Labor-market volatility and financial development in the advanced OECD countries: does labor-market regulation matter?
Darcillon, Thibault
INTRODUCTION
Since the 1980s economic insecurity has grown, especially for
low-skilled workers, with higher unemployment rates, less stable
employment and more volatile wages (OECD, 2007). At the same time, most
advanced OECD economies also have experienced a sharp increase in their
financial activities over the last three decades: financial markets,
financial institutions and financial innovations have been rapidly grown
across countries since the early 1980s. A large literature has explored
the potential impact of the international economic integration on the
increased uncertainty faced by workers (Rodrik, 1997; Buch and
Schlotter, 2011). For instance, Rodrik (1997) shows that the
international economic integration has contributed to the increase in
the elasticity in labor demand, especially for low-skilled workers,
making wages and the number of hours worked more volatile. The aim of
this paper is to analyze the relationship between financial development
and labor-market volatility in the OECD countries. Some recent
contributions have investigated the role of increasing financialization
in higher labor-market volatility. Pagano and Pica (2012) have analyzed
the impact of financial development on labor reallocation across
industries. More specifically, Buch and Pierdzioch (2014) have
investigated on the role of financial globalization in labor-market
volatility. This paper focuses on the relationship between domestic
financial development (and its impact of corporate strategies) and
labor-market volatility.
To analyze the impact of financial development on labor-market
volatility, I argue that higher financial development should affect
corporate strategies and be positively correlated to labor-market
volatility. Thesmar and Thoenig (2004) argue that financial development
has directly affected corporate strategies. For instance,
shareholder-oriented firms are likely to become more sensitive to the
financial market fluctuations. Increasing financial development has
reinforced investors' bargaining power at the global and domestic
levels, pushing for pro-minority shareholder corporate governance
reforms in most OECD countries (Darcillon, 2015). As a consequence,
recent changes in corporate governance thus have contributed to the
shift of risk from shareholders to wage earners. This will then affect
how firms will determine wages and the number of hours worked. However,
there are large cross-country differences in labor-market volatility,
whereas all OECD countries have experienced a large increase in their
financial activities. As the result, the increase in labor-market
volatility has been more modest in countries that have maintained strong
labor-market regulation. In this regard, it has been well recognized in
the literature that specific labor-market regulations (such unemployment
benefits or job-protection laws) are designed to reduce temporary
fluctuations of income (letter et al, 2013; Bertola and Lo Prete, 2015).
Individuals' preferences on stable wages and employment are closely
related to risk aversion. I will test the argument that financial
development should be positively correlated to higher labor-market
volatility in countries with weaker labor-market regulation. In this
paper, labor-market regulation refers to two dimensions: first, the
level of government in the labor market; second, the degree of
generosity of social and welfare policies.
Using panel data on 15 OECD countries from 1974 to 2007, I first
run fixed-effects regressions to analyze the effect of higher financial
development on labor-market volatility. Then, I estimate a threshold
regression model in panel data using a reduced sample of countries from
1986 to 2005, proposed by Hansen (1999), to test whether this effect
depends on labor-market regulation. I measure labor-market instability
using indices of volatility. 1 calculate how much wages and the number
of hours worked change over a 5-year window. The greater the variance of
short-term wages and employment changes, the more volatility there is. I
use two measures of financial development: the stock market
capitalization ratio (as a percentage of GDP) and the employment share
in the financial sector. I find evidence that higher employment in
finance is positively correlated to increased labor-market volatility,
whereas stock market development has no significant impact. Results do
not, however, show any significant differences across skill levels.
Then, results from threshold regressions indicate that this positive
effect is larger particularly in countries with weaker labor-market
regulation (which includes less generous social and welfare policies).
This second result is particularly more robust when explaining
employment volatility of low-skilled workers and when explaining wage
volatility for all categories of workers.
The next section of the paper presents my conceptual framework. In
the subsequent section, I present data on recent trends in labor-market
volatility and detail my measures of financial development and the
indicators for labor-market regulation. Then, I explain my empirical
strategy in the section after that. Estimation results and some
robustness checks are reported in the penultimate section. The final
section provides some concluding remarks and policy implications.
CONCEPTUAL FRAMEWORK
Financial development and labor-market volatility
The aim of my paper is to analyze the relationship between
financial development and labor-market volatility. To do so, I rely on
some recent theoretical and empirical contributions examining how an
increasing influence of the financial markets is likely to be associated
with higher labor-market volatility. I focus on two different
theoretical mechanisms to which financial development may be connected
to higher labor-market volatility: (1) the capital ownership structure
will impact the allocation of risk between workers and shareholders and
(2) the use of pay incentive schemes induced by financial development
should be related to higher demand volatility.
First, in line with Jacoby (2008), I focus on corporate governance
to explain how financial development is correlated to higher
labor-market volatility. In this regard, the capital ownership structure
will impact the allocation of risk between workers, which is strongly
related to labor-market volatility. Corporate governance refers to the
'private rules by which owners, managers, and workers influence a
firm's strategic decisions, including the distribution of rents and
risk' (Jacoby, 2008, p. 8). Jacoby (2008) argues that the relation
between financial and labor markets is shaped by regulation and
corporate governance. I argue that any change in corporate governance is
likely to be related to higher labor-market volatility. An increasing
influence of the financial markets should affect the corporate strategy
by altering the allocation of risks between owners and corporate
stakeholders (including creditors, suppliers or employees) and then
should affect how firms will determine how wages and the number of hours
worked are set. As underlined in a large literature (Roe, 2003; Pagano
and Volpin, 2005), labor-market regulation is strongly related to the
structure of capital ownership. The allocation of risk between owners
and stakeholders is largely determined by the capital ownership
structure, and this because blockholders (ie, large shareholders with at
least 5% of shares) and minority shareholders may have opposed risk
preferences. In the post-war era, corporations in most OECD countries
were owned by blockholders. During this era, workers accepted to cede
authority to managers in exchange for stable jobs and pay increase
(Gourevitch and Shinn, 2005; Darcillon, 2015). Roe (2003) shows that
'blockholding' is associated with strong regulation in the
economy that creates higher incentives to workers to invest in specific
human capital. Then, during the 1970s and the 1980s, financial markets
were liberalized and capital controls were abolished in most OECD
countries: all the advanced OECD countries have since then adopted
reforms intended to strengthen the power of minority shareholders within
firms (Darcillon, 2015). Accordingly, financial development and recent
changes in corporate governance have contributed to the shift of risk
from shareholders to wage earners. Prominority shareholder reforms
increase short-term corporate profitability in the sense that a more
shareholder-oriented strategy focuses on narrow financial objectives
such as movements in stock prices and short-termism (Sjoberg, 2009).
Second, pro-minority shareholder reforms also reduce the role of stable
financing through 'patient capital' by an extended firms'
use of external credit, making them more vulnerable to credit market
volatility. The emergence of a large number of financial innovations
(such as securitization) in the 1980s/1990s played a central role in the
change in the traditional role of banks as 'patient capital'
providers. (1) As a consequence, 'as investors press for larger
returns, employees faced greater risks (...) wage and employment
volatility have risen considerably' (Jacoby, 2008, p. 18).
