-2.37

-0.47

-0.98

-1.02

-0.40

-0.54

-0.02

-0.24

0.87

-0.72

0.50

-4.15

Continued on next page

-3.50%

-6.90%

-11.40%

3.50%

-9.00%

-15.50%

-6.60%

6.00%

-3.80%

3.00%

-5.30%

-12.10%

-1.50%

-9.80%

-0.57

-1.10

-2.20

0.59

-1.14

-2.15

-0.37

0.79

-0.61

0.50

-0.96

-1.74

-0.41

-2.44

35

Page 38

Table 2 – continued from previous page

Country

Region average

Europe-Emerging

Greece

Poland

Portugal

Turkey

Region average

Middle East - Emerging

Israel

Region average

North American

Canada

United States

Region average

South American -Emerging

Argentina

Brazil

Chile

Mexico

Peru

Region average

All markets pooled

All markets pooled, ex. U.S.

Country average

Emerging

Developed

Country average, ex. U.S.

AGSPREAD

99.70%

t-statSPREAD

-3.88%

t-statSTDSPRD

-3.76%

t-statSLOPE

-3.42%

t-stat VWSPRD

-5.20%

t-stat

113.21%

88.12%

59.03%

140.25%

100.15%

7.29

9.17

10.38

15.56

-4.47%

-3.27%

6.12%

4.27%

0.66%

-0.50

-0.26

1.05

0.93

-0.12%

-7.25%

13.33%

2.91%

2.22%

-0.01

-0.45

1.37

0.87

1.54%

-2.83%

9.47%

0.17%

2.09%

0.19

-0.24

1.19

0.05

-9.80%

1.70%

15.40%

-5.70%

0.40%

-0.47

0.28

1.73

-0.69

53.88%

53.88%

9.946.92%

6.92%

1.0811.19%

11.19%

1.136.75%

6.75%

0.8713.10%

13.10%

1.23

151.80%

141.24%

146.52%

10.23

10.79

-4.52%

-9.65%

-7.09%

-0.94

-3.25

-3.18%

-5.68%

-4.43%

-1.07

-3.04

-1.69%

-4.51%

-3.10%

-0.61

-2.22

-16.60%

-6.30%

-11.50%

-3.04

-2.39

56.01%

102.09%

68.39%

66.90%

36.69%

66.01%

99.61%

98.06%

95.28%

87.44%

102.73%

94.13%

4.54

13.72

10.10

9.42

12.10

5.81%

0.64%

2.53%

0.29%

-1.83%

1.49%

-2.68%

-2.42%

-1.91%

-0.54%

-3.22%

-1.72%

0.72

0.12

0.63

0.04

-0.20

-0.55%

0.60%

5.02%

4.99%

-9.26%

0.16%

-1.94%

-1.80%

-1.42%

-0.34%

-2.46%

-1.32%

-0.04

0.11

0.77

0.36

-0.36

1.14%

1.44%

6.84%

10.43%

-17.37%

0.49%

-2.18%

-2.10%

-2.26%

-2.01%

-2.50%

-2.21%

0.10

0.36

0.88

1.00

-1.10

-5.50%

13.90%

-1.80%

1.70%

1.90%

2.00%

-2.60%

-2.46%

-1.20%

2.41%

-4.64%

-1.08%

-0.39

1.73

-0.22

0.20

0.21

41.84

40.84

21.44

14.64

16.42

21.38

-3.43

-3.02

-2.62

-0.47

-3.77

-2.38

-1.63

-1.46

-1.39

-0.21

-1.88

-1.26

-2.15

-1.99

-2.11

-1.05

-2.32

-2.01

-2.12

-1.94

-0.89

1.20

-3.07

-0.78

This table reports various measures of the asset growth effect by country and by region. We sort stocks into deciles portfolios based on

asset growth rates in June of year t, where the asset growth rate is defined as the percentage changes of total assets from fiscal year t − 2

to fiscal year t−1. Portfolios are held over the 1-year holding period from July of year t to June of year t+1. Stock returns are the 1-year

buy-and-hold returns evaluated in local currencies. AGSPREAD is the time-series average of the differences in the asset growth rates

between the top and bottom asset growth deciles. SPREAD is the time-series averages of the difference in 1-year stock returns between the

top and bottom asset growth deciles. STDSPRD is the time series average of the 1-year stock return spreads between the top and bottom

asset growth deciles scaled by the difference in the asset growth rate between the two deciles. SLOPE is the time series average of the

coefficients on asset growth when we cross-sectionally regress 1-year buy-and-hold returns on asset growth. VWSPRD is the time-series

averages of the difference in the value-weighted 1-year returns between the top and bottom asset growth decile portfolios. We also pool

together firms across countries to compute measures of the asset growth effect in all markets, in all markets excluding the U.S., in all

emerging markets, and in all developed markets.

36

Page 39

Table 3: Robustness of the Asset Growth Effect: Alternative Horizons and Weighting, and

Controlling for Firm Characteristics

Panel A: All countries excluding U.S.

Year 1

EqualScaled

20.12***18.42***

(4.73)(3.33)

-2.99*** -1.36**

(-3.00) (-2.12)

-0.30-0.14

(-0.93) (-0.26)

2.65*** 3.46***

(3.07)(3.73)

6.66*** 6.15***

(3.93)(3.23)

21.3 26.18

Panel B: U.S.

Year 1

Equal Value

30.33*** 17.04***

(5.59) (4.04)

-8.78*** -4.19***

(-5.89) (-3.72)

-1.60** -0.13

(-2.01)(-0.24)

4.32*** 0.32

(3.34) (0.25)

-4.173.05

(-1.59)(0.71)

3.47 7.03

Horizon

Weighting

Intercept

1-Month

Equal

1.22***

(4.19)

-0.30***

(-4.93)

0.02

(0.51)

0.23***

(5.35)

0.96***

(6.41)

17.63

Year 2Year 3

Scaled

1.11***

(3.09)

-0.20**

(-1.96)

0.03

(0.89)

0.29***

(4.78)

0.88***

(4.11)

23.83

Equal

19.44***

(5.18)

-1.82

(-1.03)

-0.26

(-0.84)

2.76***

(2.94)

-3.51***

(-3.33)

21.49

Scaled

18.01***

(3.18)

0.03

(0.01)

-0.08

(-0.15)

3.75***

(5.83)

-4.87***

(-3.32)

26.63

Equal

17.86***

(5.83)

-2.21

(-0.90)

-0.07

(-0.29)

2.41***

(2.82)

-3.86***

(-3.11)

21.96

Scaled

17.29***

(3.40)

-2.19

(-0.81)

0.05

(0.09)

2.87***

(6.42)

-6.88***

(-4.87)

26.22

AG

ME

BM

MOM

Adj.R2

Horizon

Weighting

Intercept

1-Month

Equal

2.38***

(4.64)

-0.90***

(-8.85)

-0.11

(-1.60)

0.32***

(3.10)

0.02

(0.12)

2.12

Year 2 Year 3

Value

1.58***

(2.92)

-0.46***

(-3.12)

-0.03

(-0.52)

0.04

(0.34)

0.68*

(1.97)

6.04

Equal

23.79***

(4.94)

-3.34**

(-2.27)

-0.89

(-1.28)

3.57**

(2.40)

-3.08

(-1.47)

2.67

Value

13.63***

(3.06)

-2.75

(-1.38)

0.32

(0.53)

0.49

(0.35)

0.46

(0.16)

6.09

Equal

22.49***

(5.09)

-0.93

(-0.72)

-0.82

(-1.28)

1.30

(0.80)

0.54

(0.32)

2.16

Value

14.79***

(3.23)

2.96

(1.08)

0.08

(0.11)

-0.39

(-0.28)

0.30

(0.15)

5.46

AG

ME

BM

MOM

Adj. R2

This table reports the result of cross sectional regressions of individual stock returns onto asset growth

and various control variables. Panel A reports the regression results for all countries excluding the

U.S. and Panel B reports results for the U.S. The dependent variable, stock return, is measured at four

different holding horizons: 1-month, the first year, second year and third year after portfolio formation.

The control variables include ME (the natural logarithm of June-end market value), BM (the natural

logarithm of the previous year’s fiscal year-end book-to-market ratio), and MOM (the December-to-

May returns prior to return prediction months). The regressions use both equal weights and scaled

weights. The scaled weight (value weight for the U.S.) is a firm’s market value divided by the total

market capitalization of the country. The coefficients and the adjusted R2are expressed in percentage

points. The t-statistics are reported in the parentheses and significance at the 1%, 5%, and 10% level

is indicated by ***, **, and *, respectively.

37

Page 40

Table 4: Country Characteristics

Country

Africa-Emerging

South Africa

Region average

Asia-Developed

Hong Kong

Japan

Singapore

Region average

Asia-Emerging

China

Indonesia

India

South Korea

Malaysia

Philippines

Pakistan

Taiwan

Thailand

Region average

Australiasia-Developed

Australia

New Zealand

Region average

Europe-Developed

Germany

Belgium

Denmark

Spain

Finland

France

Ireland

Italy

Netherlands

Norway

Austria

Sweden

Switzerland

United Kingdom

Region average

ββr

βcf

βcf,r

UK FRGESCR2MKT C/MAD ASACCOUNT SHORT TURN

-0.143

-0.143

-0.056

-0.056

-0.108

-0.108

0.021

0.021

1

1

0

0

0

0

0

0

0.14

0.14

5.02

5.02

2.12

2.12

5

5

0.81

0.81

70

70

1

1

27.74

27.74

-0.253

0.044

0.098

-0.037

-0.095

0.010

0.042

-0.015

-0.221

0.039

0.086

-0.032

0.063

-0.005

-0.03

0.01

1

0

1

0

0

0

0

1

0

0

0

0

0.20

0.27

0.26

0.24

5.57

4.38

5.04

5.00

5.16

2.32

6.79

4.75

.. 69

65

78

1

1

0

64.94

49.40

49.62

54.65

4.5

5

4.75

0.5

1

0.75 0.670.000.33 0.0070.670.67

-0.904

-0.010

-0.048

0.692

-0.031

-0.094

-0.140

0.082

-0.068

-0.058

-0.181

-0.018

-0.088

0.112

-0.007

-0.092

-0.055

0.006

-0.060

-0.042

-0.992

-0.006

-0.066

0.651

-0.030

-0.063

-0.152

0.082

-0.044

-0.069

0.269

0.014

0.106

-0.071

0.006

0.061

0.067

-0.006

0.036

0.053

0

1

0

1

1

0

0

0

1

0

0

1

0

0

1

0

0

0

1

0

0

0

0

0

1

1

0

0

0

0

0

0

0

0

0

0

0.14

0.20

0.16

0.28

0.16

0.17

0.20

0.33

0.20

0.20

2.84

3.37

2.63

5.02

2.68

3.71

3.75

.

3.80

3.48

10.06

5.94

5.17

6.14

5.06

2.93

6.84

.

5.99

6.02

1

5

4

5

4

4

.

.

4

0.76

0.58

0.65

0.95

0.41

0.22

.

.

0.81

0.63

.0

1

0

1

0

1

0

1

1

10.00

63.72

65.20

50.80

220.97

39.52

204.92

314.74

62.51

114.71

57

.

76

.

65

62

65

64

0.44 0.220.33 0.003.8664.83 0.56

0.060

-0.112

-0.026

0.027

-0.067

-0.020

0.045

-0.110

-0.033

-0.012

0.065

0.027

1

1

1

0

0

0

0

0

0

0

0

0

0.11

0.15

0.13

4.30

3.69

3.99

3.26

2.54

2.90

4

4

4

0.76

0.95

0.855

75

70

72.5

1

1

1

48.16

42.68

45.42

-0.250

0.078

-0.127

-0.028

-0.028

-0.035

-0.013

-0.031

0.056

0.566

0.246

-0.189

0.483

0.013

0.053

-0.117

0.030

-0.078

-0.009

-0.079

-0.007

-0.011

-0.021

0.030

0.038

0.051

-0.034

0.085

0.006

-0.008

-0.187

0.037

-0.067

-0.022

0.005

-0.035

-0.003

-0.020

0.036

0.554

0.199

-0.189

0.393

0.011

0.051

0.054

0.011

0.018

0.003

0.046

0.007

0.001

0.01

-0.01

-0.026

-0.004

0.034

0.005

-0.004

0.01

0

0

0

0

0

0

1

0

0

0

0

0

0

1

0

1

0

0

1

0

0

1

1

0

1

0

0

0

1

0

0

0

0

1

0

0

0

0

0

0

1

0

0

0

1

1

0

0

0

0

0

1

0

1

0

0

0.19

0.15

0.13

0.19

0.14

0.15

0.14

0.24

0.16

0.18

0.24

0.18

0.14

0.13

0.17

2.73

3.95

3.78

4.05

3.92

3.51

4.11

3.32

4.43

3.48

3.84

4.36

5.06

4.85

3.96

2.35

2.10

2.31

2.71

2.13

2.00

5.47

1.78

2.55

3.00

3.11

2.32

1.58

2.69

2.58

2.5

3

4

3.5

3.5

3.5

5

2

2.5

3.5

5

3.5

3

5

3.54

0.21

0.54

0.46

0.46

0.38

0.28

0.79

0.42

0.2

0.42

0.37

0.33

0.27

0.95

0.43

54

61

62

77

69

62

.

62

64

74

64

83

68

78

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

49.52

24.68

64.23

51.34

63.82

118.46

59.96

80.08

95.36

79.64

174.82

78.08

91.22

50.62

77.270.140.360.210.2967.54

Continued on next page

38

Page 41

Table 4 – continued from previous page

Country

Europe-Emerging

Greece

Poland

Portugal

Turkey

Region average

Middle East - Emerging

Israel

Region average

North American

Canada

United States

Region average

South American-Emerging

Argentina

Brazil

Chile

Mexico

Peru

Region average

Cross-sectional mean

Cross-sectional stdev

ββr

βcf

βcf,r

UK FRGE SC R2 MKT C/M ADAS ACCOUNT SHORTTURN

0.045

0.357

-0.055

-0.550

-0.050

0.041

0.032

-0.020

-0.128

-0.019

0.026

0.366

-0.038

-0.498

-0.036

-0.022

-0.041

0.003

0.076

0.005

0

0

0

0

0

1

0

1

1

1

0

1

0

0

0

0

0

0

0

0

0.26

0.25

0.14

0.28

0.23

3.18

1.95

3.23

2.87

2.81

2.33

3.77

2.62

4.53

3.31

2

2

0.22

0.29

0.44

0.43

0.35

55

.

36

51

47

0

1

1

1

78.79

.

68.18

148.00

98.32

2.5

3

2.38 0.75

0.833

0.833

0.166

0.166

0.545

0.545

0.122

0.122

1

1

0

0

0

0

0

0

0.17

0.17

3.68

3.68

4.13

4.13

4

4

0.73

0.73

64

64

1

1

.

.

