Limits‐to‐arbitrage, investment frictions, and the investment effect: New evidence

DOIhttp://doi.org/10.1111/eufm.12216
Date01 January 2020
Published date01 January 2020
DOI: 10.1111/eufm.12216
EUROPEAN
FINANCIAL MANAGEMENT
ORIGINAL ARTICLE
Limitstoarbitrage, investment frictions, and
the investment effect: New evidence
F. Y. Eric C. Lam
1
|
Ya Li
2
|
Wikrom Prombutr
3
|
K. C. John Wei
4
1
Hong Kong Institute for Monetary
Research, Hong Kong Monetary
Authority, Central, Hong Kong
Email: cfylam@hkma.gov.hk
2
LSK School of Business and
Administration, The Open University of
Hong Kong, Ho Man Tin, Hong Kong
Email: yali@ouhk.edu.hk
3
Department of Finance, College of
Business Administration, California State
University, Long Beach, California,
Email: wikrom.prombutr@csulb.edu
4
School of Accounting and Finance,
Hong Kong Polytechnic University, Hong
Hum, Kowloon, Hong Kong
Email: johnwei@ust.hk
Funding information
partial financial support received from
the Research Grants Council of the Hong
Kong Special Administrative Region,
China, Grant/Award Numbers: 12501414,
299913ECS
Abstract
This study comprehensively reexamines the debate over
behavioral and rational explanations for the investment
effect in an updated sample. We closely follow the
previous literature and provide several differences. Our
tests include five prominent measures of corporate
investment and corporate profitability in qtheory and
recent investmentbased asset pricing models. Both
classical and Bayesian inferences show that limitsto
arbitrage tend to be supported by more evidence than
investment frictions for all investment measures. When
idiosyncratic volatility and cash flow volatility are used
in measuring investment frictions, the inference is more
favorable for the rational explanation.
KEYWORDS
investment, investment frictions, limitstoarbitrage, qtheory, stock
returns
JEL CLASSIFICATION
G14, G31, G32, M41, M42
Eur Financ Manag. 2020;26:343. wileyonlinelibrary.com/journal/eufm © 2019 John Wiley & Sons Ltd
|
3
We thank WeiPeng Chen, ChiaWei Huang, Roger Loh, and conference participants at the 2017 China International
Conference in Finance (CICF) in Hangzhou, China, the 2016 European Financial Management Association (EFMA)
annual meeting in Basel, and the 2016 International Symposium on Business and Management in Tokyo. We are
particularly grateful for the constructive and insightful comments and suggestions from the anonymous referees and
John A. Doukas (the Editor). Eric Lam acknowledges partial financial support received from the Research Grants
Council of the Hong Kong Special Administrative Region, China (project no. 299913ECS and project no. 12501414)
during his faculty appointment at the Hong Kong Baptist University. The views expressed in this paper are those of the
authors and do not necessarily reflect those of the Hong Kong Institute for Monetary Research or the Hong Kong
Monetary Authority. All errors are ours.
1
|
INTRODUCTION
Titman,Wei,andXie(2004)andCooper,Gulen,andSchill(2008),amongothers,findthatstocks
of firms with high capital investment or high total asset growth underperform those of firms with
low capital investment or low total asset growth, which is generally referred to as the investment or
asset growth effect. Lam and Wei (2011) compare the predictions of the mispricing hypothesis on
the asset growth effect with the limitstoarbitrage suggested by Shleifer and Vishny (1997) and q
theory with investment frictions, as proposed by Li and Zhang (2010). While Li and Zhang (2010)
show that limitstoarbitrage tend to be more important, the extensive direct comparisons by Lam
and Wei (2011) show that there is evidence supporting all the hypotheses. Whether one
explanation is empirically more important for the investment effect thus remains unclear. We
address this unresolvedissue in our study, whichdiffers from Lam and Wei(2011) in several ways.
First, we extend the sample period, examining data from July 1963 to December 2017, to
increase the power of the tests. Second, in addition to total asset growth (Cooper et al., 2008),
which is the only investment measure used by Lam and Wei (2011), we examine four other
common measures: investmenttoassets (Lyandres, Sun, & Zhang, 2008; Titman et al., 2004), net
operating assets(Hirshleifer, Hou, Teoh, & Zhang, 2004),net share issuance (Pontiff & Woodgate,
2008), and composite share issuance (Daniel & Titman, 2006). Third, we use the same 10 proxies
of limitstoarbitrage and four proxies of investment frictions as Lam and Wei, but construct a
composite index for each friction category, instead of individual measures, to make a fair and
precise comparison. Fourth, unlike Lam and Wei, we control for corporate profitability to analyze
investment frictions, thus aligning the tests more closely with the prediction of qtheory. Fifth,
Lam and Wei (2011) perform individual or joint tests from subsamples split by arbitrage frictions
and/or investment frictions. We instead estimate the Fama and MacBeth (1973) regressions with
the interaction term between investment and limitstoarbitrage or investment frictions. Finally,
in addition to classical inferences, we take the Bayesian approach to hypothesis testing, using the
minimum Bayes factors suggested by Harvey (2017).
