Resilience, consumption smoothing and structural policies Paper prepared for the OECD Workshop ”Structural reforms and economic resilience: evidence and policy implications”
Ekkehard Ernst OECD Economics Department 2, Rue Andr´e Pascal 75775 Paris Cedex 16
[email protected]
Gang Gong Tsinghua University School of Economics and Management Beijing, China
Willi Semmler New School Unversity Dept. of Economics Schwartz Center for Economc Policy Analysis New York June 2007
Abstract A dynamic stochastic general equilibrium model with non-clearing labour markets and imperfect competition is estimated for 21 OECD countries. The model provides for parameters both for preferences and technology and for the degree of structural rigidities on labour and product markets. These parameters are matched with policy indicators, establishing a relation between the model-based employment adjustment cost parameter and the strictness of employment protection legislation, and between the average size of the mark-up on product markets and the OECD product market regulation indicator. Policy simulations are carried out by modifying the parameters relating to labour and product market rigidities in order to evaluate the impact of a more flexible markets on consumption smoothing. Moreover, the paper establishes a trade-off between more volatile employment and improved consumption smoothing possibilities. Finally, using a standard utilitarian welfare function, the model assesses the optimal stance of structural policies when taking this trade-off into account. JEL Codes: E32, C61 Keywords: Product market competition, labour market frictions, business cycles, structural reforms in OECD countries, RBC models
1
Introduction
A rapid recovery after adverse demand shocks and a vigorous adjustment to supply shocks in short a country’s resilience - have been identified as a major policy field for many OECD countries. While traditionally macroeconomic policies play the dominant role in smoothing the business cycle, the importance of structural policies on labour and product markets has been recognized in order to understand country differences in the adjustment dynamics and the transmission of shocks. In contrast to the analysis of the role of structural reforms in improving the long-run performance of employment and productivity of OECD countries, however, research regarding the role of reforms for a country’s resilience has only recently gained prominence. In this regard, the speed of adjustment of either prices or quantities to shocks occupies a prominent place in the literature in explaining the welfare costs of business cycle fluctuations. Nominal rigidities force quantities to adjust to (temporary) demand shocks, with adverse consequences on disposable income and consumption. Real rigidities, on the other had, will slow down the adjustment process to (permanent) supply shocks, thereby limiting the country’s potential to benefit from supply side improvements and protracting the slump following adverse shocks (Caballero and Hammour, 1998; Caballero, 2007). Policies that help to raise the adjustment speed to either type of shock will, hence, not only improve the short-term dynamic behaviour but also raise welfare in the long-run. A second, less-well research question that arises in the context of policies to increase resilience concerns a country’s capacity to smooth consumption over the cycle. Consumption smoothing is a welfare-enhancing reaction of risk-averse households to income shocks. To the extent that no borrowing constraints are binding the consumer on financial market, in principle all shocks can be absorbed through borrowing or lending on the credit market and the consumption path becomes a random walk, disconnected from the income stream (Hall, 1978). However, this result has to be qualified in an economy where households also decide upon their labour supply as they have some control over their income stream. With two decision variables at their disposal, households can control their consumption stream not only via adjustments to net lending on the financial market but also through adjusting hours worked and labour income. To the extent, however, that structural rigidities on labour and product markets prevent hours worked to evolve according to the household’s optimal decisions, the possibilities for consumption smoothing will equally be affected. In order to address these issues in a coherent way, the following paper presents a dynamic general equilibrium model with capital accumulation, imperfect competition on product markets and matching frictions on labour markets. It offers the perspective to estimate various structural rigidities in a model-coherent and consistent way. In particular, the model includes a rich structure of the labour market and allows to estimate parameters that reflect employment adjustment costs and wage bargaining power. It also accounts for the degree of competition on the product market, making use of a sticky information assumption regarding the price setting process. Moreover, by introducing capital stock dynamics the model allows 1
for shocks to have persistent effects on product and labour markets through the investment channel. The model builds, as Uhlig (2004) and Blanchard and Gali (2005), on frictions on the labour market and allows for price and wage stickiness. We make use of a simplified searchand-matching methodology to model the effective level of employment given (sticky) wages. In this set-up, employment evolves as a result of a match between labour demand and supply. Moreover, the product market is characterised by imperfect competition that leads to a markup in the retail sector and a socially sub-optimal level of production. (Relative) prices are set as a result of sticky information whereby firms update their expectations regarding the future developments of the market with a lag. Together, these assumptions allow to estimate the degree of real wage rigidity, employment adjustment costs and the size of the mark-up on product markets. Despite the limited number of rigidities the model allows to develop some new insights into potential effects of structural market reforms on the cyclical behavior of OECD economies, their impact on consumption smoothing, employment volatility and on households’ overall welfare. The model’s parameters are estimated for 21 OECD countries. In order to demonstrate that the estimated parameters properly reflect standard wisdom