Second, I argue that the use of pay incentive schemes induced by
financial development should be related to higher demand volatility. In
this line, Thesmar and Thoenig (2004) demonstrate that financial
development, by broadening the pool of external investors both at the
domestic and international levels, improves risk-sharing but also
encourages firms to adopt more profitable and riskier strategies, and
this including for non-listed firms. In this case, firms have higher
incentives to introduce specific income schemes such as
performance-related pay, individual or collective bonuses for employees
(eg, employees stock ownership plan) that are indexed on the firms'
profits, making incomes more volatile (OECD, 2011). In addition, Thesmar
and Thoenig (2004) show that financial development has a general
equilibrium effect on wages and price and should impact the strategies
of non-listed firms. This will result in an overall increase in the
uncertainty of sales, employment and profits in all firms. For this
reason, more developed financial markets should affect product demand
and therefore labor demand. Then, it can be argued that financial
development is more particularly related to higher elasticity of labor
demand for low-skilled workers. Low-skilled workers should have a higher
degree of risk aversion, making them probably more resistant against
greater volatility in financial and labor markets. As shown in a recent
paper by Pagano and Pica (2012), financial development favors labor
reallocation across industries from the 'weaker' to the
'stronger' industries. As the result, workers can expect
higher future labor income, but at the same time income risk can
increase because of labor reallocation. Similarly, a large literature
has shown that low-skilled workers should be more particularly exposed
to higher volatilities on wages and hours worked because labor demand
for low-skilled workers is more elastic (Rodrik, 1997; Scheve and
Slaughter, 2004). To this regard, Buch and Pierdzioch (2014) have
investigated the effect of financial globalization on labor-market
volatility across skill levels. They show that financial openness
particularly increases volatility of hours worked for low-skilled
workers: their empirical results indicate that financial globalization
is associated with a rise in output volatility that may trigger more
additional hires and fires of low-skilled workers because they can less
easily absorb productivity shocks compared with high-skilled workers.
Then, consumption volatility is also lower for this category of workers
because they have easier access to financial markets to smooth their
consumption.
Hypothesis 1: Financial development is positively correlated to
labor-market volatility, especially for low-skilled workers.
Does labor-market regulation matter?
All OECD countries have experienced a large increase in their
financial activities since the last decades. However, the increase in
labor-market volatility has been much more modest in some countries. To
explain these cross-country differences in labor-market volatility, I
argue that labor-market regulation may influence the direct relationship
between financial development and labor-market volatility. It has been
well recognized in the literature that specific labor-market regulations
(such as unemployment benefits or job-protection laws) are designed to
reduce temporary fluctuations of income (Jetter et al, 2013; Bertola and
Lo Prete, 2015). In this context, all these institutions can dampen
labor-market volatility, and this more especially for low-skilled
workers. More specifically, some labor-market institutions, such as
unemployment benefits, can also provide insurance and protect the
citizens against major economic risks (unemployment, poverty, sickness,
family and so on) and then be negatively correlated to labor-market
volatility.
A large literature in labor economics has analyzed the effect of
labor-market regulation on labor-market volatility. Some theoretical and
empirical contributions have shown that stronger labor-market
institutions have a reducing effect on labor-market fluctuations. First,
it has been shown that strong unions' bargaining power and
employment protection appear as a powerful driver for specific skill
investment (Wasmer, 2006) and can dampen labor-market fluctuations as
powerful 'automatic stabilizers'. (2) The automatic
stabilizers can better absorb the shocks associated with higher economic
volatility and instability (Stiglitz, 2000). I use the concept of
automatic stabilizers to analyze the reducing effect of social and
welfare policies on labor-market volatility. Investments in
match-specific human capital reduce the outside option for workers,
implying less incentives to separate. Cairo and Cajner (2014) show that
low-skilled workers have experienced higher employment volatility
because of higher volatile separation rates. In this vein, the OECD
(2011) notes that temporary workers are also much more likely to
experience earnings volatility, both within full-time jobs and because
of movements into and out of work. Similarly, Sala et al (2012) show
that turnover rates are higher for temporary workers, implying higher
employment and wage volatility. Moreover, the lack of the on-the-job
training is also a central determinant in explaining labor-market
volatility for temporary workers. Second, social and welfare policies
have more specifically a strong negative direct effect on labor-market
fluctuations. As noted in a report from the OECD (2011), 'tax and
welfare system scan help buffer households against volatile earnings.
Taxes play a prominent role in reducing the impact of earnings
fluctuations among full-time workers, while transfers such as
unemployment benefits and social assistance are more important when
volatility is due to movements into or out of work'. Thus, rising
volatility may be expected to reduce welfare for risk-averse workers
with limited insurance. Welfare and social policies can in this case
provide insurance and protect the citizens against major economic risks.
In this regard, Rodrik (1997) claims that workers will support social
policies that provide social insurance against external risk generated
by increased trade openness. Accordingly, trade openness should be
positively correlated to public spending in the OECD countries. (3)
The central point in this paper is to focus on the interactions
between financial structures and labor-market institutions (such as
employment protection and redistribution). I focus on two main
theoretical mechanisms: (1) the use of pay incentive schemes and (2) the
incentive for workers to invest in specific human capital, which both
are related to labor-market volatility. First, I argue that financial
markets and labor-market regulation have a joint impact on the corporate
strategies related to firms' pay strategies that will affect labor-
market volatility. According to Thesmar and Thoenig (2004), the degree
of financial development and labor-market regulation may jointly
influence the firms' adoption of a riskier corporate strategy. They
find that strong labor-market regulation increases the costs of choosing
a risky strategy by adopting some variable pay schemes. For instance,
powerful trade unions (which play a central role in pay compression) and
intermediate or sectoral wage bargaining (which reduces the possibility
of the adoption of variable pay schemes depending on individual company
performance) seem to lower the benefits of the use of such pay schemes.
In other words, firms have higher incentives to use pay incentive
schemes and to adopt a riskier strategy, especially when labor markets
are more flexible. Accordingly, regulated financial markets combined
with more regulated labor markets (which include more generous social
and welfare policies) are more likely to reduce the incentives for firms
to adopt a riskier corporate strategy (through variable pay schemes)
generating higher employment and wage volatility. By contrast, less
flexible labor markets imply riskier corporate strategies, which
increases volatility when financial development is higher. As a
consequence, in the economies with higher (resp. lower) developed
financial markets and less (resp. more) regulated labor markets, firms
have higher (resp. lower) incentives to adopt riskier corporate
strategies that will make workers' wages and employment more
volatile. Labor-market regulation can then efficiently mitigate income
and employment risk as powerful automatic stabilizers. Second,
institutional interactions between the financial structures and
labor-market regulation can also jointly affect the incentive for
workers to invest in specific human capital, which can also be related
to lower labor-market volatility. For instance, a low degree of
financial development (ie, strong capital ownership concentration
through insider monitoring) and strong labor market regulation are more
likely to promote the development of internal labor markets, which
increases the incentives for workers to invest in specific human capital
and thus implies lower labor reallocation (Wasmer, 2006), and then to
produce a lower labor-market volatility.
Hypothesis 2: The positive correlation between financial
development and labor market volatility is stronger in countries with
weaker labor market regulation.
DATA AND TRENDS
Measuring labor-market volatility
To obtain a measure of labor-market volatility, I use the EU-KLEMS
Database from the OECD that provides data on total hours worked by
persons engaged and total labor compensation by skill groups (HS for
high-skilled workers and LS for low-skilled workers) from 1974 (for some
countries) to 2007. The EU-KLEMS provides data on labor compensation and
the number of hours worked by skill levels only in the March 2008
database. Unfortunately, the updated databases do not provide any
information on shares in total hours and in total labor compensation. In
addition, data on the share of total hours and total labor compensation
by skill levels are missing for some countries in many years. Table 1
describes data coverage of my measure of labor-market volatility by
country. For most countries, this measure is only available from the
early 1980s.