Overall, our individual and joint tests using classical and Bayesian inferences yield similar
results but differ from those of Lam and Wei (2011). In general, we find greater evidence for
mispricing with limitstoarbitrage than for qtheory with investment frictions on the investment
effect. First, in ourindividual tests, without controlling for the competing hypothesis, we find that
80% of the cases support arbitrage frictions while 44% support investment frictions (80% vs. 44%).
Second, the joint tests show that 80% of the FamaMacBeth regression slopes on the interaction
term between investment and arbitrage frictions are negative and statistically significantat the 5%
level. The two insignificant slopes come from weighted least squares (WLS) regressions using net
or composite share issuance as the measure of investment. In contrast, only a small number of
cases show that the investment effect is significantly related to investment frictions as predicted
by qtheory. Only 14% of the FamaMacBeth regression slopes of the interaction term between
investment and investment frictions are negative and statistically significant at the 5% level. All
significant cases come from ordinary least squares (OLS) regressions, with investmenttoassets
generating the most evidence, followed by total asset growth.Other investment measures provide
no evidence for the investment frictions hypothesis.
The Bayesian inferences provide similar findings. We start with a prior odds ratio of 4:1; that
is, the prior probability that the null (investment frictions or arbitrage frictions are not important)
is true is 80%, and the prior likelihood that the alternative (investment frictions or arbitrage
frictions are important) is true is 20%. We find that in about 51% of the regression coefficient
estimates, the posterior probability of the arbitrage frictions null being true is less than 5%. The
4
|
EUROPEAN
FINANCIAL MANAGEMENT
LAM ET AL.
results with OLS and WLS regressions are significant when the measure of investment is total
asset growth or investmenttoassets. In contrast, the posterior probability of the investment
frictions null being true is less than 5% in only about7% of the regression coefficient estimates. All
of the significant results are exclusively derived from using investmenttoassets as the measure of
investment with OLS regressions. Interestingly, the overall results show that investmenttoassets
appears to be the most important measure of investment in terms of supporting eitherhypothesis,
followed by total asset growth, net operating assets, and then net share issuance, with composite
share issuance being the least important.
Our findings suggest that the mispricing hypothesis with limitstoarbitrage empirically
outperforms qtheory with investment frictions in explaining the investment effect. This is
consistent with Li and Zhang (2010). However, we cannot rule out the qtheory explanation. The q
theory is unsurprisingly a viable economic mechanism for understanding the return predictability
of investmenttoassets, as it focuses on corporate capital investment. The limitstoarbitrage
hypothesis, however, is more important because it can explain return predictability for a broadly
defined measure of investment, even with WLS regressions in some cases. The findings suggest that
compared with smallcap stocks, arbitrage frictions are less important, and investment frictions are
not important at all for largecap stocks. Yet, the relative importance of the two frictions depends on
the set of variables used in the construction of the indices. For example, when idiosyncratic and
cash flow volatilities are used as investment frictions measures instead of arbitrage frictions
measures, there is more evidence for qtheory but less evidence for limitstoarbitrage.
The investment effect continues to receive attention from academics and practitioners alike.
1
Lam and Wei (2011) point out that the limitstoarbitrage hypothesis should make similar
predictions as the investment frictions hypothesis, as proxies for both are highly correlated. As
the evidence drawn from individual tests may support both hypotheses, it is important to
conduct joint tests to distinguish between the two explanations. Although Lipson et al. (2011)
and Li and Sullivan (2011), (2015) provide evidence that limitstoarbitrage play a significant
role in the return predictability of total asset growth, their tests do not control for investment
frictions or profitability. Extending the joint test framework of Lam and Wei is thus crucial to
testing the limitstoarbitrage hypothesis. Our findings confirm the conclusion of Lipson et al.
(2011) and Li and Sullivan (2011), (2015) that limitstoarbitrage are important.
We make several contributions to the literature. First, we confirm that the mispricing
hypothesis with limitstoarbitrage can better explain the investment effect than can qtheory
with investment frictions. Our sample contains a longer time series than that used by Lam
and Wei (2011), and our crosssectional regressions are estimated with all stocks rather than a
subset. Instead of a horse race based on different numbers of significant arbitrage friction
measures versus significant investment friction measures, we test a composite index of
arbitrage frictions against a composite index of investment frictions. The construction of these
composite indices follows Stambaugh, Yu, and Yuan (2015), who construct a composite proxy
for stock mispricing. As each individual measure contains noise, our approach of constructing
a composite index by averaging rankings within each category of frictions should be able to
diversify the noise in the cross section, and can provide a fairer and more precise comparison
between the two hypotheses. Our conclusion is robust to both classical and Bayesian
inferences.
1
Recent papers touching upon the asset growth effect include Cooper, Gulen, and Schill (2010), Cooper and Priestley (2011), Gray and Johnson (2011), Lipson,
Mortal, and Schill (2011), Titman, Wei, and Xie (2013), Watanabe, Xu, Yao, and Yu (2013), Li and Sullivan (2011), (2015), Mao and Wei (2016), and
Papanastasopoulos (2017), among others.
LAM ET AL.EUROPEAN
FINANCIAL MANAGEMENT
|
5

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT