do firms save too much cash? evidence from a tax on ...€¦ · korean rms hold three times more...
TRANSCRIPT
Do Firms Save Too Much Cash?Evidence from a Tax on Corporate Savings∗
Hwanki Brian KimBaylor University
Woojin KimSeoul National University
Mathias KronlundUniversity of Illinois at Urbana-Champaign
October 24, 2019
∗Hwanki Brian Kim is at Baylor University; Hankamer School of Business; One Bear Place; Waco, TX76798; U.S.A.; Email: Brian [email protected]. Woojin Kim is at Seoul National University; 1 Gwanak-ro,Gwanak-gu, Bldg 59; Seoul, 08826; Korea; Email: [email protected]. Mathias Kronlund is at theUniversity of Illinois at Urbana-Champaign; Gies College of Business; 1206 South Sixth Street; Champaign, IL,61820; U.S.A.; Email: [email protected]. Yongseok Kim and SeongMyeong Kang provided excellentresearch assistance. We thank Heitor Almeida, Peter DeMarzo, Brent Glover, Kristine Hankins, and seminarand conference participants at the BYU Red Rock conference, ESSFM Gerzensee, the Korean SecuritiesAssociation, Seoul National University, the University of Illinois at Urbana-Champaign, and Virginia Techfor many helpful comments. A previous draft of this paper was circulated under the title “DiscouragingCorporate Savings”
Do Firms Save Too Much Cash?Evidence from a Tax on Corporate Savings
October 24, 2019
Abstract
Corporations have accumulated record amounts of cash. Are these savings optimal or
excessive? We examine this question by exploiting a Korean tax reform that sought
to discourage cash savings by imposing a surtax on earnings that were not paid out
to shareholders or invested. This tax applied only to firms with book equity above 50
billion wons or that belonged to a Chaebol. Difference-in-differences tests show that
the treated firms reduced cash savings and increased payouts, wages, and investments.
These additional investments appear profitable, and an event study analysis shows that
shareholders viewed the reform as value-enhancing. These results are consistent with
the accumulation of excessive savings before the reform, and we find evidence consistent
with both agency-based and behavioral channels underlying such excessive savings.
JEL category: G32, G35, G38
Keywords : Corporate cash, investment, wages, payout policy, natural experiment, tax
1 Introduction
Corporate cash balances have risen dramatically over the last few decades. In the United
States, cash on corporate balance sheets now totals over $3 trillion (Faulkender, Hankins,
and Petersen, 2019), and the trend toward growing corporate cash balances is prevalent also
in many other countries (Kalcheva and Lins, 2007; Pinkowitz, Stulz, and Williamson, 2015).
This phenomenon is often referred to as a “corporate savings glut” or “cash hoarding.”1 Still,
it remains an open question whether all of this cash represents excessive hoarding or optimal
savings.
It is challenging to determine whether corporations are saving too much and investing
too little, as the “cash hoarding” narrative would suggest. Corporate decisions about cash
policies, investments, and payouts are endogenously determined and driven by multiple factors
that are difficult to control for in empirical tests. To address the question of whether firms
save too much, we ideally need both a shock to firms’ cash policies—where some companies
exogenously change their savings behavior—and a way to measure the value impact of these
changes.
The main objective of this paper is to address this question through the lens of a recent
natural experiment in South Korea, where the government intentionally sought to stem the
tide of corporate cash by imposing a new 10% surtax on any “excess” cash savings above
a specific threshold of earnings.2 This tax reform became effective on January 1, 2015. A
crucial feature of the reform for identification purposes is that it applied only to firms that
had a level of shareholder’s equity above 50 billion Korean wons (equivalent to around $50
million USD), and to firms that were part of large business conglomerates (chaebol), whereas
non-chaebol firms below the 50-billion threshold were not subject to the new tax. This setting
1For example, a 2016 Financial Times article argued that “The failure of companies to invest their cashpile has frustrated investors who say companies are not plowing enough back into their underlying businesses,in research and development”; www.ft.com/content/368ef430-1e24-11e6-a7bc-ee846770ec15
2In the next section, we discuss the exact formula that determined the surtax and relevant exclusions.
1
allows us to study how the firms that were treated by the tax law changed their savings and
investment policies, and to measure the resulting consequences for firm value. This evidence
thus helps us assess whether saving less enhanced firm value, or whether the law merely
distorted corporate cash and investment policies.
Korean firms are among the most cash-heavy in the world: Compared with U.S. firms,
Korean firms hold three times more cash as a fraction of GDP: Corporate cash balances are
equivalent to 50% of GDP in Korea, compared with around 15% in the U.S..3 The law was
motivated by the idea that money sitting on corporate balance sheets could be used more
productively in the hands of shareholders or employees, or in the form of new investments.4
Our empirical strategy employs a difference-in-differences analysis, in which we compare
changes from two years before the tax reform (2013–14) to two years after (2015–16) in firms’
cash savings, payouts, and investments between the firms that were treated by this reform
versus firms that were not treated. An attractive feature of the 50-billion-wons threshold that
determined treatment is that this threshold is plausibly exogenous to firms’ future savings
and investment policies, and also unrelated to any other governmental reforms of which we
are aware that could have differentially affected the treated and non-treated firms around the
same time. Crucially, this surtax on cash savings was also significant in economic magnitude:
a ten-percentage-point surtax implies an almost 50% increase on top of the standard corporate
tax rate, which was 22% during our sample period.
Our results show that the treated firms responded by significantly cutting savings. The
economic magnitude of this effect is also large: the savings rate among the treated firms was
3This phenomenon is most evident among the largest business groups (chaebol). In 2017, the top fourchaebol together held the equivalent of around 500 Billion USD in cash. See The Economist, September 27,2014, at www.economist.com/leaders/2014/09/27/a-25-trillion-problem; and Financial Times, June 11, 2017,at www.ft.com/content/966fcbd8-4f13-11e7-bfb8-997009366969
4From an economy-wide perspective, money on corporate balance sheets is not “idle”; some of it is directlyinvested in other corporations’ stocks and bonds (Gilbert et al. 2017), and cash that is held as bank depositsis deployed as a source of lending.
2
cut by 0.3–0.6 percentage-points as a fraction of assets, compared with an average savings
rate of around 1 percent of assets.
A firm must necessarily either spend money that is not saved or pay it out to investors.
We show that the treated firms responded by significantly increasing payouts to shareholders.
Specifically, the treated firms pay out 0.2 percentage points more (as a fraction of assets)
after the reform, about a 30% increase relative to the sample mean of 0.7%. We also find
that the investment rate for the treated firms increased by around 20
To determine whether these actions were value-improving—and thus, whether the treated
firms’ savings before the reform might have been excessive—we next study the profitability
and valuation consequences for the treated firms. First, we show that the relative increases in
investments among the treated firms were especially concentrated among firms and industries
that exhibited high profitability before the reform. This implies that the marginal investments
were focused primarily in areas where these investments were likely to be more profitable.
Even if the additional investments by the treated firms were profitable; however, that does
not rule out the possibility that the shadow value of savings was even higher. We therefore
study the valuation consequences for the firms that were treated by the tax reform. Using
an event study design, we find that the cumulative abnormal returns (CAR) for the treated
firms around the days when the law was proposed and passed were positive. This finding is
consistent with investors believing that firms’ cash savings before the reform were excessive
and that saving less would raise firm value. The positive valuation effects are particularly
notable because a direct negative valuation effect for the treated firms is that they are subject
to higher taxes; this implies that investors viewed the responses by the treated firms as more
than outweighing these direct costs.
Cross-sectionally, we find stronger responses in terms of changes to savings, payout, and
investments among the firms that had high pre-reform earnings and that, therefore, would be
3
required to decrease their savings to the greatest extent to avoid the surtax. These firms also
experience more positive valuation effects around the passage of the law, which is consistent
with investors anticipating that these firms would change their behavior to the greatest
extent.
We finally explore two possible channels through which firms’ cash savings could have
been excessive before the reform and why being incentivized to save less might have improved
firm value. First, firms could have saved excessively because of agency conflicts (Bertrand and
Mullainathan, 2003; Gao, Harford, and Li, 2013; Nikolov and Whited, 2014). Second, firms
might have employed overly cautious cash policies because of behavioral biases (Malmendier,
Tate, and Yan, 2011); this would have occurred, for example, if managers have been scarred
by past crises such as the Asian Financial Crisis of 1997–98. In the final part of the paper, we
investigate these two possible mechanisms that might have induced firms to save too much,
and we find evidence consistent with both of these channels at work.
We conduct a battery of robustness tests for our findings. The crucial identification
assumption—as in any difference-in-differences design—is the “parallel trends” assumption.
To support this assumption, we show that there are no significant differential pre-trends
between the treated and control firms. We also conduct a “placebo” test by exploiting the
feature that the 50 billion threshold is relevant only for non-chaebol firms, by showing that
there are no effects around this threshold on firms that belong to chaebol and are therefore
treated regardless of their size. These results indicate that our results are not driven by
differential trends that might be related to a firm’s size.
We further show that our difference-in-differences regressions are robust to a variety of
controls, including controls for the distance to the treatment threshold (as in a regression
discontinuity), size, leverage, cash flow, and cash levels. The results are also similar when we
focus on a range around the threshold (firms with a shareholder’s equity between 10 and 90
4
billion wons). These tests also help ensure that any effects we find are not spuriously driven
by differential trends from the smallest or largest firms in Korea.
The Korean setting represents a unique opportunity to study the effects of a government
policy that was aimed at curtailing corporate savings. A historical episode involving a
similar reform was the “undistributed profits tax,” which raised taxes on retained corporate
earnings in the United States between 1936 and 1937. Poterba et al. (1987) show that the
undistributed profits tax resulted in more payouts. Christie and Nanda (1994) use an event
study methodology to show, similarly to our findings, that the undistributed profits tax was
associated with positive abnormal returns, and they argue that this tax also alleviated agency
problems whereby firms had been paying out too little of their profits. The recent experiment
in Korea has at least two primary benefits compared with studies of the undistributed profits
tax in the U.S. First, crucially, the Korean experiment has a control group of firms that were
not subject to the new tax, whereas the undistributed profits tax had no such natural control
group. Second, the Korean experiment takes place in recent times when the nature of firms’
balance sheets differ substantially from what it was in the 1930s.
Our study further contributes to the literature on the value of cash. Faulkender and
Wang (2006) estimate that the marginal value of cash on average is $0.94, but that this varies
across firms, depending on their financial constraints. Dittmar and Mahrt-Smith (2007) find
further that the value of cash is lower for poorly governed firms and Harford (1999) shows
that cash-rich firms tend to do more acquisitions of other firms and that these acquisitions
tend to be value-decreasing.5
We also contribute to the literature on the determinants of corporate cash. Studies
have cited several reasons that firms hold more cash than they once did, including more
precautionary savings (Opler, Pinkowitz, Stulz, and Williamson, 1999; Bates, Kahle, and
5However, Opler, Pinkowitz, Stulz, and Williamson (1999) do not find that excess cash predicts moreacquisitions, capital expenditures, or payouts.
5
Stulz, 2009), lower opportunity costs (Azar, Kagy, and Schmalz, 2016), increased competition
(Lyandres and Palazzo, 2016), costs of repatriating foreign earnings (Foley, Hartzell, Titman,
and Twite, 2007; Faulkender, Hankins, and Petersen, 2019), and higher levels of intangible
capital (Falato, Kadyrzhanova, Sim, and Steri, 2018).6 While most of this literature has
focused on cash levels rather than savings, two exceptions include Almeida, Campello, and
Weisbach (2004) and Riddick and Whited (2009), who examine what we can learn about
firms’ financial frictions from the sensitivity of firms’ savings to cash flows.
Finally, this experiment also relates to more recent studies on the “tax holiday” of the
“American Job Creation Act”(AJCA) of 2004, which allowed firms to repatriate foreign
earnings. Blouin and Krull (2009), Dharmapala, Foley, and Forbes (2011), and Faulkender
and Petersen (2012) examine what firms do after the ACJA when they enjoy (cheaper) access
to previously saved foreign cash. Blouin and Krull (2009) and Dharmapala, Foley, and Forbes
(2011) show that most of the money that was repatriated during the tax holiday was spent on
buybacks, while Faulkender and Petersen (2012) also find an increase in domestic investment
after the AJCA.
Our paper proceeds as follows. In the next section, we describe the legislative history
and the implementation of the law. In section 3, we describe our data on Korean firms and
summary statistics. In section 4, we present results on firms’ responses to the tax reform,
and in section 5, we describe several robustness tests. In section 6, we examine the valuation
effects. In section 7, we analyze cross-sectional variation in firms’ responses and valuation
effects; these tests shed light on the possible channels that could drive firms to save excessively.
We conclude in Section 8.
6The literature on the determinants of cash is surveyed by Almeida, Campello, Cunha, and Weisbach(2014).
6
2 Legislative History and Institutional Background
On July 16, 2014, the newly appointed then-Minister of Strategy and Finance in Korea, Choi
Kyoung-Hwan, announced in his inaugural address a plan for tax reforms. Among them was
the new tax on corporate savings. He pronounced that his team was considering taxes on
firms’ retentions of unspent earnings, with the aim of incentivizing corporate investments
and boosting domestic demand. The prevailing view was that money sitting on corporate
balance sheets could be put to more productive use if only firms were incentivized to spend
their cash on plants, equipment, labor, or payouts to shareholders.
The official proposal for the tax reform was made on August 6, 2014, the National
Assembly passed the law on December 2, 2014, and the law took effect on January 1, 2015.
The law did not target all firms but applied only to those that satisfied at least one of the
following criteria: 1) having shareholders’ equity greater than or equal to 50 billion Korean
wons as of the previous fiscal year, or 2) belonging to a large business group, i.e., a chaebol.
Firms that did not satisfy either criterion were not subject to the new tax.
For the treated firms, the reform imposed a 10% surtax on “excess saving,” defined as
earnings above a certain threshold after deductions for investments, wage increases, and
payouts to shareholders. The first surtaxes were assessed in April 2016 based on end-of-
2015-fiscal-year accounting information. Firms could choose between two separate rules for
determining the tax amount:
(Rule A): tax amount = [net profit×0.8
− (investments + wage increases + payouts)]×0.1
(Rule B): tax amount = [net profit×0.3 - (wage increases + payouts)]×0.1
7
For example, according to Rule A, treated firms that spent more than 80% of their net
income on investments, wage increases, or payouts during a given year would be fully exempt
from the tax, while treated firms that saved more than the threshold amount would be taxed
at 10% on the residual. Rule A was the relatively more favorable rule for firms that tend to
make substantial investments,7 whereas Rule B was more favorable to firms that do not make
large investments. Even though firms could choose which of the rules they would adhere to
initially, they were bound to their chosen rule for at least the next three years.8
This surtax was also significant in economic magnitude: A ten-percentage-point surtax
implies an almost 50% increase on top of the standard corporate tax rate, which was 22%
during our sample period. Also, note that this surtax was not a tax on cash balances, but
instead the tax applied to excessive accumulation of retained earnings as non-invested capital.9
The law made important exclusions related to the deductions for wage increases and
investments that were allowed. The law explicitly did not allow wage increases awarded to
top managers to count as deductions. Regarding the deduction for investments, there was a
concern that firms might seek to “store” money in the form of physical investments rather
than cash, such as by buying land. The law therefore required that firms taking a deduction
for buying land also would be required to start building facilities on the land by the end of
the year. The law further imposed a retroactive tax if a firm sold or leased any such land
for a period of up to two years after the completion of any construction on the land. While
payouts to shareholders was an allowed deduction, firms nevertheless were not allowed to
claim an equivalent deduction for paying back debt.
This reform was controversial and received significant pushback from the private sector.
On the one hand, an argument supporting the reform was that this measure was designed
7Specifically, for firms whose capital expenditures represents more than 50% of net earnings.8We do not have data indicating which rule each treated firm chose to follow, but we can nevertheless
calculate a lower bound on the tax amount based on the most favorable rule for each firm.9Cash balances already incur an implicit tax penalty because interest is subject to corporate taxes (e.g.
Gamba and Triantis (2008)).
8
to push corporate cash savings toward households through investments, wage increases, or
payouts. One concern quite broadly shared by politicians and policymakers was that Korean
firms had accumulated cash excessively after the financial crisis.10 A common view was also
that a corporate tax cut that had been passed in 2009 had resulted in even higher cash
accumulation in place of the intended effect of increasing investment activities. An implicit
assumption behind this argument was that firms were saving more cash than was optimal,
and that directing more of this cash towards households would result in higher aggregate
consumption.11
On the other hand, other observers raised doubts about the effectiveness of the new
tax. These views were based primarily on the assumption that retention rates are optimally
determined, and that any additional taxation or “forced” spending on payouts or invest-
ments would drive firms to make suboptimal decisions. It was also argued that the reform
would effectively increase the corporate tax rate, which would eventually suppress corporate
investment and growth.
3 Data and Empirical Strategy
In this section, we discuss our empirical strategy, sample construction, and variable definitions,
and we report summary statistics.
10As of 2014, Korean firms reported the second-largest corporate savings among OECD member nations asa fraction of GDP, trailing only Japan.
11For example, when he was a nominee as the minister, Choi emphasized in a press interview that he wouldmeasure success by welfare improvements for the household economy (https://thediplomat.com/2014/06/south-koreas-incoherent-economic-policy/).
9
3.1 Empirical strategy
Studying the effects of changes to corporate savings is generally challenging because it is
difficult to find a source of exogenous variation in savings across firms. To overcome this
challenge, we exploit the Korean tax reform as a natural experiment. We employ a difference-
in-differences strategy by comparing firms that were treated by this reform with firms that
were not treated, from two years before the reform (2013–14) to two years after (2015–16).
An important advantage of using the Korean reform as a natural experiment is that the
assignment of treatment is plausibly unrelated to omitted variables that could influence
changes in future corporate savings and investment policies and thereby invalidate the parallel
trends assumption.
We define an indicator variable ’Treated ’ to denote a firm that is treated by the tax reform,
and estimate the following baseline difference-in-differences model:
yi,t = θ + β0Treatedi + β1Aftert
+ β2Treatedi × Aftert + η′ · Xi,t + ϕi + τt + ψj,t + εi,t,
where i indexes firms, j indexes industries,12 and t indexes years. yi,t denotes outcome
variables, such as investments, wage increases, and payouts; After is an indicator variable
for the post-reform years; X is a vector of control variables based on firm characteristics.13
ϕi denotes firm fixed effects to control for unobserved time-invariant firm heterogeneity; τt
denotes year fixed effects to control for unobserved market-wide shocks for each year; and
12The Korean industry classification uses a six-digit code. To ensure that we have sufficiently many firms ineach industry classification, we use a five-digit industry code that roughly corresponds to two- or three-digitaggregation of industries using U.S. SIC codes.
13Because of a possible “bad controls” concern (Angrist and Pischke (2008), p. 64), we first report resultsfrom these difference-in-differences regressions without these additional control variables. We then report theresults with the control variables later for a robustness check.
10
ψj,t denotes industry-by-year fixed effects to control for unobserved industry-wide shocks for
each year.14
The main coefficient of interest is β2, which captures the average treatment effects for
each of the outcome variables. We cluster standard errors throughout at the firm level for
stand-alone firms or, alternatively, at the Chaebol level for firms belonging to a chaebol group.
All variables are winsorized at the 1% and 99% levels.
Because treatment is based on a firm’s shareholders’ equity, one may nevertheless be
concerned about the possibility that a firm’s equity size could be spuriously correlated with
differential trends; for example, suppose that larger firms tended to perform better after
the reform, but for reasons unrelated to the reform. To address this concern, we repeat
the baseline difference-in-differences estimation while limiting the sample to firms for which
the pre-reform shareholders’ equity is within a narrower bandwidth around the treatment
threshold of 50 billion Korean wons. Specifically, we employ a bandwidth that ranges between
10 billion and 90 billion Korean wons to make it sufficiently narrow so that firms are more
comparable and less likely to be affected by differential trends, while also maintaining a
reasonably large sample to ensure sufficient statistical power.
3.2 Data and sample construction
Most of the data for this study, including accounting information, stock prices (for firms
publicly listed on the stock exchange), and other firm characteristics, are collected from
DataGuide, which is a comprehensive database on both private and public firms in Korea.
To determine more precisely which firms are affiliated with chaebol groups, we manually
double-check the accuracy of our data using an annual report by the Korean Fair Trade
Commission. We obtain data on corporate governance scores from the Korea Corporate
14Because we include both firm fixed effects and year fixed effects in all specifications, the Treated andAfter variables are in practice subsumed by those fixed effects.
11
Governance Service, and data on the probability of default for public firms from the Risk
Management Institute of the National University of Singapore (NUS RMI).
Beginning with these data on all public and private firms in Korea, we employ several
filters to further refine the sample. We first eliminate firm-year observations for which data on
shareholders’ equity are missing, as this variable determines treatment. We further exclude
financial firms and utilities. We next screen out firms that have total assets of less than
1 billion Korean wons (approximately $1 million) and firms with very negative levels of
capital expenditures (less than -10% of total assets). As a final filter, we drop three specific
firms—Hyundai Motors, Kia Motors, and Hyundai Mobis—as these firms made a massive
investment in real estate during our sample period but that had been planned a long time
before the tax reform.15 After restricting our sample period to four years around the tax
reform, i.e. from 2013 through 2016, our final sample consists of 80,494 firm-year observations
across 20,916 unique firms.
3.3 Main variables and summary statistics
Because the tax reform was aimed directly at discouraging savings while encouraging payouts,
investments, and wage growth, we start our empirical analysis by studying changes to these
outcomes as dependent variables. Cash savings (∆Cash/assets), payouts (Payout), investment
(Investment), and wage growth (Wage increase), are defined, in turn, as changes in cash
holdings during the year, total cash dividends plus share repurchases, capital expenditures,
and changes in a firm’s wage bill for non-executive employees during the year.16 All of these
15Including these firms would make the estimated effects on investments stronger, but would less accuratelyreflect the impact of the tax reform.
16As described in section 2, wage increases for executives did not count as a deduction from the surtax.
12
variables are normalized by lagged total assets. Definitions for all variables are described in
Table A9 in the Internet Appendix.17
While the definitions of most variables are straightforward, it is worth paying extra
attention to that of our measures of share repurchases: We follow Fama and French (2001)
and Almeida et al. (2016) and focus on purchases of stock, because the tax reform did not
penalize sales of stock. In other words, we measure repurchases as purchases of stock if they
are not missing, and replace them with increases in common Treasury stock if stock purchases
are missing or zero and Treasury stock is not missing for the current and prior year. If neither
data on purchases of stock nor changes in Treasury stock are non-missing, repurchases are
set to zero.
Table 1 summarizes the panel data over our sample period (2013 through 2016). In Panel
A, we report summary statistics on firm characteristics.
Table 1 About Here
In Panel B of Table 1, we report statistics on how many firms were treated and not treated
by the reform by each criterion for treatment. Of the sample firms, 1,052 are chaebol -affiliated,
and 3,451 firms report shareholders’ equity that is greater than 50 billion Korean wons, while
599 firms are treated based on both of these dimensions. In total, 3,904 firms were treated,
i.e. subject to the tax reform. We will exploit this “two-dimensionality” of the treatment as
part of our identification strategy and in our “placebo” robustness tests.
17We multiply these asset-scaled variables by 100 throughout to aid interpretation, as this defines thevariables in percentage terms.
13
4 Main Results
In this section, we employ the difference-in-differences methodology to study the effects that
this reform had on firms’ financial and investment policies. These results crucially show
whether the reform had any “bite” in affecting how firms save, and the results also shed light
about the margins along which firms respond most strongly. It is nevertheless important to
note that these results regarding how firms respond to the law do not tell us whether these
responses enhance firm value or not; we will turn to that crucial question in the following
section.
4.1 Effects on cash accumulation
We begin by studying how the reform affected corporate cash savings. In a simple theoretical
framework, firms should equate the marginal cost of increasing cash holdings and thus
potentially sacrificing valuable investment opportunities today with the marginal benefit
of having available internal cash to invest in the future (e.g., Opler et al. (1999); Almeida,
Campello, and Weisbach (2004)). The surtax raises the marginal cost of saving, so we might
naturally expect firms to save less.
This simple theoretical prediction is nevertheless made somewhat less evident because
the law provided no symmetric tax credit for dissaving (negative cash accumulation). The
reform thus creates a wedge between saving and dissaving: Saving above a certain threshold
of earnings is taxed, but payouts or investments that result in dissaving in another year does
not provide firms with an equivalent credit. As a result, we might expect treated firms to
smooth their savings to a greater extent and avoid years of dissaving. The overall effects of
the reform on savings are thus ambiguous.
In Table 2, we report difference-in-differences results on how the tax reform affected cash
savings in the treated firms. The dependent variable is defined as changes in cash holdings
14
scaled by total assets. In Columns 1 and 2 we report the results for the full sample and in
columns 3 and 4 we focus on firms that fall within a narrower bandwidth of shareholders’
equity around the treatment threshold (in a range between 10 and 90 billion wons). We
control throughout for firm fixed effects, as well as year or industry-by-year fixed effects.