I build two different measures of labor-market volatility by
following several steps conducting the same methodology used by Buch and
Pierdzioch (2014):
1. First, I calculate the number of hours worked by persons engaged
across skill groups ([h.sub.fit]) by computing the total hours worked by
persons engaged ([H_EMP.sub.it]) in country i for the year t weighted by
the share of hours worked, respectively, by high-skilled and low-skilled
persons engaged in total hours ([H.sub.fit]/100) with f = {HS;LS}).
Then, I calculate average hourly wages for each skill level
([w.sub.fit]) by computing labor compensation converted into constant US
dollar ([LAB.sub.it]) divided by the number of hours worked
([H_EMP.sub.it]) weighted by the share of high-skilled (respectively,
low-skilled) labor compensation in total labor compensation
([LAB.sub.fit]/100) with f = {HS;LS}). (4)
2. Then, as it is very usual in the literature (Buch and
Pierdzioch, 2014; letter et al, 2013), I apply a Hodrick-Prescott (HP)
filter that decomposes the series ([x.sub.t]) into a cyclical
([y.sub.t]) and a trend ([[tau].sub.t]) component. As suggested by
Ravn and Uhlig (2002), I use a value of the smoothing parameter A
of 6.25 for annual data. The parameter determines the relative
importance of the trend and the cyclical component.
3. Finally, I calculate the growth rate of the cyclical trend of
hours worked and of average wages. I compute the rolling standard
deviation for growth rates over a 5-year window by skill groups.
Computing volatility over a chosen rolling window is a common
measure of volatility in the empirical literature. However, as
underlined by Broto et al. (2011), this method has some drawbacks: (1)
this method can generate some problems of endogeneity and serial
correlation; (2) there is a loss of observations corresponding to the
number of years used in the rolling window; and (3) the choice of the
rolling window is somewhat arbitrary. As an alternative measure of
volatility, Broto et al. (2011) propose the Generalized Autoregressive
Conditional Heterogeneity (GARCH) model. However, as noted by Broto et
al. (2011), using an unbalanced panel sample can entail serious caveats
because of data scarcity (leading to convergence errors). Besides, the
GARCH model is based on maximum likelihood estimates that can contain
considerable biases for small samples. For these reasons, I decided to
compute volatility over a chosen rolling window.
Table 1 reveals some differences in the level of volatility of
hours worked by skill groups. First, some Anglo-Saxon countries (such as
Ireland, the United Kingdom and the United States) have experienced
higher volatilities of hours worked for high-skilled and for low-skilled
workers. By contrast, some Northern European countries (such as Belgium,
Denmark and Finland) share low levels of volatility of hours worked. In
addition, it is striking that in some countries (ie, Ireland, Italy, the
Netherlands and the United States) low-skilled workers have been
particularly confronted with increasing volatility compared with
high-skilled workers. As regards the volatility of wages, descriptive
statistics indicate some differences in employment volatility: very
surprisingly, one can see that some countries with low employment
volatility (Austria, Belgium, Germany and Japan) have higher levels of
wage volatility. By contrast, Ireland, the United Kingdom and the United
States are characterized by lower volatility of wages. Finally, some
countries (such as France, Italy, Spain, the United Kingdom and the
United States) present higher levels for low-skilled workers than
high-skilled workers. On average, I find, as in Buch and Pierdzioch
(2014), that the volatility of hours worked has been higher for
high-skilled workers than for low-skilled workers. The main objective of
this paper is to explain cross-national differences in labor-market
volatility, and this across different skill levels.
Measures of financial development and labor-market regulation
Measuring financial development
The main objective of this paper is to examine the relationship
between financial development and labor-market volatility in the OECD
countries. My main argument is based on the idea that the development of
financial markets should be related to corporate strategies and finally
to how firms will determine wages and the number of hours worked. This
argument would require to isolate the component of financial development
that is related to a change in corporate governance. Unfortunately, no
such variable is currently available in a time-series cross-sectional
analysis in the OECD countries. Instead, I use two overall measures of
financial development: the stock market capitalization ratio and the
employment share in the financial sector. (5)
I use as a first financial variable the stock market capitalization
ratio as a percentage of GDP. This variable is provided by the Financial
Structure Database from the World Bank (Beck et al, 2010). This first
variable gives me information on the size of stock markets as a share of
GDP. More developed stock markets are supposed to increase the
investors' opportunities for risk diversification. Most Anglo-Saxon
countries (such as Australia, the United Kingdom and the United States)
have higher stock market capitalization ratio. Anglo-Saxon countries
traditionally have well-developed stock markets that have been
deregulated in the late 1970s or in the early 1980s. Stock markets were
deregulated in other countries in the late 1980s and grew more
particularly during the 1990s.
As argued in the section 'Conceptual framework',
financial development is not limited to stock market development but can
include some recent evolutions in the banking structures. To capture the
importance of the financial markets and financial institutions in the
whole economy, I refer to the concept of 'financialization'. I
use one aggregate indicator measuring the extent of financialization in
a time-series cross-sectional analysis: the share of employment in
finance in the total employment. Shares are computed by using the
EU-KLEMS Database from the OECD that provides data on employment across
sectors based on national accounts from 1970 to 2007. The financial
sector here refers to 'financial intermediation' that includes
financial intermediation, except insurance and pension funding,
insurance and pension funding, except compulsory social security and
activities auxiliary to financial intermediation following the NACE
classification.
Measurement of labor-market regulation
According to Checchi and Garcia-Penalosa (2008], labor-market
institutions encompass different aspects of government policies to
employee organizations, such as the employment protection legislation,
the union density and coverage, the degree of centralization or
coordination of wage bargaining. Labor-market institutions also refer to
unemployment benefits and more broadly to redistribution policies. These
two dimensions of labor-market regulation are taken into account in this
paper.
First, I use the trade-union density rate to measure the first
dimension. This variable is calculated by the OECD as the proportion of
union members among workers. Data are available from 1980 to 2007 and
are provided by the OECD. (6) Then, I use alternative measures for the
second dimension that refers to the protection of the citizens against
major economic risks in the areas of unemployment, poverty, sickness,
family, active labor-market programs, housing and so on. In this regard,
I use three different variables: (1) the public social expenditure as a
percentage of GDP; (2) an overall index of welfare generosity; and (3) a
measure of redistribution. First, I use public social expenditure as a
percentage of GDP that covers all public expenditure, including old age,
survivors' incapacity, health, family, active labor-market
programs, unemployment and housing. This variable is provided by the
OECD. Then, I use an overall index of welfare generosity. This aggregate
index provided by the Comparative Welfare Entitlement Database is a
computation of the net replacement rates of unemployment benefits,
sickness benefits and pension insurance, the extent of program coverage
and its duration (Scruggs et al, 2014). (7) Alternatively, I use an
index of redistribution. This index provided by the Standardized World
Income Inequality Database (Solt, 2009) estimates redistribution by the
percentage reduction in gross income inequality, that is, the difference
between the market and net income inequality, divided by market income
inequality, multiplied by 100. Whereas the share of public social
expenditure in GDP measures welfare effort, that is, how much of it is
spent on social policies, the index of welfare generosity reflects a
measure of eligibility and generosity. Similarly, redistribution
measures how social policies are successful in reducing income
inequality and reflects welfare outcome.