-0.593

-0.319

-0.456

-0.125

-0.054

-0.090

-0.613

-0.308

-0.460

0.145

0.043

0.094

1

1

1

0

0

0

0

0

0

0

0

0

0.11

0.15

0.13

4.33

4.63

4.48

2.80

3.13

2.96

4

3

0.64

0.65

0.65

74

71

72.5

1

1

1

65.64

125.30

95.473.5

0.043

0.068

0.051

-0.094

0.023

0.018

-0.007

0.308

0.042

0.032

0.024

-0.052

0.014

0.012

-0.016

0.069

0.021

0.051

0.043

-0.073

0.025

0.013

-0.015

0.286

-0.02

-0.015

-0.016

0.031

-0.016

-0.007

0.025

0.059

0

0

0

0

0

0

1

1

1

1

1

1

0

0

0

0

0

0

0

0

0

0

0

0

0.23

0.13

0.16

0.20

0.12

0.17

0.18

0.05

2.87

3.14

4.36

3.19

2.83

3.28

3.79

0.82

2.36

2.54

4.98

2.36

2.29

2.90

3.61

1.83

2

5

4

3

0.34

0.27

0.63

0.17

0.45

0.37

0.52

0.24

45

54

52

60

38

49.8

63.72

10.80

1

1

1

1

0

14.42

61.14

10.00

34.42

21.65

28.33

77.55

61.82

3.5

3.5

3.61

1.05

0.80

0.83

0.38

0.34

0.48

0.37

0.49

0.20

0.40

0.10

0.30

This table reports the country characteristic variables. β, βr, βcfand βcf,rare the overall beta, discount rate beta, cash flow beta,

and interaction beta estimated from the investment-investment return relation (following the procedure described in Section 3). UK, FR,

GE, and SC are dummy variables for a country’s legal origin (English, French, German, and Scandinavian, respectively). R2 is stock

return synchronicity, measured by the average R-square of the market model for stock returns in each country. MKT is the equity market

capitalization to GDP ratio (in percentage point). C/M is the domestic bank credit volume to equity market capitalization ratio. AD is the

revised Anti-Directed rights index, and AS is the anti-self dealing index. SHORT is a dummy variable equal to 1 if short-selling is allowed

and zero if not allowed. TURN is the average monthly stock trading during the previous year.

39

Page 42

Table 5: Correlations of Country Characteristics

βcf,r

R2MKTC/MUKVariablesββcf

βr

FR GESC ADASACCOUNTSHORTTURN

β 1.00

βcf

0.98

(0.00)

0.89

(0.00)

-0.60

(0.00)

0.15

(0.00)

0.09

(0.01)

-0.03

(0.30)

0.05

(0.10)

-0.02

(0.48)

-0.09

(0.00)

0.07

(0.02)

0.30

(0.00)

0.01

(0.75)

0.11

(0.00)

0.19

(0.00)

0.053

(0.10)

1.00

βr

0.85

(0.00)

-0.69

(0.00)

0.17

(0.00)

0.09

(0.01)

-0.04

(0.21)

0.03

(0.33)

-0.01

(0.88)

-0.11

(0.00)

0.10

(0.00)

0.33

(0.00)

-0.02

(0.43)

0.10

(0.00)

0.22

(0.00)

0.056

(0.08)

1.00

βcf,r

-0.63

(0.00)

0.15

(0.00)

0.13

(0.00)

-0.06

(0.10)

0.07

(0.02)

0.03

(0.38)

-0.11

(0.00)

-0.02

(0.53)

0.24

(0.00)

0.03

(0.36)

0.02

(0.44)

0.11

(0.00)

0.059

(0.07)

1.00

R2-0.21

(0.00)

-0.12

(0.00)

0.09

(0.01)

0.03

(0.25)

-0.12

(0.00)

0.19

(0.00)

-0.10

(0.00)

-0.30

(0.00)

0.14

(0.00)

0.07

(0.02)

-0.22

(0.00)

-0.074

(0.02)

1.00

MKT-0.15

(0.00)

0.21

(0.00)

-0.13

(0.00)

0.04

(0.18)

0.11

(0.00)

-0.03

(0.45)

-0.03

(0.33)

-0.08

(0.03)

-0.01

(0.80)

-0.02

(0.42)

0.266

(0.00)

1.00

C/M-0.50

(0.00)

0.35

(0.00)

-0.27

(0.00)

-0.12

(0.00)

0.04

(0.28)

0.37

(0.00)

0.35

(0.00)

0.54

(0.00)

0.22

(0.00)

-0.098

(0.01)

1.00

UK -0.10

(0.01)

-0.02

(0.59)

0.19

(0.00)

-0.04

(0.29)

-0.17

(0.00)

-0.06

(0.10)

-0.25

(0.00)

-0.06

(0.06)

-0.027

(0.47)

1.00

FR-0.56

(0.00)

-0.35

(0.00)

-0.24

(0.00)

0.54

(0.00)

0.76

(0.00)

0.45

(0.00)

0.05

(0.06)

-0.053

(0.10)

1.00

GE-0.36

(0.00)

-0.25

(0.00)

-0.28

(0.00)

-0.47

(0.00)

-0.62

(0.00)

-0.05

(0.08)

-0.14

(0.00)

1.00

SC-0.15

(0.00)

-0.34

(0.00)

-0.23

(0.00)

-0.04

(0.16)

-0.12

(0.00)

0.29

(0.00)

1.00

AD0.00

(0.96)

-0.15

(0.00)

0.34

(0.00)

0.15

(0.00)

-0.05

(0.16)

1.00

AS0.47

(0.00)

0.41

(0.00)

0.15

(0.00)

0.07

(0.03)

1.00

ACCOUNT 0.41

(0.00)

-0.11

(0.00)

-0.27

(0.00)

1.00

SHORT0.18

(0.00)

0.06

(0.06)

1.00

TURN -0.13

(0.00)

1.00

This table reports cross-country correlations of characteristic variables. β, βr, βcfand βcf,rare the overall beta, discount rate beta, cash flow beta, and

interaction beta estimated from the investment-investment return relation (following the procedure described in Section 3). UK, FR, GE, and SC are dummy

variables for a country’s legal origin (English, French, German, and Scandinavian, respectively). R2 is stock return synchronicity, measured by the average

R-square of the market model for stock returns in each country. MKT is the equity market capitalization to GDP ratio (in percentage point). C/M is the

domestic bank credit volume to equity market capitalization ratio. AD is the revised Anti-Directed rights index, and AS is the anti-self dealing index. SHORT

is a dummy variable equal to 1 if short-selling is allowed and zero if not allowed. TURN is the average monthly stock trading during the previous year. The

p-values for the pairwise correlations are reported in the parentheses.

40

Page 43

Table 6: βs and the Asset Growth Effect: Cross-Country Regressions

Panel A: STDSPRD as Dependent Variable

Intercept -0.014

(-1.38)

β 0.074**

(2.57)

βcf

-0.013

(-1.34)

-0.009

(-0.89)

-0.011

(-0.99)

0.072**

(2.36)

βr

0.326**

(2.53)

βcf,r

-0.136

(-0.70)

Panel B: SLOPE as Dependent Variable

Intercept -0.022**

(-2.20)

β 0.063**

(2.38)

βcf

-0.022**

(-2.14)

-0.017*

(-1.72)

-0.017

(-1.54)

0.065**

(2.41)

βr

0.319**

(2.45)

βcf,r

-0.241

(-1.27)

Panel C: VWSPRD as Dependent Variable

Intercept-0.011

(-0.88)

β0.102***

(3.45)

βcf

0.107***

-0.010

(-0.82)

-0.007

(-0.52)

-0.008

(-0.63)

(2.96)

βr

0.328*

(1.85)

βcf,r

-0.160

(-0.62)

This table reports the results of cross-country regression of the asset growth effects. The dependent

variable is one of the three measures of the asset growth effect: STDSPRD, SLOPE, and VWSPRD.

The dependent variable is one of the four betas: β, βcf, βrand βcf,r. Bootstrap standard errors are

used to compute the t-statistics (in parentheses) and significance at the 1%, 5%, and 10% level is

indicated by ***, **, and *, respectively.

41

Page 44

Table 7: βs and the Asset Growth Effect: Subsample Regressions

Panel A: STDSPRD as Dependent Variable

βcfas Explanatory Variable

R2

lowhigh

-0.011 -0.011-0.013

(-0.73)(-0.81)(-0.84)

0.113** 0.014

(2.56)(0.27)

2219

0.218 0.004

Panel B: SLOPE as Dependent Variable

βcfas Explanatory Variable

R2

lowhigh

-0.029** -0.008-0.030*

(-1.99) (-0.55)(-1.91)

0.102**-0.003

(2.30) (-0.10)

2219

0.177 0.000

Panel C: VWSPRD as Dependent Variable

βcfas Explanatory Variable

R2

low high

-0.011-0.010

(-0.54)(-0.66)

0.108 0.104***0.089**

(1.49) (2.68)

2219

0.101 0.173

β as Explanatory Variable

MKT

low

-0.011-0.013

(-0.78)(-0.87)

0.0110.058

(0.26) (1.10)

19 22

0.0030.076

βras Explanatory Variable

MKT

highlow

-0.010 -0.007

(-0.75)(-0.47)

0.053 0.259

(0.26) (1.20)

19 22

0.003 0.074

R2 C/M MKTC/M R2C/M

low

-0.013

(-0.93)

0.115***

(3.66)

22

0.268

high high

-0.016

(-1.21)

0.102**

(2.21)

19

0.218

low high

0.005

(0.22)

0.069

(0.42)

12

0.074

lowhigh

-0.015

(-1.22)

0.104*

(1.72)

19

0.198

lowhigh

0.006

(0.26)

0.073

(0.46)

12

0.095

low

-0.007

(-0.52)

0.488***

(3.52)

22

0.254

high

-0.013

(-1.03)

0.479***

(2.88)

19

0.230

low

-0.016

(-1.47)

0.368**

(2.28)

29

0.176

high

0.008

(0.36)

0.278

(0.68)

12

0.063

Intercept -0.022**

(-1.99)

0.082**

(2.21)

29

0.176

-0.021**

(-2.01)

0.080*

(1.90)

29

0.128

β 0.055

(0.90)

22

0.057

Obs

R2

β as Explanatory Variable

MKT

low

-0.008-0.031*

(-0.56) (-1.93)

-0.004 0.069

(-0.13) (1.45)

1922

0.0000.088

βras Explanatory Variable

MKT

highlow

-0.008-0.024

(-0.60) (-1.39)

0.019 0.321

(0.10)(1.59)

19 22

0.0000.095

R2C/MMKT C/M R2C/M

low high high

-0.012

(-0.91)

0.047

(1.29)

19

0.054

lowhigh

-0.007

(-0.34)

0.113

(0.70)

12

0.166

lowhigh

-0.011

(-0.85)

0.046

(0.94)

19

0.046

low high

-0.007

(-0.31)

0.114

(0.61)

12

0.193

lowhigh

-0.010

(-0.78)

0.258*

(1.78)