Table 2 About Here
Our results show that treated firms responded by significantly reducing cash savings to
a greater extent than the non-treated firms after the tax reform: The coefficients on the
Treated × After interaction terms are significantly negative in all specifications. The economic
magnitude of this effect is also substantial: the savings rate among the treated firms was cut
by around 0.3 percentage points when normalized by assets, relative to an average savings
rate of around 1 percent. These results are qualitatively similar but slightly larger when we
focus on the narrower range of firms around the treatment threshold, as seen in columns 3–4.
In Panel A of Figure 1, we show this effect graphically. In this figure, we match each
treated firm to one non-treated control firm that is most similar (its “nearest neighbor”) before
the reform, and the figure shows that while these treated and control firms displayed similar
cash savings behavior in every year before the reform, the treated firms saved significantly
less in the years after the reform.18
Figure 1 About Here
18In the matching procedure, we require an exact match on industry, and further match on the pre-periodCash/assets, payouts, wage increases, and investment based on the Mahalanobis distance.
15
4.2 Effects on payout policy
In the previous section, we showed that the treated firms significantly reduced cash savings
after the tax reform. Given this reduction in cash accumulation, we now examine how firms
spent the cash that was not saved. We start by examining how payout policy is affected by
the reform, and in the following sections, we study effects on wage increases and investments,
respectively. It is nevertheless important to note that cash savings, payouts, wages, and
investments do not exhaust all the possible uses (or sources) of corporate cash flows, and
thus, the total effects on these alternative uses of savings do not necessarily add up precisely
to the estimated decrease in savings. We nevertheless study these specific effects as they are
the ones explicitly targeted by the law.
We measure payout as the sum of cash dividends and share repurchases scaled by total
assets (multiplied by 100, so that any effects can be interpreted as a percent fraction of
assets). Table 3 reports results. The results reported in columns 1 and 2 of Table 3 show
that the treated firms increased payouts following the tax reform: the economic magnitude
is around 0.2 percentage points of assets, which is around 30% of the sample mean payout
of 0.7%. In columns 3 and 4, we restrict the sample to a narrower bandwidth around the
regulatory threshold and find similar results. Panel B of Figure 1 also shows these results
graphically for the matched treatment-control sample.
Table 3 About Here
We next examine the components of payouts, i.e. dividends and repurchases separately.
The results reported in Table A1 in the Internet Appendix indicate that while both forms of
payout significantly increased for treated firms following the tax reform, dividends increased
more than repurchases in absolute magnitude. Guay and Harford (2000) and Jagannathan,
Stephens, and Weisbach (2000) argue that firms tend to use dividend increases to distribute
16
cash from relatively permanent shocks, whereas they use repurchases to distribute transitory
shocks. Thus, to the extent that dividends signal payouts that are more permanent, this result
suggests that the treated firms established a permanently higher level of payouts after the
surtax came into effect. The differences between the increases in dividends and repurchases
can also reflect the fact that dividends overall represent a much larger share of total payouts
in this sample, which in turn is related to the fact that the majority of the sample firms are
private. Average dividend levels equal 0.59% of assets, whereas average repurchases equal
only 0.1%, which means that the repurchases increased more in percent (by around 50%)
than dividends (30%).
4.3 Effects on wage increases and investments
We next focus on the effects on two real outcome variables: wage increases and investments.
Higher wages and investments were also among the outcomes that were explicitly targeted by
the tax reform, as these were allowed to offset the surtax on retained earnings.
In Table 4, we report how the growth in wage bills changed around the tax reform for
the treated firms. We measure wage increases as changes in firms’ non-executive wage bills
scaled by total assets. The results reported in columns 1 and 2 indicate that treated firms
increased wage bills. These results continue to hold when we consider the narrower sample
around the threshold (see columns 3 and 4).19
19For a subset of firms (around 10% of the sample), we have data on both the total wage bill and thenumber of employees, which further allows us to break down these wage bill changes into wages for new hiresand wage increases to existing employees. In Table A2 in the Internet Appendix, we report the results weobtain after separately examining the effects on these two outcome variables: wage per employee and thenumber of employees. These results are nevertheless not entirely conclusive. When we do not limit the samplebased on shareholder’s equity being relatively close to the threshold, the results indicate that the treatedfirms did raise wages to existing employees while there is no significant effect on the number of employees.On the other hand, when we consider the narrower sample around the treatment threshold, we instead findlarger effects on the number of employees.
17
Table 4 About Here
We next examine the effects on investments. We measure investments as changes in
tangible assets plus depreciation scaled by total assets. The results reported in columns
1 and 2 of Table 5 indicate that the treated firms increased their investments compared
with non-treated firms. The economic magnitudes are significant as the investment rate
for the treated firms relative to untreated firms increased by around 20%-30%. In columns
3–4, we report the results for the narrower sample around the treatment threshold, and
the estimated effects on investment here are even greater in magnitude. Panels C and D
of Figure 1 graphically show the results on wage increases and investment for the matched
treatment-control sample.
Table 5 About Here
One advantage of the data on Korean firms is that it further allows us to identify what
kinds of investments are made. When we focus on these separate components of capital
expenditures, we find that the treated firms increased investments about equally in Land,
Buildings, and Equipment. These results are reported in Table A3 in the Internet Appendix.20
Why did firms increase investments? On the one hand, these results are consistent
with firms’ having available positive-NPV projects that previously were unfunded (Fazzari,
Hubbard, and Petersen, 1988). If the treated firms had no such investments, they could
alternatively avoid any surtax by increasing payouts instead of increasing investments. The
fact that firms did invest more thus implies that firms viewed the value of these investments
as sufficiently high to outweigh the alternative of paying out more money to shareholders. On
20The results for the narrower sample are similar, but each of the separate components of investment is nolonger individually statistically significant.
18
the other hand, it is also possible that investing more could be a sign of an agency problem
in the form of “empire-building.” In Section 7.3, we examine this possibility in detail.
4.4 Other effects: Leverage and profitability
Saving less could potentially have negative consequences for a firm’s risk of finding itself
in financial distress. Did these changes contribute to higher risk? We find that the treated
firms have higher leverage after the reform went into effect and report the results in Table 6
(columns 1 and 2), which is consistent with higher risk. The economic magnitude is around
one percentage point in higher book leverage.
Table 6 About Here
Observing higher leverage for the treated firms is not surprising, as these firms start
saving less cash and paying out more.21 Whether or not this leverage increase is concerning
depends on the firms’ initial leverage, which nevertheless was relatively conservative at a
mean of 31% (with a median of 27%).
We see evidence of such muted effects on default risk when considering the estimated
treatment effects on the probabilities of default.22 That is, despite the higher leverage of the
treated firms, we observe an effect on the probability of default that is both economically
and statistically indistinguishable from zero. This result is similar if we measure distress risk
over a two-year horizon (see columns 3 and 4 of Table 6) and a three-year horizon (columns
5–6). If the trends in lower savings and higher payouts continue for the treated firms also
21The reform notably treated payments to debt and equity investors in a non-neutral way, as it does notallow debt repayment to count as a tax-reducible item.
22The probability-of-default measure is based on data from the NUS Risk Management Institute. Thismeasure is calculated based on their methodology, which is described in detail at http://www.rmicri.org. Onedownside of the probability of default data is that it is only available for a subset of the public firms.
19
in the long run, however, we would eventually expect the divergence in leverage to increase
further, potentially causing significantly added risks of distress.
What effect did these actions have on firms’ overall levels of profitability? In Table A4
in the Internet Appendix, we report results pertaining to the operating performance of the
treated firms, measured by Return on Invested Capital (ROIC, measured as EBITDA scaled
by total invested capital) as well as by profit margin (net sales less the cost of goods sold as
a percentage of net sales).
We find that the treated firms display improved profitability. These results suggest that
the actions that firms took in response to the law did not hurt profitability and that these
actions, if anything, improved their overall returns on capital. The economic magnitude of
these profitability changes is also meaningful. These results do not, however, by themselves
represent conclusive evidence that these actions were value-improving—a question that we
will analyze more formally in section 6.
5 Robustness
In this section, we conduct a battery of robustness tests of our findings.
We begin by exploiting the “double-dimensionality” in the assignment of treatment,
i.e., the fact that treatment was based on satisfying one of two inclusion criteria. As is
typical in a difference-in-differences setting, the key identification assumption is the “parallel
trends” assumption. One might nevertheless be concerned that particularly the chaebol firms
could follow different trends from the non-chaebol firms, which in turn could be spuriously
contributing to our results. To account for this possibility, in Panel A of Table 7 we report
results from the difference-in-differences tests while restricting the sample to only non-chaebol
firms that fall in the narrower range around the treatment threshold. For these firms,
“treatment” is based solely on whether shareholders’ equity is above 50 billion wons. The
20
results show that the economic magnitudes and statistical significance of the effects remain
largely similar to what we found with the full sample. Thus, chaebol firms do not drive the
overall effects, and these results are robust to examining effects within non-chaebol firms.
Table 7 About Here
Next, even within the relatively narrow range between 10 and 90 billion wons, we might
be concerned that the slightly larger firms above the 50 billion threshold might be following
a different trend that was unrelated to the reform, compared with the trend that the slightly
smaller firms below the threshold follow. To address this concern, we therefore limit the
sample to only chaebol firms in the 10–90 billion range and report the results in Panel B of
Table 7. This test thus serves as a “placebo” test, because the 50 billion threshold has no
impact on treatment for these firms. Consistent with this idea, we find no effect (or, in the
case of investment, even an effect that is the opposite of what we observe with our baseline
results) around the threshold across all of the dependent variables.
In sum, the results reported in Panels A and B of Table 7 show that the 50 billion
threshold predicts an effect among the non-chaebol firms while we find no effect within the
“placebo” sample that uses the same threshold among only the chaebol firms. These findings
reinforce the results according to which the documented effects are driven by the treatment
from the tax reform, and not spuriously driven by differences caused by a violation of parallel
trends that are related to shareholders’ equity or chaebol membership.
To examine whether the results for the full sample are robust to controlling for other
variables, we also report the results from separate specifications with additional control
variables, including Shareholder’s equity - 50 billion wons (i.e., distance to threshold, as in a
regression discontinuity design), size, cash flow, debt, and Cash/assets, and report the results
in Panel C. Here, the results show that the estimated effects remain the same or become
21
even stronger when we control for these additional firm characteristics. To address a concern
about “bad controls” that could confound these results (Angrist and Pischke, p. 64), we do
not use these results as our baseline specification.23
In Panel D, we report the results employing a difference-in-differences matching procedure.
For each treated firm, we pick one control firm that is most similar (the “nearest neighbor”);
we require an exact match on industry, and further match on the pre-period Cash/assets,
payouts, wage increases, and investment based on the Mahalanobis distance. Then, using the
sample of the matched treated-control pairs of firms, we estimate the baseline difference-in-
differences model. Panel D shows that this matching procedure generates similar estimated
treatment effects even despite using a smaller sample.
To further support the parallel trends assumption, in Table 8, we report results pertaining
to differences in “pre-trends” among our main outcome variables between the treated group
and the non-treated group. Panel A reports the results for all firms while Panel B focuses on
the firms that fall within the narrower range around the treatment threshold.
Table 8 About Here
These tests also show that our results are unlikely to be confounded by a violation of
the parallel trends assumption, as we do not observe any significant differential pre-trends
between the treated and untreated firms. This holds both across the full sample, and also
when we focus on the narrower range around the threshold. If anything, we typically observe
pre-trends that run in the opposite way of our main results; this is, for example, the case for
our investment variable in the narrow range, where the treated firms see slightly lower trends
in investment in the years leading up to the law.
23The full sets of the estimated coefficients from our baseline difference-in-differences specifications arereported in Table A5 in the Internet Appendix.
22
We may additionally be concerned about whether firms close to the threshold intentionally
shrink to avoid the tax, and whether such behavior may have an impact on our results.
We address this concern in two ways. First, we examine how many firms move across the
threshold to examine whether a large number of firms show behavior consistent with avoiding
the tax. We find little evidence that firms are shrinking to avoid the surtax; while 39 firms
move from above to below the threshold between 2014 and 2015, 185 firms move from below to
above (the fact that more firms are growing is consistent with overall growth in the economy
over this period). Second, we re-estimate our results using a “Donut RD” empirical strategy,
where we focus on the 10–90B wons range, but further exclude firms in the 45–55 billion
wons range for whom the incentives to shrink to avoid the tax may be the strongest. The
results, which are reported in Table A6 in the Internet Appendix, are similar, and if anything,
generally stronger than in our baseline tests.