EMPIRICAL STRATEGY
Panel data analysis of financial development on labor-market
volatility
First, I run fixed-effect estimations to analyze the direct effect
of higher financial development on employment and wage volatility. 1
will estimate the following relationship to test my first hypothesis
(Hypothesis 1):
[Y.sub.it] = [[beta].sub.1] x [FIN.sub.it] + [[beta].sub.k] x
[X.sub.k,it] + [[lambda].sub.i] + [[eta].sub.t] + [[epsilon].sub.it] (1)
where [Y.sub.it] denotes my measures of labor-market volatility for
each skill group. [FIN.sub.it] is a set of two variables capturing
financial development (stock market development and employment share in
the financial sector), and [X.sub.k,it] is a vector of time-varying
country controls. The specification includes a set of country
fixed-effects ([[lambda].sub.i]) and year fixed-effects ([[eta].sub.t]).
[[epsilon].sub.it] is a disturbance term.
I first test for the pooling restrictions: if parameters of
equation 1 are equal across countries, time-series and cross-sectional
data is more appropriate in this case. I run a Breusch-Pagan test. The
null hypothesis in this test is that variances across entities are zero.
This test implies that equation 1 includes country individual effects
and shows strong evidence of significant differences across countries. I
decided to run fixed-effects estimations as suggested by a Hausman test.
My choice of control variables is in line with the existing
literature (Buch and Pierdzioch, 2014). First, I control for
unemployment rates (Unemployment rate) as a determinant of labor-market
volatility, in particularly for low-skilled workers. Low-skilled workers
are more likely to have a higher risk of becoming unemployed or jobless
and with more persistent periods of unemployment, then affecting their
labor-market volatility. A positive relationship between higher
unemployment rates and my two measures of labor-market volatility is
expected, and this especially for low-skilled workers. Then, I control
for the volatility of total factor productivity (TFP): variations in TFP
could be important in explaining the observed labor-market volatility.
More specifically, it is very usual to consider differences in
employment and wages across skill levels by the differences in
productivity level. Volatility of TFP should increase in output and in
hours worked, especially for high-skilled workers, because of lower
costs adjusting hours worked per labor compensation (Buch and
Pierdzioch, 2014). Accordingly, higher TFP volatility should be
positively correlated to my dependent variables. The EU-KLEMS Database
provides information on TFP growth, value-added based. I compute the
rolling standard deviation for TFP growth over a 5-year window
[Volatility of TFP). Finally, I also control for output volatility that
is expected to impact labor market and then to increase their
volatility. I calculate the volatility of output growth by computing the
rolling standard deviation for growth rates of real GDP over a 5-year
window [Output volatility).
Panel threshold regression analysis of financial development on
labor-market volatility
Then, I want to analyze how financial structures may interact with
labor-market regulation to explain cross-country differences in
labor-market volatility. To capture the interdependences between
financial and labor-market variables, the use of an interaction term is
very common. But this method raises two different problems. First, the
linkages between financial development, labor-market variables and
labor-market volatility are very complex to analyze. Financial
development may depend on household demand for private insurance, which
in turn depends on the availability of social insurance. Therefore, it
is very difficult to disentangle all these variables. Second,
introducing a misspecified interaction term between financial and
labor-market variables may produce biased coefficients because of
potential multicollinearity (Chatelain and Ralf, 2014). To avoid these
two different problems, I use a sample-split (depending on labor-market
institutional features) to account for different skill levels with
respect to Pagano and Pica (2012). Hansen (1999) proposes a sample-split
and threshold regression technique appropriate for panel data analysis.
In other words, I suspect a non-linear relationship between my
financial variables and my measures of labor-market volatility. I argue
in the section 'Conceptual framework' that the effect of
financial development on labor market should be impacted by labor-market
institutions. More particularly, I assume that the increasing effect of
financial development should be undermined by the reducing effect of
labor-market regulation. In line with my second hypothesis (Hypothesis
2), I decide to split my sample depending on specific labor-market
institutional features, expressed as follows from equation 1:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (2)
with [LAB.sub.it] denoting labor-market variables. These variables
are used in my analysis as threshold variables in order to sort the data
in different regimes or groups of countries. The threshold model permits
the regression parameters ([[beta]'.sub.1] and
[[beta]'.sub.2]) to switch between regimes depending on whether
[LAB.sub.it] is smaller or larger than the threshold value [gamma].
Hypothesis 2 implies that [[beta]'.sub.1] > [[beta].sub.2].
Equation 2 can be rewritten as in a single equation form with the
introduction of the indicator function I(*):
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (3)
This method has several advantages. First, this model endogenously
identifies the thresholds for LAB at which the relationship between
labor-market volatility and financial development switches. In other
words, the model searches over all values of [LAB.sub.it] for [gamma]
sequentially until sample- splitting value y is found. The procedure
proposed by Hansen (1999) also allows for testing the statistical
significance of the threshold effect (ie, the null hypothesis of no
threshold [H.sub.0]:[[beta].sub.1] = [[beta].sub.2] against the
alternative hypothesis of having at least one threshold [H.sub.1]:
[[beta].sub.1] [not equal to] [[beta].sub.2]. To do this, Hansen (1999)
recommends a bootstrap procedure to determine critical values of the
test statistics. The critical values for determining the 95% confidence
interval of the threshold values are given by:
[GAMMA] = {[gamma]: LR([gamma]) [less than or equal to] C([alpha])}
(4)
where C([alpha]) is the 95% percentile of the asymptotic
distribution of the like-lihood ratio (LR([gamma])). The null hypothesis
of no threshold effect will be rejected if the bootstrap estimate of the
p-value for the likelihood ratio (LR) test is smaller than the desired
critical values (eg, 5%). Then, once a threshold is found, Hansen (1999)
recommends to test the presence of two thresholds with the help of
another LR test. Second, as earlier mentioned, this method deals with
the multicollinearity problem raised by the use of misspecified
interaction terms (Chatelain and Ralf, 2014).
A potential problem with this approach to identifying the
thresholds is that the welfare state and labor-market variables are
potentially endogenous while the Hansen (1999) method requires that
independent variables are strictly exogenous. Labor-market institutions,
and more particularly social and welfare policies, can be endogenous to
economic conditions: as noted by Rodrik (1998) and Fatas and Mihov
(2001), governments should be capable of stabilizing labor-market
fluctuations if such economies are more volatile. In order to take into
account the effect of labor-market volatility on my different
labor-market variables, and hence to address this issue of potential
reversed causality, I run IV-2SLS regressions (producing
autocorrelation-robust covariance matrix and standard errors) and
implement several endogeneity tests for the potential endogenous
variables. I fail to reject the null and conclude that all my
labor-market variables are actually exogenous.
ESTIMATION RESULTS
Is financial development positively correlated to labor-market
volatility?
At this first stage of the analysis, my sample is composed of 15
OECD countries (Australia, Austria, Belgium, Denmark, Finland, France,
Germany, Ireland, Italy, the Netherlands, Spain, Sweden, the United
Kingdom and the United States) from 1974 to 2007 when using the
employment share in the financial sector as a financial variable and
restricting from 1989 to 2007 when using the stock market capitalization
ratio. Table A1 in Appendix describes all the variables used in my
regression analysis.
Fixed-effects regression results showing the impact of stock market
development and employment share in the financial sector are displayed
in Table 2. I estimate equation 1 separately for each skill group
(high-skilled and low-skilled workers). Columns (1) and (2) show results
when using the volatility of hours worked as dependent variable, whereas
columns (3) and (4) present results when using the volatility of wages
as dependent variable. Table 2A presents the results of the impact of
stock market development on labor-market volatility. In Table 2B, I use
the employment share in the financial sector as an alternative measure
of financial development. (8)
To start with, I find that some of my control variables have the
expected sign. First, I find contrasted evidence of a positive effect of
unemployment on labor-market volatility. Results show that higher
unemployment rates are associated with higher wage volatility with a
larger effect for high-skilled workers. I find that higher unemployment
is also negatively correlated to employment volatility for high-skilled
workers, suggesting that high-skilled workers can more easily absorb
productivity shocks. Second, when statistically significant, I find that
TFP and output volatilities are, as expected, associated with higher
labor-market volatility. Volatility of TFP should increase in output and
in hours worked, especially for high-skilled workers, because of lower
costs adjusting hours worked per labor compensation (Buch and
Pierdzioch, 2014). I find that volatility of TFP exerts a positive
effect on employment volatility but only when considering the employment
share in the financial sector as a financial variable (Table 2B) with a
larger effect for high-skilled workers. Finally, higher output
volatility is more likely to affect low-skilled workers than
high-skilled workers (models (2) and (4) in Table 2A).
When focusing on my two different financial variables, I find that
higher stock market development has no significant impact on
labor-market volatility (Table 2A), whereas higher employment share in
the financial sector is strongly associated with increased labor-market
volatility (Table 2B). (9) Second, when I compare the effect across
skill levels, I find a larger effect on employment volatility for
high-skilled workers and a larger effect on wage volatility for
low-skilled workers. In order to compare the respective effect of
employment share in the financial sector on labor-market volatility
across skill groups, I compute a Wald test of coefficient equality,
which suggests that the difference across skill groups is not
statistically significant. (10) To conclude, Hypothesis 1 is partially
validated: I find evidence that higher employment in finance is
positively correlated to increased labor-market volatility. Results do
not, however, show any significant differences across skill levels.
Do labor-market institutions influence the labor-market volatility
effect of financial development?
In order to test my second hypothesis assuming a non-linear
relationship between financial development and labor-market volatility
depending on labor-market institutional features, I use a fixed-effect
threshold regression model proposed by Hansen (1999). This model,
however, requires a strongly balanced panel, restricting my sample to 13
OECD countries (excluding Ireland and Sweden) from 1986 to 2005 when
using the employment share in finance and from 1989 to 2005 when using
the stock market capitalization ratio.
Fixed-effect threshold regression results are displayed in Tables 3
and 4. The estimated threshold values ([[??].sub.1] and [[??].sub.2])
for each test are also reported. For each labor-market variable, I test
for a single and a double threshold where a number of 300 bootstrap
replications were used for each of two bootstrap LR tests. (11) I only
report the results for the variables where the null hypothesis of no
threshold [H.sub.0]: [[beta].sub.1] = [[beta].sub.2] is rejected (at the
1 % and 5% levels of significance). For most specifications, I estimate
a panel data model with one single threshold. In specification (6) in
Table 4, the LR test suggests that there are two different threshold
values. I estimate in this case a panel data with double threshold.
Fixed-effect threshold regression results showing the impact of
stock market development are displayed in Table 3. Table 4 reports the
results when using the employment share in the financial sector as an
alternative financial variable.
When considering stock market capitalization ratio as a financial
measure in Table 3, I find in most specifications strong evidence that
the positive correlation between labor-market volatility and financial
development is stronger in countries with weaker labor-market
regulation, supporting Hypothesis 2 where [[beta]'.sub.1] >
[[beta]'.sub.2]. In other words, I find a larger effect of
financial development on labor-market volatility in models (3) and (4)
for employment volatility and in models (7) and (8) for wage volatility
for low values of labor-market variables. (12) By contrast, I find in
some specifications a larger effect for higher values of labor-market
regulation (models (1), (2), (5) and (6)), refuting Hypothesis 2. What
this finding could suggest is that labor-market institutions may be
detrimental to employment, especially for high-skilled workers.
According to Buch and Schlotter (2011), employment volatility could then
increase with the level of labor-market regulation by affecting
adjustments of the labor force, especially for high-skilled workers. As
a consequence, higher labor-market regulation could be associated with
an increase in the elasticity of labor demand and could positively
affect the volatility of employment.
Then, I also find some evidence of Hypothesis 2 in most
specifications in Table 4 when using the employment share in the
financial sector as the explanatory variable. In other words, I find
that the positive correlation between financial development and
labor-market volatility is stronger in countries with weaker
labor-market regulation. Surprisingly, Hypothesis 2 is, however, not
validated when using the degree of overall welfare generosity as a
measure of labor-market regulation (models (3) and (6)) where I find a
larger effect of financial development on the dependent variables for
higher values of overall welfare generosity.
To conclude, I find some evidence supporting my two hypotheses: (i)
financial development, when measured by higher employment share in the
financial sector, is strongly associated with increased labor-market
volatility, but with no statistical differences across skill levels and
(ii) this positive effect is larger particularly in countries with
weaker labor-market regulation. Results are particularly robust when
using the employment share in the financial sector as financial
indicator.
Robustness checks
In this subsection, I use an alternative measure of financial
sector depth: the share of domestic credit to private sector by banks in
GDP, which refers to financial resources provided to the private sector
by other depository corporations (deposit taking corporations except
central banks), such as through loans, purchases of non-equity
securities, and trade credits and other accounts receivable, which
establish a claim for repayment. A large literature has emphasized the
role of financial intermediation in instability. For instance, Stiglitz
(2000) argues that financial deregulation and the extension of the
banking activities have not contributed to the reduction in the
asymmetrical information problem and thus have participated to the
increase in the differences in credit availability across households. In
the same line, Greenwood and Jovanovic (1990) argue that the use of
financial intermediation does not hamper poor but favor rich people.
In comparison with Table 2, I find very similar fixed-effect
estimation results: higher banking credit ratio is positively and
significantly associated with higher labor-market volatility only for
low-skilled workers. Then, I find similar and robust results on a
positive correlation between financial development (measured by banking
credit) and labor-market volatility, and this conditional on specific
levels of different dimensions of labor-market regulation, validating
Hypothesis 2 in most specifications.
CONCLUSION AND POLICY IMPLICATIONS
The aim of the paper was to assess how financial development is
positively correlated to higher volatility on average wages and
employment over the period of 1974-2007 in 15 advanced OECD countries.
The originality of this paper is to explain cross-country differences in
labor-market volatility using a threshold regression model in panel data
proposed by Hansen (1999).
First, my analysis on panel data suggests that the increasing
influence of the financial markets is positively correlated to higher
labor-market volatility. Despite higher possibilities of
risk-diversification, higher financial development may be related to a
change in firms' strategies and modify how firms will determine
wages and hours worked. Using different overall measures of financial
development (which are, however, not specifically related to a change in
corporate governance), fixed-effect regression estimations indicate that
higher employment share in the financial sector is significantly
correlated to my two dependent variables. In addition, results show no
significant differences across skill levels, whereas I expected a larger
effect for low-skilled workers than for high-skilled workers. Second,
using a threshold regression model in panel data on a reduced sample, I
find some evidence that this positive correlation between financial
development and labor-market volatility is higher (resp. lower) in
countries with weaker (resp. stronger) labor-market regulation (ie, the
level of government intervention in the labor market and the degree of
generosity of social and welfare policies). Results are more robust when
explaining employment volatility for low-skilled workers and when
explaining wage volatility for all categories of workers.
These interactions between financial development and labor-market
regulation can be useful to explain cross-national variations in
labor-market volatility. As a result, the increase in employment and
wage volatility has been more modest in countries with more regulated
labor markets. This result has strong policy implications at least in
the short run. Whereas a large literature has shown that higher economic
growth implies larger macroeconomic volatility, this paper stresses the
consequences of financial development on labor-market volatility in the
short run. In this regard, less government intervention is associated
with higher economic growth in the long run (eg, Erauskin, 2011).