19

0.078

lowhigh

-0.001

(-0.06)

0.500

(1.37)

12

0.168

Intercept-0.031**

(-2.24)

0.101***

(2.69)

22

0.206

-0.027**

(-2.29)

0.049*

(1.79)

29

0.060

-0.026**

(-2.24)

0.044

(1.35)

29

0.035

-0.025*

(-1.72)

0.479***

(3.53)

22

0.245

-0.023**

(-1.96)

0.266*

(1.90)

29

0.087

β 0.072

(1.38)

22

0.082

Obs

R2

β as Explanatory Variable

MKT

low

-0.009 0.020

(-0.60)(1.13)

0.093***0.088***

(2.96)(2.93)

19 22

0.1590.119

βras Explanatory Variable

MKT

high low

-0.0040.026

(-0.25)(1.39)

0.2100.301

(1.11)(1.17)

19 22

0.038 0.068

R2C/MMKT C/MR2C/M

low

-0.013

(-0.63)

0.107*

(1.87)

22

0.120

high highlow high

0.026

(1.03)

0.090

(0.56)

12

0.099

low

0.020

(1.17)

high low high

0.026

(1.05)

0.084

(0.45)

12

0.098

low

-0.009

(-0.46)

0.393

(1.31)

22

0.084

highlow

-0.021

(-1.56)

0.328

(1.22)

29

0.073

high

0.032

(1.21)

0.453

(0.98)

12

0.130

Intercept-0.049***

(-3.22)

0.150***

(3.44)

19

0.306

-0.027*

(-1.90)

0.120***

(3.60)

29

0.198

-0.047***

(-3.29)

0.162***

(3.40)

19

0.311

-0.026*

(-1.92)

0.134***

(3.36)

29

0.189

-0.044***

(-3.08)

0.670***

(3.02)

19

0.290

β

(2.22)

22

0.103

Obs

R2

This table reports the results of cross-country regressions on the asset growth effect within country subsamples. The dependent variable is one of the three measures of the asset growth effect, STDSPRD,

SLOPE, and VWSPRD. The explanatory variable is one of the three betas, β, βcf, and βr. Countries are classified into two subsamples based on whether a market efficiency indicator is below (low) or above

(high) the cross-country mean. The three market efficiency indicators are stock return synchronicity (R2), market capitalization to GDP ratio (MKT), and domestic bank credit volume to market capitalization

ratio (C/M). Bootstrap standard errors are used to compute the t-statistics (reported in parentheses) and significance at the 1%, 5%, and 10% level is indicated by ***, **, and *, respectively.

42

Page 45

Table 8: Cross-country Analysis of STDSPRD: βs with Control Variables

Panel A: β as Explanatory Variable

coefft-stat

-0.086***(-2.77)

0.076** (2.32)

0.071** (2.27)

0.083*** (3.45)

0.066* (1.77)

coefft-stat

(-4.71)

(2.63)

(3.16)

(-3.58)

(2.36)

coeff

-0.061

0.081**

0.077***

0.081***

0.060*

-0.006

t-stat

(-1.50)

(2.50)

(2.98)

(3.39)

(1.65)

(-0.62)

coeff

0.000

0.085**

0.065**

0.064*

0.086*

t-stat

(0.00)

(2.49)

(2.27)

(1.74)

(1.96)

coefft-stat

(-3.71)

(2.52)

(3.35)

(3.81)

(2.41)

coefft-stat

(-4.52)

(2.50)

(3.06)

(3.54)

(2.47)

Intercept

β

UK

FR

GE

AD

AS

ACCOUNT

SHORT

TURN

-0.085***

0.080***

0.073***

0.083***

0.079**

-0.106***

0.075**

0.076***

0.087***

0.084**

-0.098***

0.073**

0.067***

0.083***

0.081**

0.004 (0.07)

-0.001 (-0.63)

0.021(0.865)

0.199 (1.06)

Panel B: βcfas Explanatory Variable

-0.088***(-2.93)

0.076** (2.18)

0.072**(2.34)

0.085*** (3.40)

0.068* (1.82)

Intercept

βcf

UK

FR

GE

AD

AS

ACCOUNT

SHORT

TURN

-0.086***

0.081**

0.076***

0.085***

0.081**

(-4.24)

(2.39)

(3.01)

(3.37)

(2.43)

-0.060

0.082**

0.080***

0.082***

0.063*

-0.007

(-1.35)

(2.24)

(3.12)

(3.50)

(1.67)

(-0.62)

-0.00

0.09**

0.068**

0.066*

0.089**

(-0.02)

(2.24)

(2.43)

(1.83)

(1.98)

-0.106***

0.075**

0.078***

0.088***

0.086**

(-3.37)

(2.31)

(2.98)

(3.51)

(2.34)

-0.099***

0.077**

0.069***

0.085***

0.085**

(-4.30)

(2.56)

(2.84)

(3.57)

(2.43)

0.007(0.12)

-0.001 (-0.63)

0.02 (0.82)

0.193(1.11)

Panel C: βras Explanatory Variable

-0.075** (-2.19)

0.306** (1.99)

0.064* (1.90)

0.076*** (2.70)

0.06 (1.50)

Intercept

βr

UK

FR

GE

AD

AS

ACCOUNT

SHORT

TURN

-0.073***

0.322**

0.066**

0.075***

0.073**

(-3.25)