Finally, we analyze whether these results differ depending on whether we focus on private
or public firms, or whether the results are unique to either of these groups of firms. Previous
research has shown that private firms overall tend to hold less cash than public firms (Gao
et al., 2013). We nevertheless find no significant differences between private and public firms
in their responses to the law.24
6 Valuation consequences
Our results thus far have shown that the treated firms responded to the reform by saving less
cash, paying out more to shareholders, and spending more on investments and wages. The
question nevertheless remains whether or not firms were saving optimally before the law, and
thus whether the treated firms’ responses to the tax reform represent an improvement or not.
In other words, whether discouraging savings is desirable from a policy perspective depends
24These results are not separately tabulated in the paper but available from the authors.
23
crucially on why firms are saving in the first place: If high savings rates before the law were
optimal, then using the tax system to discourage savings will merely destroy firm value by
distorting firms’ cash policies.25
Theoretically, firms should equate the marginal cost of saving cash with the marginal
benefit of doing so (e.g., Opler et al. (1999); Almeida, Campello, and Weisbach (2004)).
The main cost of saving involves potentially sacrificing profitable investment projects today
and a “double-taxation” penalty from earning interest, whereas a principal benefit is that a
cash buffer can enable firms to better weather recessions and respond to future investment
opportunities (Lins et al., 2010). This is especially true when a firm may face financial
constraints that can make internal financing significantly cheaper than external financing
(Almeida et al., 2004; Acharya et al., 2007). It is challenging to assess whether the treated
firms’ post-reform savings and investment behavior represent an improvement because it is
hard to measure the (shadow) value of cash savings and of any marginal investments. In this
section, we seek to address this question by employing an event study analysis around the
proposal and the passage of the law.
We analyze how investors viewed the valuation consequences of the law for the treated
firms, by employing an event study methodology around the date when the law was proposed
and the date when it was passed. The idea is that we may be able to distinguish between
excessive and optimal savings by looking at abnormal returns around the time the law was
passed. These returns can plausibly capture investors’ expectations regarding the long-term
consequences of the law for firm value, also accounting for expectations for how firms will
react to the law regarding any changes to their savings, payouts, and investments.26
25Previous empirical evidence (Brav et al., 2018; Kaplan, 1989) as well as agency theories (Jensen, 1986)might instead suggest that over-investment rather than excessive savings could be the larger problem.
26Studying announcement returns around several dates related to the passage of the same law in Koreathat we exploit for our study, Semaan (2017) finds that firms that were more likely to be subject to surtaxesexperienced negative abnormal returns, but that this effect was muted for firms that were more likely toengage in tax avoidance.
24
To the extent that high savings and high cash balances before the law were optimal, then
using the tax system to distort firms’ savings and investment policy must destroy firm value,
and we would expect any valuation effects for the treated firms to be negative. This negative
effect consists of two parts: an indirect effect from the tax change distorting corporate policies,
plus a direct effect from paying higher taxes to the extent that firms do not change their
policies sufficiently to avoid the surtax completely. On the other hand, if investors viewed
any changes that firms take in response to the reform to be value-improving, we might expect
somewhat less negative announcement returns to the extent that these indirect effects from
changes in firm policies cancel out the direct negative effects of higher taxes, or even positive
announcement returns if these indirect effects are viewed as sufficiently value-improving and
firms are able to make sufficiently large changes to mostly avoid any surtax. That is, if firms’
savings were excessive before the reform, then discouraging savings through the tax system
could result in improved firm outcomes, and possibly even higher firm values for the treated
firms despite the prospect of higher taxes.
To examine investors’ reactions to the tax reform, we compute the three-day cumulative
abnormal returns (CAR), from one day before to one day after the proposal of the law on
August 6, 2014, as well as the three-day-CARs around the passage of the law on December
2, 2014. We choose these two days because, even though rumors of such a law had existed
before the proposal, these are the first days when the information indicating which firms
would be treated and which firms would not be treated was released. We then take the sum
of these two CARs for each firm and regress it on the Treated indicator.
In Table 9, we report the results from the event study analysis.
Table 9 About Here
25
The results show that abnormal returns (CAR) for the treated firms around the days
when the law was proposed and passed were positive. This implies that investors believed that
the cash accumulation before the law was excessive and that encouraging firms to increase
payouts and investments enhances firm value. The economic magnitudes are estimated to be
around 2% of firm value. These effects are similar regardless of whether we include industry
fixed effects or not, whether we use the full or narrower sample, and whether we limit the
sample to only non-chaebol firms. These results are notably similar to the findings reported
by Christie and Nanda (1994), who find positive market reactions of about 1% around the
announcement of the undistributed profits tax of 1936–37 in the U.S.
The positive CAR for the treated firms is especially notable because the tax itself has a
direct negative valuation effect, as at least some treated firms are likely to pay higher taxes—
specifically, those firms that do not avoid the tax by increasing payouts and investments
sufficiently to avoid the surtax). The positive returns among the treated firms imply that
investors expected the treated firms’ responses to sufficiently enhance their value to more
than outweigh the direct negative effects from higher taxes. These results broadly support
previous findings regarding that investors tend to view a marginal dollar in the hands of
a corporation as being worth less than a dollar (Faulkender and Wang, 2006; Dittmar and
Mahrt-Smith, 2007).
7 Cross-sectional variation and channels
We next investigate whether firms that had high earnings before the reform responded
differently to the tax, as these firms were likely to be more heavily affected by the new surtax.
In the following sections, we use cross-sectional splits to study the mechanism through which
the tax reform affected corporate policies, which in turn can help shed light on why firms
might have been saving excessively in the first place.
26
7.1 High earnings
Firms with high earnings are more heavily affected by the new surtax, and the law also requires
these firms to effect relatively larger changes in their investment and savings behaviors to
minimize the surtax. We might thus expect these firms to change their savings and investment
policies to a greater extent around the reform.
To examine the differential effects on firms with high earnings, we split the sample based
on firms’ pre-reform earnings, measured as operating cash flows (the ratio of EBITDA to
total assets). We define a variable High earnings based on whether these earnings are above
the median. We then estimate triple-difference regressions for the effects on savings, payouts,
and investments, where we interact Treated and After with High earnings. We report the
results in Panel A of Table 10.
Table 10 About Here
Overall, we observe that the treatment effects on corporate policies are significantly more
pronounced for firms that garnered higher earnings in the pre-reform period.
Because the high-earnings firms change their savings and investment policies after the law
more extensively, we might also expect investors to react more strongly to the passage of the
law for these firms—assuming that investors correctly anticipate these reactions. Consistent
with this hypothesis, the results we report in Panel B of Table 10 show that the positive
valuation effects are indeed more pronounced for high-earnings firms.
7.2 Behavioral bias
We next study a possible channel that might explain why Korean firms prior to the enactment
of the law may have exhibited excessive savings in the first place, which is related to possible
27
behavioral biases caused by past experiences. Previous studies have documented links between
past experiences and future financial decisions. For example, Malmendier and Nagel (2011)
and Knupfer, Rantapuska, and Sarvimaki (2017) find that individuals who experienced
depression periods are more likely to be risk-averse in their investments in financial assets.
In the corporate setting, Dittmar and Duchin (2015) find that firms with CEOs who have
worked at a company that has undergone financial difficulties hold more cash.
We similarly examine the role of behavioral biases reflecting a “crisis mentality” among
firms and executives that might have been formed during the 1997 Asian financial crisis. The
idea is that if firms depended heavily on external financing and experienced the crisis, they
might have come out of the crisis with the belief that they need to build ample cash buffers to
weather the next crisis. The basic idea is that either a firm’s CEO may personally remember
the effects of the crisis, or even if the CEO has changed, the firm may have an “institutional
memory” from the crisis. Consistent with higher pre-reform savings for firms with memories
of a past crisis, in Panel A of Table A8 in the Internet Appendix, we report results supporting
a correlation between CrisisMemory and the pre-reform level of cash. Controlling for firm
size and cash flow, we observe in general that firms with a stronger memory from the crisis
did accumulate more cash before the tax reform.
To classify firms according to whether they may suffer from biases from a past crisis,
we define an indicator variable CrisisMemory as equal to one if either 1) the firm’s CEO
was an executive of a firm during the Asian crisis that was operating in an industry that
depends heavily on external financing, or 2) if the firm itself was established before the
crisis and operated in an industry marked by heavy dependence on external financing. The
degree to which firms in an industry depends on external financing is defined, following
Rajan and Zingales (1998), as capital expenditures minus operating cash flow; the use of this
measure within the specific context of the Asian financial crisis is based on Almeida, Kim,
and Kim (2015), who find that Korean firms that operate in industries that depend heavily
28
on external financing experienced more severe liquidity shocks during the crisis. We employ
triple-difference regressions and employ this indicator variable as an additional interaction
term in the baseline difference-in-differences. We hypothesize that firms that fit the definition
of the CrisisMemory variable were more likely to hoard cash excessively before the tax reform
and thus were more heavily affected by the new tax law.
We report the results of our analysis of the CrisisMemory variable in Table 11. In Panel A
we report the triple-difference results regarding the effects on savings, payouts, wage increases,
and investments. We see that firms with a crisis memory reduced cash savings to a greater
extent and also increase payouts, wages, and investments to a greater extent after the tax
reform, although some of these results are statistically significant at the 10% level only. This
suggests that the tax reform had a larger effect in terms of nudging these firms to change
their behavior.
Table 11 About Here
The results we report in Panel B in Table 11 show the differences in market valuation
responses around the law announcements for the crisis-memory firms vs. those firms without
such memories from the previous crisis. The results indicate that the positive announcement
returns (CAR) for treated firms are concentrated among firms with a crisis memory.
As a robustness test, we next confirm that our results capture not only industry-level
differences related to external financing needs, but are related specifically to a firm’s or CEO’s
memories from the past. In Table A7 in the Internet Appendix we report results from a
placebo test where we replace the CrisisMemory variable with a variable indicating whether
a firm operates in an industry with high dependence on external financing but either the
firm or CEO were not operating in 1997 when the Asian financial crisis hit. These placebo
tests show that there is no relationship between firm responses around the reform and their
29
dependence on external financing, but that the results in Table 11 are indeed driven by the
interaction of both being dependent on external finance and remembering the previous crisis.
The results we report in Table A7 similarly show that there are no effects merely involving
“old” firms or CEOs (defined as having been operating in 1997) unless the firm also relied
heavily on external financing.
The results discussed in this section together indicate that firms operating with a crisis
mentality saved more cash before the new reform, but also responded more strongly overall
to the reform, and moreover that these responses were viewed more positively by the market.
These results are consistent with a behavioral hypothesis whereby some firms and CEOs may
have been traumatized by the crisis and therefore engaged in excessive savings.
7.3 Corporate governance
Finally, we study the extent to which differences in firms’ levels of corporate governance
affected how firms and investors responded to the tax reform. Agency conflicts and poor
governance are one of the main reasons why we might expect firms to engage in overly
cautious precautionary savings before the reform (Bertrand and Mullainathan, 2003; Gao,
Harford, and Li, 2013; Nikolov and Whited, 2014). It is thus possible that the reform had
a larger impact on firms with relatively worse governance quality. The results we report in
Table A8 (Panel B) in the Internet Appendix show that poorly governed firms indeed tended
to carry more cash on their balance sheets before the reform.
To measure corporate governance quality, we employ a governance index score created by
the Korea Corporate Governance Service (KCGS). KCGS calculates these scores every year
for all public firms listed in the KOSPI market, and this measure is therefore limited to only
the subset of public firms. This governance score is calculated as a sum that encompasses
four distinct aspects of governance: 1) the protection of shareholder rights, which is given a
30
score between 0 and 81, 2) the internal workings and processes of the board, which is given a
score between 0 and 69, 3) the workings of monitoring organizations, which is given a score
between 0 and 43, and 4) transparency in disclosures, which is given a score between 0 and
47. These scores are then summed to create a total governance score. Note that higher scores
denote increasing governance quality, differing from the GIM-index or E-Index scores in
the U.S., which denote decreasing governance quality. We then define an indicator variable
High-G as equal to one if a firm is assigned an above-median total governance score.
In Panel A of Table 12 we first report results about how differences in corporate governance
quality affected post-reform changes in cash savings, investments, and payouts for the treated
firms.
Table 12 About Here
We find that there are no significant differences in changes in cash savings between poorly
governed and well-governed firms. That is, treated firms in both of these groups reduce their
savings equally after the law was enacted.
How firms used this money that was not saved differs in a predictable pattern, however,
depending on the quality of governance. The positive treatment effects on payouts are
concentrated among well-governed firms: treated firms demonstrating good governance on
average increase payouts following the tax reform by approximately 0.35 percentage points of
assets while poorly governed treated firms do not. On the other hand, the positive treatment
effects on investments and wage increases are much stronger among poorly governed firms.27
This evidence is broadly consistent with “empire-building” among these poorly governed
firms.
27The magnitudes of the average differences in the outcome variables between well-governed and poorlygoverned firms are quite large relative to those of our baseline treatment effects. This is partly a result of thefact that the magnitudes of the “baseline” treatment effects are even larger among this subsample of firms forwhich the governance score is available.
31
As seen in Panel B of Table 12, we further find that the investors’ value reaction to the
reform was much higher for the well-governed firms. This suggests that investors expected the
better-governed firms to respond more efficiently to the tax reform, apparently realizing that
the reform could encourage empire-building, particularly among poorly-governed firms. The
coefficients on the interaction term indicate that the announcement day CAR for a treated
firm is about 2% lower among poorly-governed firms.
This result thus reveals a dark side of the reform. Even if some firms were saving too
much, pushing firms to do something else with that money may not necessarily improve firm
value. For example, instead of hoarding cash, firms could alternatively overinvest or engage
in other forms of “empire-building,” which can result in even worse consequences.
8 Conclusion
Corporations all over the world have accumulated unprecedented levels of cash on their
balance sheets. Many observers say this trend has gone too far and increasingly criticize
firms for “hoarding” cash. Does this cash represents optimal precautionary savings by firms,
or could this money instead be put to more productive uses if firms saved less and instead
spent more on payouts, wages, or new investments? Can tax policy reduce corporate savings?
And if so, what are the consequences? To investigate these questions, we exploit a unique
tax reform in South Korea that explicitly sought to curb corporate savings by imposing a
surtax on excessive savings.
We employ a difference-in-differences methodology to exploit this natural experiment and
show that firms that were discouraged from saving instead spent more on payouts, wages, and
investments. Whether saving less and spending more on payouts and investments is desirable
depends on whether the treated firms’ savings and investment policies were optimal before
the law. Exploiting an event study methodology, we find that the valuation consequences
32
from being treated by the law were positive. This result suggests that investors expected
firms’ responses to the law to be value-improving, indeed sufficient enough to more than
outweigh the direct negative valuation consequences of facing higher taxes.
Overall, these findings are consistent with firms having saved excessively before the
law was enacted. When we investigate two channels that might explain why firms might
have been saving too much, we find evidence consistent with both behavioral biases and
agency conflicts as drivers of excessive savings prior to the reform. Both firms and investors
responded more strongly when a treated firm had a memory from the Asian financial crisis.
Merely encouraging internal investments instead of saving cash, however, is not a silver
bullet: poorly governed firms tend not to increase payouts and instead focus on investments,
and they experience lower valuation effects that are consistent with being perceived as
practicing inefficient empire-building. These findings contribute to the literature on the value
of corporate cash and offer important lessons for future policy.
33
References
Acharya, V. V., H. Almeida, and M. Campello. 2007. Is Cash Negative Debt? A Hedging
Perspective on Corporate Financial Policies. Journal of Financial Intermediation 16:515–
554.
Almeida, H., M. Campello, I. Cunha, and M. S. Weisbach. 2014. Corporate Liquidity
Management: A Conceptual Framework and Survey. Annual Review of Financial Economics
6:135–162.
Almeida, H., M. Campello, and M. S. Weisbach. 2004. The Cash Flow Sensitivity of Cash.
The Journal of Finance 59:1777–1804.
Almeida, H., V. Fos, and M. Kronlund. 2016. The Real Effects of Share Repurchases. Journal
of Financial Economics 119:168–185.
Almeida, H., C.-S. Kim, and H. B. Kim. 2015. Internal Capital Markets in Business Groups:
Evidence from the Asian Financial Crisis. The Journal of Finance 70:2539–2586.
Angrist, J. D., and J.-S. Pischke. 2008. Mostly harmless econometrics: An empiricist’s
companion. Princeton university press.
Azar, J. A., J.-F. Kagy, and M. C. Schmalz. 2016. Can Changes in the Cost of Carry
Explain the Dynamics of Corporate “Cash” Holdings? The Review of Financial Studies
29:2194–2240.
Bates, T. W., K. M. Kahle, and R. M. Stulz. 2009. Why Do US Firms Hold so Much More
Cash Than They Used To? The Journal of Finance 64:1985–2021.
Bertrand, M., and S. Mullainathan. 2003. Enjoying the Quiet Life? Corporate Governance
and Managerial Preferences. Journal of Political Economy 111:1043–1075.
34
Blouin, J., and L. Krull. 2009. Bringing It Home: A Study of the Incentives Surrounding the
Repatriation of Foreign Earnings under the American Jobs Creation Act of 2004. Journal
of Accounting Research 47:1027–1059.
Brav, A., W. Jiang, S. Ma, and X. Tian. 2018. How Does Hedge Fund Activism Reshape
Corporate Innovation? Journal of Financial Economics 130:237–264.
Christie, W. G., and V. Nanda. 1994. Free Cash Flow, Shareholder Value, and the Undis-
tributed Profits Tax of 1936 and 1937. The Journal of Finance 49:1727–1754.
Dharmapala, D., C. F. Foley, and K. J. Forbes. 2011. Watch What I Do, Not What I Say:
The Unintended Consequences of the Homeland Investment Act. The Journal of Finance
66:753–787.
Dittmar, A., and R. Duchin. 2015. Looking in the rearview mirror: The effect of managers’
professional experience on corporate financial policy. The Review of Financial Studies
29:565–602.
Dittmar, A., and J. Mahrt-Smith. 2007. Corporate Governance and the Value of Cash
Holdings. Journal of Financial Economics 83:599–634.
Falato, A., D. Kadyrzhanova, J. Sim, and R. Steri. 2018. Rising Intangible Capital, Shrinking
Debt Capacity, and the US Corporate Savings Glut. Working Paper.
Fama, E. F., and K. R. French. 2001. Disappearing Dividends: Changing Firm Characteristics
or Lower Propensity to Pay? Journal of Financial Economics 60:3–43.
Faulkender, M., and M. Petersen. 2012. Investment and Capital Constraints: Repatriations
under the American Jobs Creation Act. The Review of Financial Studies 25:3351–3388.
Faulkender, M., and R. Wang. 2006. Corporate Financial Policy and the Value of Cash. The
Journal of Finance 61:1957–1990.
35
Faulkender, M. W., K. W. Hankins, and M. A. Petersen. 2019. Understanding the Rise in
Corporate Cash: Precautionary Savings or Foreign Taxes. The Review of Financial Studies
.
Fazzari, S., R. G. Hubbard, and B. Petersen. 1988. Investment, Financing Decisions, and
Tax Policy. The American Economic Review 78:200–205.
Foley, C. F., J. C. Hartzell, S. Titman, and G. Twite. 2007. Why Do Firms Hold so Much
Cash? A Tax-based Explanation. Journal of Financial Economics 86:579–607.
Gamba, A., and A. Triantis. 2008. The Value of Financial Flexibility. The Journal of Finance
63:2263–2296.
Gao, H., J. Harford, and K. Li. 2013. Determinants of Corporate Cash Policy: Insights from
Private Firms. Journal of Financial Economics 109:623–639.
Guay, W., and J. Harford. 2000. The Cash-flow Permanence and Information Content of
Dividend Increases Versus Repurchases. Journal of Financial Economics 57:385–415.
Harford, J. 1999. Corporate Cash Reserves and Acquisitions. The Journal of Finance
54:1969–1997.
Jagannathan, M., C. P. Stephens, and M. S. Weisbach. 2000. Financial Flexibility and
the Choice between Dividends and Stock Repurchases. Journal of Financial Economics
57:355–384.
Jensen, M. C. 1986. Agency Costs of Free Cash Flow, Corporate Finance, and Takeovers.
The American Economic Review 76:323–329.
Kalcheva, I., and K. V. Lins. 2007. International Evidence on Cash Holdings and Expected
Managerial Agency Problems. The Review of Financial Studies 20:1087–1112.
36
Kaplan, S. 1989. The effects of management buyouts on operating performance and value.
Journal of Financial Economics 24:217 – 254.
Knupfer, S., E. Rantapuska, and M. Sarvimaki. 2017. Formative experiences and portfolio
choice: Evidence from the Finnish great depression. The Journal of Finance 72:133–166.
Lins, K. V., H. Servaes, and P. Tufano. 2010. What Drives Corporate Liquidity? an
International Survey of Cash Holdings and Lines of Credit. Journal of Financial Economics
98:160–176.
Lyandres, E., and B. Palazzo. 2016. Cash Holdings, Competition, and Innovation. Journal of
Financial and Quantitative Analysis 51:1823–1861.
Malmendier, U., and S. Nagel. 2011. Depression Babies: Do Macroeconomic Experiences
Affect Risk Taking? The Quarterly Journal of Economics 126:373–416.
Malmendier, U., G. Tate, and J. Yan. 2011. Overconfidence and Early-Life Experiences:
The Effect of Managerial Traits on Corporate Financial Policies. The Journal of Finance
66:1687–1733.
Nikolov, B., and T. M. Whited. 2014. Agency Conflicts and Cash: Estimates from a Dynamic
Model. The Journal of Finance 69:1883–1921.
Opler, T., L. Pinkowitz, R. Stulz, and R. Williamson. 1999. The Determinants and Implica-
tions of Corporate Cash Holdings. Journal of Financial Economics 52:3–46.
Pinkowitz, L., R. M. Stulz, and R. Williamson. 2015. Do US Firms Hold More Cash Than
Foreign Firms Do? The Review of Financial Studies 29:309–348.
Poterba, J. M., R. E. Hall, and R. G. Hubbard. 1987. Tax Policy and Corporate Saving.
Brookings Papers on Economic Activity 1987:455–515.
37
Rajan, R. G., and L. Zingales. 1998. Financial Dependence and Growth. The American
Economic Review 88:559–586.
Riddick, L. A., and T. M. Whited. 2009. The Corporate Propensity to Save. The Journal of
Finance 64:1729–1766.
Semaan, S. 2017. Tax Avoidance, Income Diversion, and Shareholder Value: Evidence from
a Quasi-Natural Experiment. Working Paper.
38
Figure ITime-series Trends in Outcome Variables
This figure plots differential trends over the years in cash savings (Panel A), payout (PanelB), wage increases (Panel C), and investments (Panel D) between treated and control firms.For each treated firm, we pick one control firm that is most similar (its “nearest neighbor”);we require an exact match on industry, and further matches on the pre-period ∆Cash/assets,payouts, wage increases, and investment based on the Mahalanobis distance. All outcomevariables are defined in Table A9 in the Internet Appendix.
39
Table 1Summary Statistics
In this table we report summary statistics. In Panel A we report firm-year-level statistics onfirm characteristics. In Panel B we report the number of firms in each combination of thetreatment criteria as of 2014. All variables are defined in Table A9 in the Internet Appendix.