However, labor-market regulation through automatic stabilizers is more
likely to play an active role in cushioning the effects of the financial
markets on labor-market volatility in the short run. This is indeed
widespread concern about higher economic uncertainty linked to
employment and wage volatility. From this perspective, faced with higher
insecurity, workers could support higher redistribution (Rodrik, 1997).
Acknowledgements
The author wants to thank Bruno Amable, Karim Azizi, Christophe
Rault, Antoine Reberioux, all participants of the 25th Annual Conference
of the Society for the Advancement of Socio-Economics (SASE) at the
University of Milan, 27-29 June 2013, two anonymous referees and the
Editor for valuable comments on a previous version.
REFERENCES
Beck, T, Asli Demirguc-Kunt, A and Levine, R. 2010: A new database
on financial development and structure. World Bank,
http://econ.worldbank.org/WBSITE/EXTERNAL/EXTDEC/EXTRESEARCH/ 0.,
contentMDK:20696167 ~
pagePK:64214825-piPK:64214943~theSitePK:469382,00.html, accessed 25
January 2016.
Bertola, G and Lo Prete, A. 2015: Labor market reforms, finance,
and the current account. Review of International Economics 23(3):
469-488.
Broto, C, Diaz-Cassou, J and Erce, A. 2011: Measuring and
explaining the volatility of capital flows to emerging countries.
Journal of Banking and Finance 35(8): 1941-1953.
Buch, CM and Schlotter, M. 2011: Regional origins of employment
volatility: Evidence from German states. Empirica 40(1): 1-19.
Buch, CM and Pierdzioch, C. 2014: Labor market volatility, skills,
and financial globalization. Macroeconomic Dynamics 18(5): 1018-1047.
Cairo, I and Cajner, T. 2014: Human Capital and Unemployment
Dynamics: Why More Educated Workers Enjoy Greater Employment Stability,
Finance and Economics Discussion Series, Federal Reserve Board, 2014-09.
Chatelain, JB and Ralf, K. 2014: Spurious regressions and
near-multicollinearity, with an application to aid, policies and growth.
Journal of Macroeconomics 39(A): 85-96.
Checchi, D and Garcia-Penalosa, C. 2008: Labour market institutions
and income inequality. Economic Policy 23 (56): 600-651.
Darcillon, T. 2015: Corporate governance reforms and political
partisanship: An Empirical Analysis in 16 OECD Countries. Business and
Politics 17(4): 661-676.
Erauskin, I. 2011: Financial openness, volatility, and the size of
productive government. SERIES 2(2): 233-253.
Fatas, A and Mihov, I. 2001: Government size and automatic
stabilizers: International and intranational evidence. Journal of
International Economics 55(1): 3-28.
Gourevitch, PA and Shinn, J. 2005: Political power and corporate
control: The new global politics of corporate governance. Princeton
University Press: Princeton.
Greenwood, J and Jovanovic, B. 1990: Financial development, growth,
and the distribution of income. Journal of Political Economy 98(5):
1076-1107.
Hansen, B. 1999: Threshold effects in non-dynamic panels:
Estimation, testing, and inference. Journal of Econometrics 93(2):
345-368.
Jacoby, SM. 2008: Finance and labor: Perspectives on risk,
inequality, and democracy. Comparative Labor Law & Policy Journal
30(1): 17-65.
Jetter, M, Nikolsko-Rzhevskyy, A and Smith, WT. 2013: The effects
of wage volatility on growth. Journal of Macroeconomics 37(1): 93-109.
Macit, F. 2010: The role of labor market institutions on wage and
inflation dynamics: Empirical evidence from OECD economies. Economic
Analysis & Policy 40(1): 49-62.
OECD. 2007: OECD workers in the global economy: Increasingly
vulnerable? OECD employment outlook 2007. OECD Publishing: Paris.
OECD. 2011: Earnings volatility: Causes and consequences. OECD
employment outlook 2011. OECD Publishing: Paris.
Pagano, M and Pica, G. 2012: Finance and employment. Economic
Policy 27(69): 5-55.
Pagano, M and Volpin, P. 2005: The political economy of corporate
governance. American Economic Review 95(4): 1005-1030.
Ravn, MO and Uhlig, H. 2002: On adjusting the Hodrick-Prescott
filter for the frequency of observations. The Review of Economics and
Statistics 84(2): 371-380.
Rodrik, D. 1997: Has globalisation gone too far? Institute for
International Economics: Washington DC.
Rodrik, D. 1998: Why do more open economies have bigger
governments? Journal of Political Economy 106(5): 997-1032.
Roe, MJ. 2003: Political determinants of corporate governance.
Political context, corporate impact. Oxford University Press: Oxford.
Sala, H, Silva, JI and Toledo, M. 2012: Flexibility at the margin
and labor market volatility in OECD countries. Scandinavian Journal of
Economics 114(3): 991-1017.
Scheve, K and Slaughter, MJ. 2004: Economic insecurity and the
globalization of production. American Journal of Political Science
48(4): 662-674.
Scruggs, L, Detlef, J and Kati, K. 2014: Comparative welfare
entitlements dataset 2. Version 2014-03. University of Connecticut &
University of Greifswald, http://cwed2.org/, accessed 25 January 2016.
Sjoberg, O. 2009: Corporate governance and earnings inequality in
the OECD countries 1979-2000. European Sociological Review 25(5):
519-533.
Solt, F. 2009: Standardizing the world income inequality database.
Social Science Quarterly 90(2): 231-242.
Stiglitz, JE. 2000: Capital market liberalization, economic growth,
and instability. World Development 28(6): 1075-1086.
Thesmar, D and Thoenig, M. 2004: Financial market developments and
the rise in firm level uncertainty. Centre for Economic Policy Research:
London, http://www.cepr.org/active/publications/
discussion_papers/dp.php?dpno=4761, accessed 25 January 2016.
Wasmer, E. 2006: Interpreting Europe-US labour market differences:
The specificity of human capital investments. American Economic Review
96(3): 811-831.
APPENDIX
Table Al: Variables description
Variables Description Time N
Dependent variables
Employment volatility Rolling standard 1974-2007 437
deviation for growth
rates of the number of
hours worked by persons
engaged over a 5-year
window for high-skilled
and low-skilled workers
(Source: EU-KLEMS
Database)
Wage volatility Rolling standard 1974-2007 437
deviation for growth
rates of the average wage
over a 5-year window for
high-skilled and
low-skilled workers
(Source: EU-KLEMS
Database)
Financial variables
Stock market Stock market 1989-2007 278
development capitalization to GDP
(Source: Financial
Structure Database)
Employment share in Share of employment in 1970-2007 567
the financial sector the financial sector in
the total employment
(Source: EU-KLEMS)
Banking credit Domestic credit to 1970-2007 567
private sector by banks
to GDP (Source: Financial
Structure Database)
Welfare state and labor-market variables
Trade union density Proportion of union 1970-2007 570
members among workers
(Source: OECD)
Social security Public social 1970-2007 570
expenditure expenditures as a
percentage of GDP (old
age, survivors'
incapacity, health,
family, active
labor-market programs,
unemployment and housing)
from 1970 to 2007
(Source: OECD Social
Expenditure Database
(SOCX))
Overall welfare Composite index 1971-2007 525
generosity index calculated by the sum of
subindexes of
unemployment, sickness
and pension generosity
(Source: Comparative
Welfare Entitlement
Database)
Redistribution Difference between the
pre-tax and post-tax
income inequality,
divided by pre-tax income
inequality, multiplied by
100) as an estimate of
redistribution (Source:
Solt, 2009)
Control variables
Unemployment rate Standardized unemployment 1970-2007 570
rates (Source: OECD Main
Economic Indicators
Database)
TFP volatility Rolling standard 1970-2007 418
deviation for TFP growth
rates (value-added based)
over a 5-year window
(Source: EU-KLEMS)
Output volatility Rolling standard 1974-2007 510
deviation for real GDP
growth rates over a 5-
year window (Source: OECD
Economic Outlook
Database)
(1) For instance, the traditional compromise between blockholders,
managers and trade unions in Germany was undermined when large German
banks (such as Deutsche Bank, Dresdner Bank) in the mid-1980s ceased to
play their traditional role of 'patient capital' providers to
benefit from the internationalization of the Anglo-Saxon banks.