(2.373)

(2.44)

(2.72)

(2.05)

-0.062

0.313**

0.067**

0.074***

0.057

-0.003

(-1.42)

(2.20)

(2.36)

(2.83)

(1.43)

(-0.31)

0.004

0.300**

0.058*

0.056

0.083*

(-0.03)

(1.99)

(1.89)

(1.59)

(1.86)

-0.100***

0.306**

0.070**

0.081***

0.080**

(-2.87)

(2.30)

(2.53)

(2.92)

(2.11)

-0.089***

0.256*

0.060**

0.076***

0.074*

(-3.44)

(1.94)

(2.28)

(2.89)

(1.95)

0.003(0.04)

-0.001 (-0.61)

0.026(1.07)

0.204 (1.10)

This table reports the results of cross-country regressions with STDSPRD as the dependent variable. The explanatory variables include one of the three betas, β

(Panel A), βcf(Panel B), βr(Panel C), and various country characteristics: legal origin dummies UK, FR, and GE (SC dropped to avoid multi-collinearity), the

Revised Anti-Director rights index (AD), the Anti-Self Dealing index (AS), accounting standard index (ACCOUNT), the short-selling dummy, and stock trading

turnover (TURN). Bootstrap standard errors are used to compute the t-statistics (reported in parentheses) and significance at the 1%, 5%, and 10% level is indicated

by ***, **, and *, respectively.

43

Page 46

Table 9: Cross-country Analysis of SLOPE: βs with Control Variables

Panel A: β as Explanatory Variable

coefft-stat

-0.064(-1.34)

0.066** (2.09)

0.061(1.27)

0.074** (2.05)

0.066* (1.81)

coefft-stat

(-2.71)

(2.25)

(1.42)

(2.08)

(2.09)

coeff

-0.058

0.073**

0.051

0.073**

0.062

-0.007

t-stat

(-1.20)

(2.34)

(1.45)

(2.06)

(1.64)

(-0.62)

coeff

-0.129

0.047

0.058*

0.093**

0.109**

t-stat

(-0.83)

(1.42)

(1.68)

(2.00)

(2.38)

coefft-stat

(-3.08)

(1.95)

(1.68)

(2.52)

(2.44)

coefft-stat

(-2.38)

(1.70)

(1.24)

(2.11)

(2.09)

Intercept

β

UK

FR

GE

AD

AS

ACCOUNT

SHORT

TURN

-0.085***

0.070**

0.048

0.076**

0.081**

-0.152***

0.055*

0.058*

0.088**

0.097**

-0.087**

0.064*

0.044

0.075**

0.084**

-0.043(-0.48)

0.001(0.32)

0.070*(1.84)

0.078(0.31)

Panel B: with βcfas Explanatory Variable

(-1.17) -0.067 (-1.34)

(2.15)0.068* (1.84)

(1.56) 0.062(1.26)

(1.98) 0.075**(2.05)

(1.67)0.068* (1.81)

(-0.74)

-0.04 (-0.44)

Intercept

βcf

UK

FR

GE

AD

AS

ACCOUNT

SHORT

TURN

-0.083**

0.075**

0.051

0.077**

0.084**

(-2.56)

(2.03)

(1.43)

(2.04)

(2.13)

-0.056

0.078**

0.054

0.075**

0.064*

-0.008

-0.131

0.047

0.059

0.094**

0.111**

(-0.85)

(1.27)

(1.64)

(2.01)

(2.38)

-0.152***

0.055*

0.059*

0.089**

0.099**

(-3.05)

(1.67)

(1.65)

(2.44)

(2.55)

-0.088**

0.070*

0.046

0.077**

0.087**

(-2.43)

(1.71)

(1.30)

(2.01)

(2.03)

0.001 (0.33)

0.070* (1.90)

0.071 (0.28)

Panel C: βras Explanatory Variable

-0.054 (-1.17)

0.317**(2.28)

0.053 (1.06)

0.067*(1.92)

0.062 (1.64)

Intercept

βr

UK

FR

GE

AD

AS

ACCOUNT

SHORT

TURN

-0.071**

0.328**

0.042

0.068*

0.077*

(-2.27)

(2.32)

(1.23)

(1.89)

(1.95)

-0.055

0.337**

0.041

0.066*

0.059

-0.005

(-1.17)

(2.41)

(1.19)

(1.81)

(1.47)

(-0.46)

-0.125

0.207

0.054

0.088*

0.107**

(-0.80)

(1.53)

(1.47)

(1.91)

(2.21)

-0.145***

0.284**

0.053

0.083**

0.095**

(-2.99)

(2.42)

(1.51)

(2.36)

(2.41)

-0.077**

0.287*

0.038

0.069*

0.079*

(-2.11)

(1.92)

(1.07)

(1.87)

(1.78)

-0.042 (-0.48)

0.001(0.34)

0.073**(2.08)

0.075 (0.29)

This table reports the results of cross-country regressions with SLOPE as the dependent variable. The explanatory variables include one of the three betas, β

(Panel A), βcf(Panel B), βr(Panel C), and various country characteristics: legal origin dummies UK, FR, and GE (SC dropped to avoid multi-collinearity),

the Revised Anti-Director rights index (AD), the Anti-Self Dealing index (AS), accounting standard index (ACCOUNT), the short-selling dummy, and stock

trading turnover (TURN). Bootstrap standard errors are used to compute the t-statistics (reported in parentheses) and significance at the 1%, 5%, and 10%

level is indicated by ***, **, and *, respectively.