Mean SD P1 P25 P50 P75 P99 N
Panel A. Firm characteristics (All firms)
Treated 0.17 0.37 0.00 0.00 0.00 0.00 1.00 80494After 0.53 0.50 0.00 0.00 1.00 1.00 1.00 804941{Equity ≥ 50 billion} 0.15 0.35 0.00 0.00 0.00 0.00 1.00 80494Chaebol 0.05 0.21 0.00 0.00 0.00 0.00 1.00 80494∆Cash/assets (%) 1.08 7.94 -23.67 -1.40 0.10 2.67 37.08 74016Payout (%) 0.75 2.38 0.00 0.00 0.00 0.00 16.03 74390
Dividend (%) 0.56 1.84 0.00 0.00 0.00 0.00 12.44 74390Repurchase (%) 0.10 0.61 0.00 0.00 0.00 0.00 5.04 74390
Wage increase (%) 0.36 1.46 -4.17 -0.13 0.14 0.67 7.30 71809Investment (%) 3.05 10.08 -16.20 -0.79 0.26 3.58 41.57 70592
Investment in land (%) 1.20 6.33 -15.83 0.00 0.00 0.00 38.21 72626Investment in building (%) 1.08 4.80 -9.29 0.00 0.00 0.16 30.18 72627Investment in equipment (%) 1.24 4.18 -11.20 0.00 0.00 1.07 24.33 72631
Earnings (%) 2.51 11.11 -43.36 -0.19 2.38 6.83 38.42 74322Size 24.23 1.15 22.44 23.41 23.92 24.78 28.31 80494Debt (%) 31.75 27.87 0.00 6.71 27.95 48.95 118.79 80494G-index 103.55 22.56 52.00 91.00 101.00 113.00 174.00 2513ROIC (%) 7.74 15.61 -44.68 1.50 5.65 11.79 81.80 74390Profit margin (%) 21.27 20.17 -37.29 9.75 17.03 28.83 91.77 73653PD in 2yrs (%) 0.38 0.55 0.00 0.06 0.20 0.46 3.25 6467PD in 3yrs (%) 0.58 0.72 0.00 0.13 0.36 0.73 4.04 6467CARproposal+CARpassage 0.01 0.07 -0.18 -0.03 0.00 0.04 0.23 6569
Panel B. Number of firms in each combination of treatment criteria
Chaebol Non-chaebol TotalSEQ ≥ 50 billion 599 2,852 3,451SEQ < 50 billion 453 17,012 17,465Total 1,052 19,864 20,916
40
Table 2Treatment Effects on Changes in Cash
In this table we report the results of difference-in-differences regressions regarding the treatmenteffects of the tax reform on changes in cash holdings. We use the entire sample to obtain theresults reported in columns 1 and 2 while for columns 3 and 4 we use firms in the narrowerbandwidth around the 50-billion-wons equity threshold. All variables are defined in Table A9in the Internet Appendix. We include firm and year or industry-year fixed effects as indicated.t-statistics (in parentheses) are calculated using standard errors that are heteroskedasticity-robust and clustered by firm or chaebol -group. ∗∗∗, ∗∗ , and ∗ indicate statistical significance atthe 1%, 5%, and 10% levels, respectively.
Dependent variable: ∆Cash/assets
(1) (2) (3) (4)
Sample All firms Firms in [10B, 90B]
Treated × After -0.282** -0.291* -0.449** -0.442**(-2.06) (-1.95) (-2.23) (-2.03)
Firm FE Yes Yes Yes YesYear FE Yes No Yes NoIndustry × Year FE No Yes No Yes
N 71,476 70,504 32,094 31,027R2 0.1910 0.2447 0.1880 0.2781
41
Table 3Treatment Effects on Payouts
In this table we report the results of difference-in-differences regressions regarding the treatmenteffects of the tax reform on total payouts. To obtain the results reported in columns 1 and 2 weuse the entire sample while for columns 3 and 4 we use firms in the narrower bandwidth aroundthe 50-billion-wons equity threshold. All variables are defined in Table A9 in the InternetAppendix. We include firm and year or industry-year fixed effects in the specifications. t-statistics (in parentheses) are calculated using standard errors that are heteroskedasticity-robustand clustered by firm or chaebol -group. ∗∗∗, ∗∗ , and ∗ indicate statistical significance at the1%, 5%, and 10% levels, respectively.
Dependent variable: Payout
(1) (2) (3) (4)
Sample All firms Firms in [10B, 90B]
Treated × After 0.212*** 0.216*** 0.208*** 0.219***(5.27) (5.03) (3.12) (3.06)
Firm FE Yes Yes Yes YesYear FE Yes No Yes NoIndustry × Year FE No Yes No Yes
N 71,883 70,912 32,134 31,071R2 0.6206 0.6452 0.6275 0.6679
42
Table 4Treatment Effects on Wage Increases
In this table we report the results of difference-in-differences regressions regarding the treatmenteffects of the tax reform on wage increases. To obtain the results reported in columns 1 and 2 weuse the entire sample while for columns 3 and 4 we use firms in the narrower bandwidth aroundthe 50-billion-wons equity threshold. All variables are defined in Table A9 in the InternetAppendix. We include firm and year or industry-year fixed effects in the specifications. t-statistics (in parentheses) are calculated using standard errors that are heteroskedasticity-robustand clustered by firm or chaebol -group. ∗∗∗, ∗∗ , and ∗ indicate statistical significance at the1%, 5%, and 10% levels, respectively.
Dependent variable: Wage increase
(1) (2) (3) (4)
Sample All firms Firms in [10B, 90B]
Treated × After 0.173*** 0.161*** 0.122*** 0.098***(8.02) (6.53) (3.68) (2.69)
Firm FE Yes Yes Yes YesYear FE Yes No Yes NoIndustry × Year FE No Yes No Yes
N 69,329 68,412 31,580 30,552R2 0.3889 0.4358 0.3946 0.4616
43
Table 5Treatment Effects on Investments
In this table we report the results of difference-in-differences regressions regarding the treatmenteffects of the tax reform on investments. To obtain the results reported in columns 1 and 2 weuse the entire sample while for columns 3 and 4 we use firms in the narrower bandwidth aroundthe 50-billion-wons equity threshold. All variables are defined in Table A9 in the InternetAppendix. We include firm and year or industry-year fixed effects in the specifications. t-statistics (in parentheses) are calculated using standard errors that are heteroskedasticity-robustand clustered by firm or chaebol -group. ∗∗∗, ∗∗ , and ∗ indicate statistical significance at the1%, 5%, and 10% levels, respectively.
Dependent variable: Investment
(1) (2) (3) (4)
Sample All firms Firms in [10B, 90B]
Treated × After 0.494*** 0.480*** 0.694*** 0.712***(2.89) (2.69) (2.90) (2.59)
Firm FE Yes Yes Yes YesYear FE Yes No Yes NoIndustry × Year FE No Yes No Yes
N 65,887 64,934 30,447 29,416R2 0.4077 0.4461 0.3934 0.4575
44
Table 6Financial Distress
In this table we report the results of difference-in-differences regressions regarding the treatmenteffects of the tax reform on financial distress proxied by financial leverage or the probability ofdefault. To obtain the results reported in Panel A we use the entire sample while for Panel B weuse firms in the narrower bandwidth around the 50-billion-wons equity threshold. All variablesare defined in Table A9 in the Internet Appendix. We include firm and year or industry-yearfixed effects in the specifications. t-statistics (in parentheses) are calculated using standarderrors that are heteroskedasticity-robust and clustered by firm or chaebol -group. ∗∗∗, ∗∗ , and ∗
indicate statistical significance at the 1%, 5%, and 10% levels, respectively.
(1) (2) (3) (4) (5) (6)
Dependent variable Leverage PD in 2yrs PD in 3yrs
Panel A: All firms
Treated × After 1.093*** 0.896*** 0.000 0.000 0.000 0.000(5.30) (3.74) (1.42) (0.89) (1.44) (0.80)
N 84,885 83,983 6,465 5,581 6,465 5,581R2 0.8713 0.8794 0.7253 0.7816 0.7398 0.7944
Panel B: Firms in [10B, 90B] (Discontinuity test)
Treated × After 0.581* 0.773** 0.000 -0.000 0.000 -0.000(1.90) (2.34) (0.64) (-0.09) (0.65) (-0.14)
N 35,744 34,714 3,187 2,449 3,187 2,449R2 0.8450 0.8625 0.6581 0.7629 0.6791 0.7776
Firm FE Yes Yes Yes Yes Yes YesYear FE Yes No Yes No Yes NoIndustry × Year FE No Yes No Yes No Yes
45
Table 7Robustness of the Treatment Effects
In this table we report the results of difference-in-differences (DID) regressions regarding thetreatment effects of the tax reform on each of our dependent variables across several combina-tions of samples and specifications: the baseline DID regressions within only non-chaebol firmsin the narrower bandwidth around the 50-billion-wons equity threshold (Panel A), a “placebotest” among chaebol firms around the 50 billion threshold within the narrower bandwidth(Panel B), the baseline DID regressions while controlling for additional firm characteristics,including distance to threshold, size, earnings, debt, and Cash/assets (Panel C), and matchingDID regressions where we match each treated firm with one control firm of the nearest neighbor(based on the Mahalanobis distance) on industry as well as on pre-period cash, payouts, wageincreases, and investments (Panel D). All variables are defined in Table A9 in the InternetAppendix. We include firm and industry-year fixed effects as indicated. t-statistics (in paren-theses) are calculated using standard errors that are heteroskedasticity-robust and clustered byfirm or chaebol -group. ∗∗∗, ∗∗ , and ∗ indicate statistical significance at the 1%, 5%, and 10%levels, respectively.
Dependent variable ∆Cash/assets Payout Wage increase Investment
(1) (2) (3) (4)
Panel A: Non-chaebol firms in [10B, 90B]
1{Equity ≥ 50 billion} × After -0.489* 0.250*** 0.082** 0.591**(-1.70) (3.16) (2.06) (1.97)
N 29,700 29,735 29,287 28,264R2 0.2794 0.6691 0.4589 0.4530
Panel B: Placebo, Chaebol firms in [10B, 90B]
1{Equity ≥ 50 billion} × After -0.104 -0.203 -0.060 -3.958**(-0.04) (-0.46) (-0.36) (-2.12)
N 758 758 730 623R2 0.4303 0.7523 0.6083 0.6914
Firm FE Yes Yes Yes YesIndustry × Year FE Yes Yes Yes Yes
46
Dependent variable ∆Cash/assets Payout Wage increase Investment
(1) (2) (3) (4)
Panel C: Additional control variables
Treated × After -0.737*** 0.289*** 0.074** 1.566***(-3.64) (4.89) (2.12) (6.06)
Firm FE Yes Yes Yes YesIndustry × Year FE Yes Yes Yes YesControls Yes Yes Yes Yes
N 70,517 70,611 68,340 64,862R2 0.6148 0.6488 0.4463 0.4790
Panel D: Matched treated-control pairs
Treated × After -0.447* 0.149** 0.081* 0.807***(-1.93) (2.48) (1.77) (2.69)
N 24,172 24,193 24,133 24,054R2 0.1908 0.6307 0.3970 0.3903
47
Table
8.
Pre
-tre
nds
Inth
ista
ble
we
rep
ort
the
resu
lts
ofcr
oss-
sect
ional
regr
essi
ons
onth
ediff
eren
ces
inpre
-ref
orm
aver
age
chan
ges
inth
eou
tcom
eva
riab
les
bet
wee
nth
etr
eatm
ent
and
the
contr
olgr
oups.
The
dep
enden
tva
riab
les
are
the
aver
age
ofan
nual
chan
ges
inth
eou
tcom
eva
riab
les
in20
13an
d20
14.
To
obta
inth
ere
sult
sre
por
ted
inP
anel
Aw
euse
the
enti
resa
mple
while
for
Pan
elB
we
use
firm
sin
the
nar
row
erban
dw
idth
arou
nd
the
50-b
illion
-won
seq
uit
yth
resh
old.
All
vari
able
sar
edefi
ned
inT
able
A9
inth
eIn
tern
etA
pp
endix
.W
ein
clude
indust
ryfixed
effec
tsin
the
spec
ifica
tion
s.t-
stat
isti
cs(i
npar
enth
eses
)ar
eca
lcula
ted
usi
ng
stan
dar
der
rors
that
are
het
eros
kedas
tici
ty-r
obust
and
clust
ered
by
firm
orch
aebo
l-gr
oup.
∗∗∗ ,
∗∗,
and
∗in
dic
ate
stat
isti
cal
sign
ifica
nce
atth
e1%
,5%
,an
d10
%le
vels
,re
spec
tive
ly.
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Dep
enden
tva
riab
leA
vg.
∆(∆
Cas
h/a
sset
s)A
vg.
∆P
ayou
tA
vg.
∆W
age
incr
ease
Avg.
∆In
vest
men
t
Pan
elA
:A
llfirm
s
Tre
ated
-0.0
78-0
.165
-0.0
50-0
.017
-0.0
34-0
.031
-0.3
75-0
.129
(-0.
22)
(-0.
45)
(-0.
89)
(-0.
28)
(-0.
67)
(-0.
58)
(-0.
87)
(-0.
27)
Con
stan
t-0
.029
-0.5
33-0
.910
**-4
.760
(-0.
01)
(-1.
28)
(-2.
16)
(-1.
25)
Con
trol
sY
esY
esY
esY
esY
esY
esY
esY
esIn
dust
ryF
EN
oY
esN
oY
esN
oY
esN
oY
es
N16
,003
15,7
4516
,095
15,8
3815
,539
15,2
9515
,298
15,0
51R
20.
0006
0.07
460.
0026
0.06
440.
0007
0.08
330.
0071
0.07
54
Pan
elB
:F
irm
sin
[10B
,90
B]
(Dis
conti
nuit
yte
st)
Tre
ated
-0.1
57-0
.171
-0.1
01-0
.054
-0.0
17-0
.039
-0.9
79*
-1.0
74*
(-0.
32)
(-0.