Gradually, banks sold the shares they held in firms, withdrew from their
management and supervisory boards and became investment banks. Obviously
these evolutions directly affected the behavior of large German firms,
becoming more sensitive from the pressures to generate 'shareholder
value'.
(2) Automatic stabilizers are specific features of government
spending to dampen the fluctuations in real GDP. Fatas and Mihov (2001)
show that government expenditure through automatic stabilizers has a
reducing effect on output volatility.
(3) Rodrik (1997) finds no empirical evidence of this argument when
country-specific characteristics are controlled for fixed-effects.
(4) Labor compensation converted into constant US dollar is
obtained by using the exchange rate series from the Penn World Tables
and by deflating by the US output price index.
(5) I use in robustness checks the share of domestic credit to
private sector by banks in the GDP as another common measure of
financial sector depth.
(6) Other common measures of labor market regulation, such as the
strictness of employment protection legislation and the degree of
coordination in wage bargaining, have been used in the analysis. But,
the threshold regression method used in the empirical analysis in the
section 'Estimation results' indicates that these two
different variables cannot be used to sample-split. I fail to reject the
null hypothesis of no threshold [H.sub.0]: [[beta].sub.1] =
[[beta].sub.2].
(7) Unemployment and sickness generosity indexes are calculated on
replacement rate, qualification period, duration, waiting days and
coverage. Pension generosity index is calculated on replacement rate,
expected pension duration years, pension qualification years and
employee pension funding ratio.
(8) F-tests have been run to test the presence of year
fixed-effects. The null hypothesis assumes that all year coefficients
are equal to zero. If I fail to reject the null, no time fixed-effects
are needed only when considering employment volatility as dependent
variable.
(9) I find very similar and robust results when calculating my
measures of labor market volatility with different values of 1 (such as
1 = 100). Using alternative filters such as the Baxter-King filter,
gives substantially similar results.
(10) Results of this test are not here reported.
(11) Using more bootstrap replications gives very similar results.
(12) Models (7) and (8) indicae that higher stock market
development is associated with a reduction (and not an increase as
expected) in labor-market volatility for higher values of welfare
generosity. Some recent contributions have shown that labor market
institutions have a reducing effect on the volatility of real wage
growth (Macit, 2010; Buch and Schlotter, 2011). Strong labor market
regulation allows to maintain workers' wages and lead to smoother
responses in real wages. Accordingly, the increasing effect of financial
development on wage volatility is undermined by the reducing effect of
welfare state and labor market variables, including for high-skilled
workers.
THIBAULT DARCILLON
Maison des Sciences Economiques, University of Paris I
Pantheon-Sorbonne, 106-112 Boulevard de l'Hopital, Paris 75013,
France.
E-mail: thibault.darcillon@univ-parisl.fr
Table 1: Country-level descriptive statistics
Employment
volatility Wage volatility
High- Low- High- Low- Data
skilled skilled skilled skilled coverage
Australia 0.0343 0.0289 0.0749 0.0644 1983-2007
Austria 0.0333 0.0184 0.1046 0.0906 1981-2007
Belgium 0.0158 0.0131 0.0951 0.0919 1981-2007
Denmark 0.0209 0.0251 0.0945 0.0894 1981-2007
Finland 0.0254 0.0284 0.0904 0.0835 1974-2007
France 0.0357 0.0185 0.0843 0.0933 1981-2007
Germany 0.0305 0.0262 0.0949 0.0933 1974-2007
Ireland 0.0526 0.0530 0.0863 0.0795 1989-2007
Italy 0.0185 0.0291 0.0977 0.1025 1974-2007
Japan 0.0225 0.0237 0.0984 0.0958 1974-2007
The Netherlands 0.0434 0.0568 0.1240 0.1004 1980-2007
Spain 0.0402 0.0276 0.0710 0.0794 1981-2007
Sweden 0.0434 0.0332 0.0883 0.0798 1982-2007
The United Kingdom 0.0473 0.0425 0.0768 0.0824 1974-2007
The United States 0.0360 0.0459 0.0233 0.0445 1974-2007
Time (by decade)
1970-1979 0.0373 0.0405 0.0658 0.0623
1980-1989 0.0313 0.0248 0.0868 0.0870
1990-1999 0.0365 0.0335 0.0975 0.0946
2000-2007 0.0280 0.0323 0.0786 0.0771
Total 0.0327 0.0311 0.0865 0.0848
Table 2: Fixed-effects estimations
Employment volatility
HS LS
(1) (2)
A. Financial variable: Stock market capitalization/GDP
Stock market capitalization/GDP 0.0012 0.0087
(0.0040) (0.0056)
Unemployment rate -0.0011 ** -0.0014 *
(0.0005) (0.0007)
TFP volatility 0.0016 -0.0025
(0.0015) (0.0021)
Output volatility 0.0037 * 0.0086 ***
(0.0020) (0.0027)
Constant 0.0235 *** 0.0196 **
(0.0062) (0.0087)
Observations 237 237
Number of country 15 15
[R.sup.2] 0.0355 0.0535
Year fixed-effect No No
B. Financial variable: Employment share in the financial sector
Employment share in finance 0.0169 *** 0.0152 ***
(0.0032) (0.0045)
Unemployment rate -0.0011 *** -0.0006
(0.0004) (0.0005)
TFP volatility 0.0031 *** 0.0004
(0.0011) (0.0015)
Output volatility 0.0018 0.0016
(0.0011) (0.0015)
Constant -0.0299 *** -0.0244 *
(0.0103) (0.0142)
Observations 358 358
Number of country 15 15
[R.sup.2] 0.0902 0.0336
Year fixed-effect No No
Wage volatility
HS LS
(3) (4)
A. Financial variable: Stock market capitalization/GDP
Stock market capitalization/GDP -0.0062 -0.0089
(0.0082) (0.0095)
Unemployment rate 0.0035 *** 0.0020 **
(0.0009) (0.0010)
TFP volatility 0.0045 * 0.0061 **
(0.0024) (0.0028)
Output volatility 0.0016 0.0076 **
(0.0032) (0.0038)
Constant 0.1225 *** 0.1225 ***
(0.0109) (0.0128)
Observations 237 237
Number of country 15 15
[R.sup.2] 0.5432 0.4438
Year fixed-effect Yes Yes
B. Financial variable: Employment share in the financial sector
Employment share in finance 0.0159 * 0.0202 **
(0.0082) (0.0086)
Unemployment rate 0.0030 *** 0.0022 **
(0.0009) (0.0009)
TFP volatility 0.0096 *** 0.0083 ***
(0.0028) (0.0029)
Output volatility -0.0035 -0.0014
(0.0027) (0.0028)
Constant 0.0039 -0.0016
(0.0262) (0.0272)
Observations 358 358
Number of country 15 15
[R.sup.2] 0.0887 0.0668
Year fixed-effect Yes Yes
Table 3: Fixed-effects threshold estimations
Employment volatility
HS HS LS
(1) (2) (3)
Single threshold
Stock market capitalization 0.0356 ***