regarding the microeconomic distortions in OECD economies, they are matched with structural indicators taken from the OECD International Regulation Database and the OECD Employment Outlook. This allows to establish a relation between (i) the model-based employment adjustment cost parameter and the indicator for employment protection legislation, and (ii) the size of the product market mark-up and the degree of product market regulation. In the second, policy-oriented part of the paper, the paper presents policy simulations based on changes in the size of the estimated parameters relating to the degree of labour market rigidities and the importance of product market imperfections in order to evaluate the impact more flexible labour and product markets may have on the dynamic properties of the model. In particular, we are looking at the correlation between consumption and employment as a measure of consumption smoothing. In the policy simulations, it is shown that reducing employment protection legislation and hence labour adjustment costs for firms, households are able to improve risk sharing over the cycle and experience a decrease in their consumption volatility by means of adjusting their labour supply (more easily) to its optimal level. When firms face high employment adjustment costs, households are precluded from such a possibility as actual employment will evolve only sluggishly in response to any shock and employment volatility is low. By way of these policy experiments the paper also assesses to what extent increasing product market competition can constitute a substitution for a decrease in employment adjustment costs. The paper shows that product market competition is less effective in improving consumption smoothing but that it complements labour market reforms as the joint effect of reforms on both markets exceeds the individual effect of each reform taken separately. Finally, we assess the trade-off that arises between improved consumption smoothing and higher employment volatility following the implementation of these structural reforms. Using a standard utilitarian welfare function, we show that under 2
certain circumstances, a less ambitious reform package may turn out second-best optimal. The remainder of this paper is organized as follows. Section 2 presents the structure of the model, which is calibrated in section 3 for 21 OECD countries. After interpreting our estimates of structural rigidities, section 4 studies the possible impacts of structural reforms in the context of such a model by undertaking simulations. Section 5 discusses the trade-off between employment volatility and consumption smoothing and the consequences for welfare-optimizing policies. Section 6 concludes the paper.
2 2.1
Sticky information and labour market frictions Setting up the model
The economy is characterized by a representative household and monopolistically competitive intermediate goods producers. Agents enter market exchanges in three markets: the product, the labour and the capital market. The household owns all factors of production and sells factor services to firms and buys (final) products for consumption or accumulation of the capital stock. The product market is assumed to be imperfectly competitive, with the individual firms facing a perceived demand curve and a sticky price (fixed at P = 1). Wage stickiness. Following Gong and Semmler (2006) wages are supposed to be sticky, with wage updating taking place at the pace of labour contract renewal supposed to be fixed at rate ξ (Calvo wage setting), corresponding to its persistence: wt = ξwt−1 + (1 − ξ) wt∗ Only a certain fraction of existing contracts will be renewed, for all other labour contracts the previous wages continue to hold. The optimal bargained wage, on the other hand, is determined through monopoly union bargaining, where households maximize their utility against the constraint of firms’ intertemporal labour demand. Et max ∗ wt
subject to
∞
i
β U (ct+i , nt+i )
i=0
1 [(1 − δ) kt+i + f (kt+i , nt+i , At+i ) − ct+i ] kt+i+1 = 1+γ wt∗ = fn (kt+i , nt+i , At+i )
Above, k is the capital stock, c the consumption, n the labour, A the technology; β designates the intertemporal preference rate; δ the depreciation rate; and γ stands for the stationarity parameter. Assuming a constant relative risk aversion (CRRA) utility function for the household’s instantaneous utility U (c, n) = ln (c) + θ ln (1 − n) 3
where c: consumption, n: employment and θ the elasticity between consumption and work effort and let f (kt , nt , At ) be the firm-level production function, the solution to this optimisation problem can be written as:
(θα − 1) wt∗ = At α 1−δ 1 α
1−α α
(1)
Implicitly the above optimisation problem may also allow us to derive a sequence of an optimal consumption plan {ct+i }∞ i=0 . However, this plan has no feedback effect on the optimum wage wt∗ but is solved for the given wt∗ as expressed in (1), On the other hand, that optimal plan may not be carried out since the actual wage wt is not necessary equal to wt∗ . Factor markets. Unlike the standard RBC model with competitive markets, factor markets in this model will be re-opened at the beginning of each period t, necessary to ensure adjustment in response to a non-clearing labour market. The non-clearing of the market is related to wage stickiness following the Calvo wage setting and the sequence of bargained wages {wt∗ }∞ t=0 . The dynamic decision process is an adaptive one. It exhibits two stages: in a first step, households determine their consumption and optimal wage pattern (and hence indirectly their labour supply), in a second step, they re-optimize their consumption plans following the realized transactions on the labour and product market. At period t, the representative household expects a series of technology shocks {Et At+i }∞ 0 and real wages and interest rates {Et wt+i , Et rt+i }∞ where i=0,1,2,. . . The decision problem 0 of the household is then to choose a sequence of planned consumption and labour effort ∞ d s ct+i , nt+i such that 0
max ∞ Et ∞ {cdt+i}=0 ,{nst+i}i=0 subject to s = kt+i+1
∞
i
βU
cdt+i , nst+i
i=0
1 s s (1 − δ) kt+i + f kt+i , nst+i , At+i − cdt+i 1+γ
where superscripts d and s stand for “demand” and “supply”. Using standard dynamic programming techniques, this optimal planning problem can be solved toyield the solution ∞ d s ; however, from each sequence only the first tupel cdt , nst is actually sequence ct+i , nt+i i=0 carried out.