42)
(-1.
22)
(-0.
61)
(-0.
27)
(-0.
57)
(-1.
83)
(-1.
76)
Con
stan
t-1
.296
-2.6
73**
-0.2
85-1
3.88
2*(-
0.21
)(-
2.53
)(-
0.34
)(-
1.85
)
Indust
ryF
EN
oY
esN
oY
esN
oY
esN
oY
es
N7,
622
7,34
77,
633
7,35
97,
514
7,25
17,
400
7,13
5R
20.
0002
0.12
040.
0040
0.10
230.
0004
0.11
800.
0051
0.11
43
48
Table 9Value Implication of the Tax Reform
In this table we report the differences in cumulative abnormal returns around the proposaland the passage of the law between treated and nontreated firms. To obtain the results re-ported in Panel A we use the entire sample while for Panel B we use firms in the narrowerbandwidth around the 50-billion-wons equity threshold. We include industry fixed effects inthe specifications. t-statistics (in parentheses) are calculated using standard errors that areheteroskedasticity-robust and clustered by firm or chaebol -group. ∗∗∗, ∗∗ , and ∗ indicate statis-tical significance at the 1%, 5%, and 10% levels, respectively.
(1) (2) (3) (4)
Dependent variable CARproposal+CARpassage (-1 to +1 days)
Sample Chaebol + Non-chaebol Non-chaebol only
Panel A: All firms
Treated 0.018*** 0.019*** 0.018*** 0.016***(4.62) (3.84) (4.53) (3.12)
Constant -0.006 -0.006(-1.59) (-1.59)
Industry FE No Yes No Yes
N 1,650 1,451 1,422 1,236R2 0.0158 0.2313 0.0165 0.2459
Panel B: Firms in [10B, 90B] (Discontinuity test)
Treated 0.017*** 0.020*** 0.016*** 0.017**(3.00) (2.63) (2.77) (2.34)
Constant -0.004 -0.004(-1.08) (-1.08)
Industry FE No Yes No Yes
N 811 641 786 616R2 0.0107 0.2411 0.0092 0.2544
49
Table 10Earnings and Treatment/Value Effects
In Panels A and B, respectively, we report the results of panel and cross-sectional regres-sions on the cross-sectional differences across firms regarding the treatment and value effectsdepending on whether a firm’s pre-reform earnings are above or below the median (High earn-ings) where earnings is proxied by earnings before interest, taxes, depreciation, and amorti-zation (EBITDA). t-statistics (in parentheses) are calculated using standard errors that areheteroskedasticity-robust and clustered by firm or chaebol -group. ∗∗∗, ∗∗ , and ∗ indicate statis-tical significance at the 1%, 5%, and 10% levels, respectively.
Panel A: Treatment effects and earnings
(1) (2) (3) (4)
Dependent variable ∆Cash/assets Payout Wage increase Investment
Treated × After 0.029 0.110** 0.060* -0.042(0.14) (2.55) (1.88) (-0.13)
High earnings × After -0.199 -0.065** -0.193*** -0.874***(-1.57) (-2.23) (-7.53) (-4.05)
Treated × After × High earnings -0.680** 0.184** 0.158*** 0.816**(-2.32) (2.51) (3.50) (1.97)
Firm FE Yes Yes Yes YesIndustry × Year FE Yes Yes Yes Yes
N 62,977 63,409 60,463 59,500R2 0.1722 0.6130 0.4009 0.3923
50
Panel B: Value implications and earnings
(1) (2)
Dependent variable CARproposal+CARpassage (-1 to +1 days)
Sample Chaebol + Non-chaebol Non-chaebol only
Treated 0.010 0.004(1.52) (0.64)
High earnings -0.005 -0.004***(-0.56) (-0.53)
Treated × High earnings 0.017* 0.021**(1.89) (2.20)
Industry FE Yes Yes
N 1,443 1,229R2 0.2383 0.2550
51
Table 11The Role of Crisis Mentality
In Panels A and B, respectively, we report the results of panel and cross-sectional regressions onthe cross-sectional differences across firms regarding the treatment and value effects dependingon whether a CEO or firm itself was operating in an industry featuring above-median externalfinance dependence (EFD) during the 1997 Asian financial crisis (CrisisMemory). t-statistics(in parentheses) are calculated using standard errors that are heteroskedasticity-robust andclustered by firm or chaebol -group. ∗∗∗, ∗∗ , and ∗ indicate statistical significance at the 1%,5%, and 10% levels, respectively.
Panel A: Treatment effects and crisis mentality
(1) (2) (3) (4)
Dependent variable ∆Cash/assets Payout Wage increase Investment
Treated × After -0.142 0.135** 0.108*** 0.371*(-0.98) (2.37) (2.75) (1.69)
CrisisMemory × After 0.189 0.057 0.014 -0.430**(0.47) (0.64) (0.21) (-2.54)
Treated × After × CrisisMemory -0.279* 0.146* 0.110** 0.599*(-1.86) (1.72) (1.98) (1.95)
Firm FE Yes Yes Yes YesIndustry × Year FE Yes Yes Yes Yes
N 71,083 71,358 68,885 67,067R2 0.1955 0.6281 0.4157 0.4153
52
Panel B: Value implications and crisis mentality
(1) (2)
Dependent variable CARproposal+CARpassage (-1 to +1 days)
SSample Chaebol + Non-chaebol Non-chaebol only
Treated 0.010 0.007(1.53) (0.97)
CrisisMemory -0.023** -0.026***(-2.42) (-2.77)
Treated × CrisisMemory 0.017* 0.018*(1.94) (1.90)
Industry FE Yes Yes
N 1,451 1,236R2 0.2316 0.2485
53
Table 12The Role of Governance
In Panels A and B, respectively, we report the results of panel and cross-sectional regressions onthe cross-sectional differences across firms regarding the treatment and value effects dependingon whether a firm’s governance score (G-index) is above the median or not (High-G). t-statistics(in parentheses) are calculated using standard errors that are heteroskedasticity-robust andclustered by firm or chaebol -group. ∗∗∗, ∗∗ , and ∗ indicate statistical significance at the 1%,5%, and 10% levels, respectively.
Panel A: Treatment effects and governance
(1) (2) (3) (4)
Dependent variable ∆Cash/assets Payout Wage increase Investment
Treated × After -0.601 0.050 0.298** 2.722*(-0.72) (0.61) (2.01) (1.91)
High-G × After 2.053 -0.283* 0.245** 2.586**(1.50) (-1.77) (2.11) (2.05)
Treated × After × High-G -1.777 0.350** -0.213** -2.661**(-1.26) (2.06) (-1.99) (-2.13)
Firm FE Yes Yes Yes YesIndustry × Year FE Yes Yes Yes Yes
N 2,521 2,521 2,514 2,471R2 0.1779 0.7595 0.4521 0.3707
54
Panel B: Value implications and governance
(1) (2)
Dependent variable CARproposal+CARpassage (-1 to +1 days)
Sample Chaebol + Non-chaebol Non-chaebol only
Treated 0.010 0.012(1.44) (1.54)
High-G -0.044 -0.038(-1.63) (-1.53)
Treated × High-G 0.021** 0.020**(2.05) (2.01)
Industry FE Yes Yes
N 494 342R2 0.3218 0.3577
55
Appendix
Table A1Treatment Effects on Payout Components
In this table we report the results of difference-in-differences regressions regarding the treat-ment effects of the tax reform on the components of total payouts: cash dividends (Panel A)and share repurchases (Panel B). To obtain the results reported in columns 1 and 2 we usethe entire sample while for columns 3 and 4 we use firms in the narrower bandwidth aroundthe 50-billion-wons equity threshold. All variables are defined in Table A9 in the Internet Ap-pendix. We include firm and year or industry-year fixed effects in the specifications. t-statistics(in parentheses) are calculated using standard errors that are heteroskedasticity-robust andclustered by firm or chaebol -group. ∗∗∗, ∗∗ , and ∗ indicate statistical significance at the 1%,5%, and 10% levels, respectively.
(1) (2) (3) (4)
Sample All firms Firms in [10B, 90B]
Panel A Dependent variable: Dividend
Treated × After 0.135*** 0.149*** 0.108** 0.117**(4.42) (4.81) (2.08) (2.16)
N 71,883 70,912 32,134 31,071R2 0.7121 0.7322 0.7115 0.7430
Panel B Dependent variable: Repurchase
Treated × After 0.037*** 0.031** 0.041** 0.041*(2.78) (2.16) (2.02) (1.84)
N 71,883 70,912 32,134 31,071R2 0.3603 0.3958 0.3519 0.4124
Firm FE Yes Yes Yes YesYear FE Yes No Yes NoIndustry × Year FE No Yes No Yes
56
Table A2Treatment Effects on Average Wages and Number of Employees
In this table we report the results of difference-in-differences regressions regarding the treatmenteffects of the tax reform on three employment-related variables: Wage per employee, defined asthe natural logarithm of total wage bills divided by the number of employees (columns 1 and2); Number of employees, defined as the natural logarithm of the total number of employees(columns 3 and 4); and N.employees/assets, defined as the total number of employees over totalassets multiplied by 1,000,000 (columns 5 and 6). We include firm and year or industry-yearfixed effects in the specifications. t-statistics (in parentheses) are calculated using standarderrors that are heteroskedasticity-robust and clustered by firm or chaebol -group. ∗∗∗, ∗∗ , and ∗
indicate statistical significance at the 1%, 5%, and 10% levels, respectively.
(1) (2) (3) (4) (5) (6)
Dependent variable Wage per employee Number of employees N.employees/assets
Panel A: All firms
Treated × After 0.072*** 0.065** 0.018 0.036 0.006 0.006(3.27) (2.46) (0.90) (1.63) (1.37) (1.19)
N 8,219 7,353 8,249 7,381 8,249 7,381R2 0.9217 0.9331 0.9650 0.9674 0.9141 0.9285
Panel B: Firms in [10B, 90B] (Discontinuity test)
Treated × After 0.029 -0.013 0.035 0.063** 0.016*** 0.022***(1.02) (-0.37) (1.44) (2.15) (3.04) (3.28)
N 3,827 3,074 3,838 3,085 3,838 3,085R2 0.8722 0.8998 0.9237 0.9394 0.8977 0.9151
Firm FE Yes Yes Yes Yes Yes YesYear FE Yes No Yes No Yes NoIndustry × Year FE No Yes No Yes No Yes
57
Table A3Treatment Effects on the Components of Investments
In this table we report the results of difference-in-differences regressions regarding the treatmenteffects of the tax reform on each component of total investments, i.e. investments in land,building, and equipment. We include all firms in the sample. All variables are defined in TableA9. We include firm and year or industry-year fixed effects in the specifications. t-statistics(in parentheses) are calculated using standard errors that are heteroskedasticity-robust andclustered by firm or chaebol -group. ∗∗∗, ∗∗ , and ∗ indicate statistical significance at the 1%,5%, and 10% levels, respectively.
(1) (2) (3) (4) (5) (6)
Dependent variable Investment in land Investment in building Investment in equipment
Treated × After 0.210** 0.143 0.192** 0.197** 0.102 0.139*(2.25) (1.40) (2.48) (2.29) (1.28) (1.75)
Firm FE Yes Yes Yes Yes Yes YesYear FE Yes No Yes No Yes NoIndustry × Year FE No Yes No Yes No Yes
N 67,905 66,979 67,907 66,981 67,912 66,986R2 0.3063 0.3468 0.3105 0.3569 0.4512 0.4873
58
Table A4Operating Performance
In this table we report the results of difference-in-differences regressions regarding the treatmenteffects of the tax reform on operating performance, as measured by return on invested capital(ROIC ) and profit margin. To obtain the results reported in Panel A we use the entire samplewhile for Panel B we use firms in the narrower bandwidth around the 50-billion-wons equitythreshold. All variables are defined in Table A9. We include firm and year or industry-yearfixed effects in the specifications. t-statistics (in parentheses) are calculated using standarderrors that are heteroskedasticity-robust and clustered by firm or chaebol -group. ∗∗∗, ∗∗ , and ∗
indicate statistical significance at the 1%, 5%, and 10% levels, respectively.