(trade union [less (0.0060)
than or equal to]
[[??].sub.1])
Stock market capitalization -0.0038
(trade union > (0.0054)
[[??].sub.1])
Stock market capitalization
(social security
expenditure [less than or
equal to] [[??].sub.1])
Stock market capitalization
(social security
expenditure >
[[??].sub.1])
Stock market capitalization -0.0071 *
(welfare generosity [less (0.0038)
than or equal to]
[[??].sub.1])
Stock market capitalization 0.0301 ***
(welfare generosity > (0.0050)
[[??].sub.1])
Stock market capitalization -0.0202 ***
(redistribution [less (0.0051)
than or equal to]
[[??].sub.1])
Stock market capitalization 0.0104 ***
(redistribution (0.0039)
[[??].sub.1])
Double threshold
Stock market capitalization
(redistribution [less
than or equal to]
[[??].sub.1])
Stock market capitalization
([[??].sub.1] <
redistribution [less
than or equal to]
[[??].sub.2])
Stock market capitalization
([[??].sub.2] <
redistribution)
[[??].sub.1] 36.90 34.64 23.75
[[??].sub.2]
Controls Yes Yes Yes
Observations 221 221 221
[R.sup.2] 0.2535 0.2021 0.2697
Number of country 13 13 13
Employment volatility
LS LS LS
(4) (5) (6)
Single threshold
Stock market capitalization
(trade union [less
than or equal to]
[[??].sub.1])
Stock market capitalization
(trade union >
[[??].sub.1])
Stock market capitalization 0.0433 ***
(social security (0.0061)
expenditure [less than or
equal to] [[??].sub.1])
Stock market capitalization -0.0003
(social security (0.0051)
expenditure >
[[??].sub.1])
Stock market capitalization -0.0019
(welfare generosity [less (0.0048)
than or equal to]
[[??].sub.1])
Stock market capitalization 0.0641 ***
(welfare generosity > (0.0069)
[[??].sub.1])
Stock market capitalization
(redistribution [less
than or equal to]
[[??].sub.1])
Stock market capitalization
(redistribution
[[??].sub.1])
Double threshold
Stock market capitalization -0.0082
(redistribution [less (0.0063)
than or equal to]
[[??].sub.1])
Stock market capitalization 0.0624 ***
([[??].sub.1] < (0.0074)
redistribution [less
than or equal to]
[[??].sub.2])
Stock market capitalization 0.0081
([[??].sub.2] < (0.0056)
redistribution)
[[??].sub.1] 11.69 37.00 34.98
[[??].sub.2] 39.50
Controls Yes Yes Yes
Observations 221 221 221
[R.sup.2] 0.3218 0.3842 0.3914
Number of country 13 13 13
Wage volatility
HS LS
(7) (8)
Single threshold
Stock market capitalization
(trade union [less
than or equal to]
[[??].sub.1])
Stock market capitalization
(trade union >
[[??].sub.1])
Stock market capitalization
(social security
expenditure [less than or
equal to] [[??].sub.1])
Stock market capitalization
(social security
expenditure >
[[??].sub.1])
Stock market capitalization 0.0206 0.0288 **
(welfare generosity [less (0.0130) (0.0144)
than or equal to]
[[??].sub.1])
Stock market capitalization -0.0540 *** -0.0510 ***
(welfare generosity > (0.0077) (0.0086)
[[??].sub.1])
Stock market capitalization
(redistribution [less
than or equal to]
[[??].sub.1])
Stock market capitalization
(redistribution
[[??].sub.1])
Double threshold
Stock market capitalization
(redistribution [less
than or equal to]
[[??].sub.1])
Stock market capitalization
([[??].sub.1] <
redistribution [less
than or equal to]
[[??].sub.2])
Stock market capitalization
([[??].sub.2] <
redistribution)
[[??].sub.1] 24.00 24.00
[[??].sub.2]
Controls Yes Yes
Observations 221 221
[R.sup.2] 0.3130 0.2478
Number of country 13 13
* P<0.1, ** P<0.05, *** P<0.01.
Note: Standard errors in parentheses. HS = high-skilled workers
and LS = low-skilled workers. Financial variable: Stock market
capitalization/GDP.
Table 4: Fixed-effects threshold estimations
Employment volatility
HS HS HS
(1) (2) (3)
Single threshold
Employment in finance 0.0376 ***
(trade union [less (0.0047)
than or equal to]
[[??].sub.1])
Employment in finance 0.0296 ***
(trade union > (0.0044)
[[??].sub.1])
Employment in finance 0.0304 ***
(social security (0.0046)
expenditure [less
than or equal to]
[[??].sub.1])
Employment in finance 0.0222 ***
(social security (0.0046)
expenditure >
[[??].sub.1])
Employment in finance 0.0218 ***
(welfare generosity (0.0046)
[less than or equal
to] [[??].sub.1])
Employment in finance 0.0287 ***
(welfare generosity (0.0046)
> [[??].sub.1])
Employment in finance
(redistribution [less
than or equal to]
[[??].sub.1])
Employment in finance
(redistribution >
[[??].sub.1])
[[??].sub.1] 23.75 12.31 37.00
Controls Yes Yes Yes
Observations 260 260 260
[R.sup.2] 0.2869 0.2333 0.2254
Number of country 13 13 13
Employment volatility
LS LS LS
(4) (5) (6)
Single threshold
Employment in finance 0.0279 ***
(trade union [less (0.0068)
than or equal to]
[[??].sub.1])
Employment in finance 0.0157 **
(trade union > (0.0064)
[[??].sub.1])
Employment in finance 0.0191 ***
(social security (0.0063)
expenditure [less
than or equal to]
[[??].sub.1])
Employment in finance 0.0027
(social security (0.0063)
expenditure >
[[??].sub.1])
Employment in finance 0.0025
(welfare generosity (0.0065)
[less than or equal
to] [[??].sub.1])
Employment in finance 0.0152 **
(welfare generosity (0.0065)
> [[??].sub.1])
Employment in finance
(redistribution [less
than or equal to]
[[??].sub.1])
Employment in finance
(redistribution >
[[??].sub.1])
[[??].sub.1] 23.75 12.31 37.00
Controls Yes Yes Yes
Observations 260 260 260
[R.sup.2] 0.2219 0.2444 0.1950
Number of country 13 13 13
Wage volatility
HS LS
(7) (8)
Single threshold
Employment in finance
(trade union [less
than or equal to]
[[??].sub.1])
Employment in finance
(trade union >
[[??].sub.1])
Employment in finance
(social security
expenditure [less
than or equal to]
[[??].sub.1])
Employment in finance
(social security
expenditure >
[[??].sub.1])
Employment in finance
(welfare generosity
[less than or equal
to] [[??].sub.1])
Employment in finance
(welfare generosity
> [[??].sub.1])
Employment in finance 0.0543 *** 0.0438 ***
(redistribution [less (0.0110) (0.0113)
than or equal to]
[[??].sub.1])
Employment in finance 0.0380 *** 0.0261 **
(redistribution > (0.0111) (0.0115)
[[??].sub.1])
[[??].sub.1] 33.19 31.94
Controls Yes Yes
Observations 260 260
[R.sup.2] 0.2935 0.2448
Number of country 13 13
* P<0.1, ** P<0.05, *** P<0.01.
Note: Standard errors in parentheses. FIS = high-skilled workers
and LS = Low-skilled workers. Financial variable: Employment share
in the financial sector.