44

Page 47

Table 10: Cross-country Analysis of VWSPRD: βs with Control Variables

Panel A: β as an Explanatory Variable

coefft-stat

-0.081**(-2.04)

0.108*** (2.96)

0.054(1.45)

0.088*** (2.74)

0.063**(2.27)

coefft-stat

(-3.70)

(3.58)

(2.07)

(2.75)

(2.53)

coeff

-0.091

0.105**

0.053

0.089***

0.065**

0.003

t-stat

(-1.38)

(2.50)

(1.61)

(2.76)

(2.02)

(0.18)

coeff

0.111

t-stat

(-0.73)

(3.58)

(1.70)

(0.79)

(1.02)

coefft-stat

(-2.79)

(3.09)

(2.11)

(2.58)

(2.59)

coefft-stat

(-3.48)

(2.30)

(1.73)

(2.64)

(2.22)

Intercept

β

UK

FR

GE

AD

AS

ACCOUNT

SHORT

TURN

-0.079***

0.109***

0.060**

0.088***

0.080**

-0.100***

0.105***

0.063**

0.092***

0.085***

-0.094***

0.089**

0.052*

0.087***

0.068**

0.134***

0.053*

0.045

0.047

0.004 (0.05)

-0.003 (-1.28)

0.021(0.70)

0.229(1.10)

Panel B: βcfas an Explanatory Variable

-0.085**(-2.12)

0.113*** (2.69)

0.056(1.40)

0.090*** (2.82)

0.067** (2.25)

Intercept

βcf

UK

FR

GE

AD

AS

ACCOUNT

SHORT

TURN

-0.081***

0.116***

0.064**

0.090***

0.084***

(-4.24)

(3.32)

(2.39)

(2.93)

(2.78)

-0.088

0.111**

0.058*

0.090***

0.069**

0.002

(-1.28)

(2.37)

(1.78)

(2.58)

(2.28)

(0.10)

0.109 (0.70)

(3.97)

(1.86)

(0.86)

(1.12)

-0.100***

0.111***

0.066**

0.093***

0.088***

(-2.72)

(2.81)

(2.21)

(2.59)

(2.62)

-0.095***

0.097**

0.054*

0.089***

0.073**

(-3.65)

(2.15)

(1.87)

(2.63)

(2.37)

0.158***

0.060*

0.048

0.051

0.01(0.11)

-0.003 (-1.26)

0.019 (0.64)

0.220(1.06)

Panel C: βras an Explanatory Variable

-0.066(-1.41)

0.324(1.39)

0.048(0.99)

0.077**(2.02)

0.054(1.61)

Intercept

βr

UK

FR

GE

AD

AS

ACCOUNT

SHORT

TURN

-0.067** (-2.18)

(1.60)

(1.39)

(1.91)

(1.90)

-0.101

0.301

0.041

0.081**

0.062*

0.009

(-1.42)

(1.30)

(1.04)

(1.99)

(1.71)

(0.56)

0.109

0.342

0.045

0.035

0.043

(-0.66)

(1.32)

(1.17)

(0.55)

(0.84)

-0.097**

0.311

0.056

0.083**

0.079**

(-2.20)

(1.49)

(1.38)

(1.96)

(2.01)

-0.087**

0.188

0.043

0.079*

0.056

(-2.42)

(0.81)

(1.12)

(1.88)

(1.51)

0.33

0.051

0.077*

0.072*

-0.002(-0.02)

-0.002 (-1.11)

0.03(1.04)

0.251(1.14)

This table reports the results of cross-country regressions with VWSPRD as the dependent variable. The explanatory variables include one of the three betas, β

(Panel A), βcf(Panel B), βr(Panel C), and various country characteristics: legal origin dummies UK, FR, and GE (SC dropped to avoid multi-collinearity), the

Revised Anti-Director rights index (AD), the Anti-Self Dealing index (AS), accounting standard index (ACCOUNT), the short-selling dummy, and stock trading

turnover (TURN). Bootstrap standard errors are used to compute the t-statistics (reported in parentheses) and significance at the 1%, 5%, and 10% level is indicated

by ***, **, and *, respectively.

45

Page 48

Figure 1: The Asset Growth Effect Across Country Groups

Panel A: STDSPRD

Panel B: SLOPE

Panel C: VWSPRD

low R2

high R2

low Mkthigh Mkt

low C/Mhigh C/M

low R2 high R2low Mkt

high Mkt

low C/Mhigh C/M

low R2high R2 low Mkthigh Mkt

low C/M high C/M

-0.04

-0.02

0

0.02

-0.04

-0.02

0

-0.03

-0.02

-0.01

0

0.01

This figure compares the magnitude of the asset growth effect across different country groups. Coun-

tries are classified into two groups each time based on one of the three market efficiency indicators:

stock return synchronicity (R2), market capitalization to GDP ratio (MKT), bank credit to market

capitalization ratio (C/M). The magnitude of the asset growth effect is measured by STDSPRD (Panel

A), SLOPE (Panel B), and VWSPRD (Panel C).

46