(1) (2) (3) (4)
Dependent variable ROIC Profit margin
Panel A: All firms
Treated × After 0.598*** 0.442* 0.839*** 0.586***(2.89) (1.91) (4.78) (3.13)
N 71,058 70,076 73,319 72,467R2 0.6412 0.6662 0.8702 0.8792
Panel B: Firms in [10B, 90B] (Discontinuity test)
Treated × After 0.809*** 0.820** 0.497* 0.491*(2.58) (2.35) (1.91) (1.85)
N 32,134 31,071 31,151 30,135R2 0.6744 0.7073 0.8983 0.9102
Firm FE Yes Yes Yes YesYear FE Yes No Yes NoIndustry × Year FE No Yes No Yes
59
Table A5Robustness of Treatment Effects: Controlling for Firm Characteristics
In this table we report the robustness results of difference-in-differences regressions regarding thetreatment effects of the tax reform while controlling for various firm characteristics including thedistance-to-the-equity-threshold, i.e. shareholders’ equity minus 50 billion wons. All variablesare defined in Table A9. We include firm and industry-year fixed effects, and firm characteristicsas control variables in the specifications. t-statistics (in parentheses) are calculated usingstandard errors that are heteroskedasticity-robust and clustered by firm or chaebol -group. ∗∗∗,∗∗ , and ∗ indicate statistical significance at the 1%, 5%, and 10% levels, respectively.
(1) (2) (3) (4)
Dependent variable ∆Cash/assets Payout Wage increase Investment
Treated × After -0.737*** 0.289*** 0.074** 1.566***(-3.64) (4.89) (2.12) (6.06)
Equity - 50 billion wons 0.000** 0.000*** 0.000 0.000***(2.23) (3.40) (0.52) (4.39)
Size 5.063*** 0.032 0.568*** 9.030***(24.48) (0.61) (14.12) (28.00)
CF 0.128*** 0.017*** 0.020*** 0.135***(15.87) (7.90) (10.82) (13.13)
Debt 0.045*** 0.001 0.000 0.082***(13.92) (1.09) (0.53) (14.44)
Cash 1.078*** 0.012*** 0.005*** -0.092***(99.37) (4.39) (2.87) (-10.81)
Equity - 50 billion wons × After -0.000 -0.000 -0.000 0.000(-0.03) (-1.16) (-0.79) (0.16)
Size × After 0.057 -0.062*** 0.010 -1.044***(0.88) (-3.30) (0.75) (-10.51)
CF × After -0.070*** 0.004 -0.007*** -0.062***(-8.01) (1.60) (-3.75) (-5.75)
Debt × After -0.013*** -0.001** -0.003*** -0.033***(-5.19) (-2.15) (-5.23) (-8.91)
Cash × After -0.179*** 0.002 -0.001 0.017**(-17.39) (0.92) (-0.64) (2.26)
Firm FE Yes Yes Yes YesIndustry × Year FE Yes Yes Yes Yes
N 70,517 70,611 68,340 64,862R2 0.6148 0.6488 0.4463 0.4790
60
Table
A6.
Tre
atm
ent
Eff
ect
sin
aDonut
Dis
conti
nuit
ySam
ple
Inth
ista
ble
we
rep
ort
the
resu
lts
ofdiff
eren
ce-i
n-
diff
eren
ces
regr
essi
ons
rega
rdin
gth
etr
eatm
ent
effec
tsof
the
tax
refo
rm.
To
obta
inth
ere
sult
sre
por
ted
inP
anel
Aw
euse
all
firm
sin
our
nar
row
erban
dw
idth
arou
nd
the
50-b
illion
-won
seq
uit
yth
resh
old
excl
udin
gth
ose
inth
en
arro
wer
ban
dw
idth
arou
nd
the
thre
shol
d,
while
for
Pan
elB
we
use
non
-cha
ebol
firm
sin
the
sam
edo
nu
tban
dw
idth
arou
nd
the
thre
shol
d.
All
vari
able
sar
edefi
ned
inT
able
A9.
We
incl
ude
firm
and
year
orin
dust
ry-y
ear
fixed
effec
tsin
the
spec
ifica
tion
s.t-
stat
isti
cs(i
npar
enth
eses
)ar
eca
lcula
ted
usi
ng
stan
dar
der
rors
that
are
het
eros
kedas
tici
ty-r
obust
and
clust
ered
by
firm
orch
aebo
l-gr
oup.
∗∗∗ ,
∗∗,
and
∗
indic
ate
stat
isti
cal
sign
ifica
nce
atth
e1%
,5%
,an
d10
%le
vels
,re
spec
tive
ly.
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Dep
enden
tva
riab
le∆
Cas
h/a
sset
sP
ayou
tW
age
incr
ease
Inve
stm
ent
Pan
elA
:A
llfirm
sin
[10B
,45
B)
U(5
5B,
90B
]
Tre
ated×
Aft
er-0
.659
**-0
.631
**0.
149*
*0.
157*
*0.
131*
**0.
105*
**0.
674*
*0.
822*
**(-
2.48
)(-
2.14
)(2
.07)
(1.9
9)(3
.73)
(2.6
4)(2
.47)
(2.6
4)
Fir
mF
EY
esY
esY
esY
esY
esY
esY
esY
esY
ear
FE
Yes
No
Yes
No
Yes
No
Yes
No
Indust
ry×
Yea
rF
EN
oY
esN
oY
esN
oY
esN
oY
es
N30
,125
29,0
3830
,163
29,0
8029
,634
28,5
8629
,226
28,1
68R
20.
1870
0.28
130.
6294
0.67
110.
3942
0.46
450.
3617
0.43
10
Pan
elB
:N
on-c
haeb
olfirm
sin
[10B
,45
B)
U(5
5B,
90B
]
Tre
ated×
Aft
er-0
.628
**-0
.554
*0.
186*
*0.
188*
*0.
120*
**0.
094*
*0.
749*
*0.
751*
*(-
2.15
)(-
1.72
)(2
.37)
(2.1
3)(3
.00)
(2.1
2)(2
.40)
(2.1
8)
Fir
mF
EY
esY
esY
esY
esY
esY
esY
esY
esY
ear
FE
Yes
No
Yes
No
Yes
No
Yes
No
Indust
ry×
Yea
rF
EN
oY
esN
oY
esN
oY
esN
oY
es
N28
,935
27,8
6528
,970
27,9
0228
,499
27,4
6628
,172
27,1
24R
20.
1883
0.28
330.
6303
0.67
250.
3912
0.46
070.
3564
0.42
83
61
Table A7Placebo Tests on the Role of Crisis Mentality
Represented in panels A and B, respectively, are the results of panel and cross-sectional regres-sions on the cross-sectional differences across firms regarding the treatment and value effects.In each panel we report the results from two separate specifications: 1) we first split the ef-fects based on whether a firm operates in an industry featuring above-median external financedependence (HighEFD) while restricting the sample to firms established after the 1997 Asianfinancial crisis; 2) we also split the effects depending on whether a firm or a CEO was op-erating in a below-median EFD industry during the crisis (LowEFDcrisis) while using thesample of all firms. t-statistics (in parentheses) are calculated using standard errors that areheteroskedasticity-robust and clustered by firm or chaebol -group. ∗∗∗, ∗∗ , and ∗ indicate statis-tical significance at the 1%, 5%, and 10% levels, respectively.
Panel A: Treatment effects and crisis mentality
(1) (2) (3) (4)
Dependent variable ∆Cash/assets Payout Wage increase Investment
Sample Firms established post-crisis
Treated × After -0.713 0.256** 0.180*** 0.302(-1.64) (2.48) (3.49) (0.64)
HighEFD × After 3.663 0.682 -0.043 -8.673(1.33) (1.06) (-1.13) (-1.53)
Treated × After × HighEFD 0.487 0.043 0.028 0.841(0.87) (0.28) (0.34) (0.89)
N 37,753 38,065 35,984 35,242R2 0.2796 0.6407 0.4839 0.4551
Sample All firms
Treated × After -0.323 0.232*** 0.166*** 0.671***(-1.58) (4.60) (5.57) (2.81)
LowEFDcrisis × After 0.291 -0.005 0.208*** 0.779***(1.42) (-0.11) (5.59) (3.26)
Treated × After × LowEFDcrisis 0.072 -0.077 -0.060 -0.209(0.22) (-0.87) (-1.15) (-0.55)
N 70,082 70,431 68,137 66,928R2 0.2452 0.6457 0.4366 0.4153
Firm FE Yes Yes Yes YesIndustry × Year FE Yes Yes Yes Yes
62
Panel B: Value implications and crisis mentality
(1) (2)
Dependent variable CARproposal+CARpassage (-1 to +1 days)
SampleFirms established post-crisis
Chaebol + Non-chaebol Non-chaebol only
Treated 0.022* 0.022*(1.96) (1.81)
HighEFD -0.006 -0.009(-0.23) (-0.34)
Treated × HighEFD -0.019 -0.017(-1.05) (-0.88)
N 391 357R2 0.3417 0.3534
SampleAll firms
Chaebol + Non-chaebol Non-chaebol only
Treated 0.024*** 0.021***(3.85) (3.29)
LowEFDcrisis 0.005 0.005(0.55) (0.49)
Treated × LowEFDcrisis -0.013 -0.014(-1.30) (-1.42)
N 1,451 1,236R2 0.2329 0.2480
Industry FE Yes Yes
63
Table A8Crisis Mentality, Governance, and Pre-reform Cash Levels
In this table we report the results of cross-sectional regressions of the pre-reform level ofcash on (CrisisMemory) (Panel A), and governance index (Panel B), controlling for firmsize and earnings. t-statistics (in parentheses) are calculated using standard errors that areheteroskedasticity-robust and clustered by firm or chaebol -group. ∗∗∗, ∗∗ , and ∗ indicate statis-tical significance at the 1%, 5%, and 10% levels, respectively.
Panel A: Crisis mentality and pre-reform cash
Dependent variable Pre-cash/assets
CrisisMemory 0.781***(4.92)
Size -0.023(-0.42)
Earnings 0.254***(26.40)
Constant 6.426***(4.74)
N 18,135R2 0.0687
Panel B: Governance and pre-reform cash
Dependent variable Pre-cash/assets
High-G -1.188**(-2.39)
Size -0.634***(-2.73)
Earnings 0.180***(3.96)
Constant 23.483***(3.62)
N 637R2 0.0425
64
Table A9Variable Definitions
Treated An indicator variable that equals one if a firm belongsto the treatment group, i.e. either shareholders equityis greater than or equal to 50 billion Korean wons or thefirm belongs to a chaebol group.
After An indicator variable that equals one for year = 2015 or2016.
1{Equity ≥ 50 billion} A indicator variable equal to one if a firm’s shareholders’equity is greater than or equal to 50 billion Korean wons.
Chaebol An indicator variable that equals one if a firm belongsto a chaebol group.
∆Cash/assets [Cash(t) - cash(t-1)]/total assets(t-1)×100.
Payout (Cash dividend(t) + max(purchase of stock(t), treasurystock(t) - treasury stock(t-1)))/total assets(t-1)×100.
Dividend Cash dividend(t)/total assets(t-1)×100.
Repurchase max(purchase of stock(t), treasury stock(t) - treasurystock(t-1))/total assets(t-1)×100.
Investment [Tangible asset(t) - tangible asset(t-1) + deprecia-tion(t)]/total assets(t-1)×100.
Investment in land [Land(t) - land(t-1)]/total assets(t-1)×100.
Investment in building [Building(t) - building(t-1)]/total assets(t-1)×100.
Investment in equipment [Equipment(t) - equipment(t-1)]/total assets(t-1)×100.
Wage increase [Wage(t) - wage(t-1)]/total assets(t-1)×100.
Equity - 50 billion wons Shareholders’ equity - 50 billion Korean wons.
Earnings Net income/total assets×100.
Size Ln(total assets).
Debt Total debts/total assets ×100.
Cash/assets Cash/total assets ×100.
65
G-index Corporate governance score provided by the Korea Cor-porate Governance Service (KCGS). The KCGS scores(using integers) each of four aspects: 1) shareholderrights protections, ranging between 0 and 81; 2) boardoperations, ranging between 0 and 69; 3) monitoring or-ganization operations, ranging between 0 and 43; and 4)transparency in disclosures, ranging between 0 and 47.The four scores are then added together to obtain thetotal governance score.
ROIC EBITDA/Total invested capital = (sales - COGS -SG&A + depreciation + amortization)/(total debts +shareholder’s equity - cash)×100.
Profit margin (Sales - COGS)/sales×100.
PD in 2yrs Measure of the probability of default within two years,developed by the Risk Management Institute of the Na-tional University of Singapore (NUS RMI).
PD in 3yrs Measure of the probability of default within three years,developed by the Risk Management Institute of the Na-tional University of Singapore (NUS RMI).
CARproposal + CARpassage The sum of cumulative abnormal returns (-1 to +1 days)around the proposal of the law by the Korean govern-ment on August 6, 2014 and cumulative abnormal re-turns around the passage of the law by the NationalAssembly on December 2, 2014.
66