In the first period t, the firm decides upon its inputs ktd , ndt given its output constraint Ey t related to its perceived demand curve. Standard (one-period) profit maximization yields the factor demand functions: ktd = fk (rt , wt , At , Eyt ) ndt = fn (rt , wt , At , Eyt ) 4
As the capital market is supposed to be perfectly competitive, the rental rate of capital, rt , adjusts in each period such as to clear the market: kt = kts = ktd . On the labour market, however, the fixed wage contract does usually not allow to clear the market1 . Wage rigidities introduce frictions on the labour market and force market participations to carry out transactions off their optimal supply and demand schedule. Following the spirit of the search-and-matching literature2 , we assume that these rigidities imply that actual employment (i.e. transactions) corresponds to a weighted average between labour supply and demand at the current wage: nt = ωndt + (1 − ω) nst , where ω measures the degree to which employment is determined by labour demand and will play a key role in the interpretation of the model and its results. This equation indicates that actual employment can be a result of a matching process whereby not all desired transactions are carried out, but where - due to the probabilistic nature of the process - firms may end up hiring more than what their current needs are. This may also happen when, for instance, employment is negotiated, when firms hoard labour in downturns, employing more than the profit-maximizing level of workers or when some other real rigidities are present. For the moment, we leave the interpretation open and only notice that observed employment may not necessarily correspond to desired levels. We give a more detailed interpretation of the ω-parameter as soon as we have estimated the model. Product markets. The final good is produced by combining intermediate goods. This process is described by the following CES function: Yt =
1 0
1ρ
yitρ di
(2)
where ρ ∈ (−∞, 1). ρ determines the elasticity of substitution between the various inputs. The producers in this sector are assumed to behave competitively and to determine their demand for each good, yit , by maximizing the static profit equation: max Pt Yt − {yitd }i∈(0,1)
1 0
Pit yit di
subject to (2). Given the general price index is supposed to remain constant and normalised to unity, the demand for intermediate goods depends only on the relative prices of intermediate goods, Pit , and the aggregate demand: 1
yitd = Pitρ−1 Yt 1
This may nevertheless happen if either the representative firm has perfect foresight of the sequence of technology shocks or the wage contract is arranged in the form of a contingency plan. Both will be excluded here; see Gong and Semmler (2006) for a discussion on this latter point. 2 We do not follow the precise set-up here, mainly for reasons of analytical simplicity.
5
The final good may be used for consumption and investment purposes. At the level of the intermediate goods production, each firm i, i ∈ (0, 1), produces an intermediate good by means of capital and labour according to a constant returns-to-scale technology: yit = Ait kit1−α nαit hence aggregating across all producers yields: 1 0
yit di = At ktα nt1−α
Moreover, from the retail demand for intermediate goods we know that: 1 0
yit di =
1 0
1 ρ−1
Pit Yt di = Yt
1 0
1
Pitρ−1 di.
Intermediate goods producers face costs to update the relevant information regarding the developments on their markets (Mankiw and Reis, 2002). To simplify matters, we want to assume that on average firms set their optimal prices and quantities according to information one period earlier, Pit = arg max Et−1 πt where πt = Pit yit − wit nit − rit kit . Firm i, then, sets its relative optimal price according to the following schedule: Pit =
sit−1 w α r 1−α where sit = α it it 1−α : real marginal costs. ρ α (1 − α)
Given that in equilibrium, all producers are setting the same price Pit = P t > 1, we will use the following definition: 1 0
yit di =
1 Yt 1 − exp (λt )
and call λt the retail margin, measuring the degree of product market competition (via the effect of ρ on prices). Consequently, the equilibrium on the goods market writes as: 1 0
yit di = At kt1−α nαt ⇔ Yt = (1 − exp (λt )) At kt1−α nαt
Calibrating the model. In order to implement the model empirically, certain specifications regarding the preference function, the technology shock and the stationarity of the time series have to be made. As noted above, the economy is represented by a consumer characterized by an instantaneous utility function over consumption, c, and leisure, l = 1 − n: U (c, n) = ln (c) + θ ln (1 − n) 6
with θ the elasticity between consumption and leisure to be estimated with the data. Moreover, technological shocks are supposed to follow and AR(1) process:
At+1 = a0 + a1 At + εt where εt ∼ N 0, σε2
The stationarity parameter, γ, can be retrieved by calculating the trend growth rate of output. Finally, employment, nt , is based on (normalised) hours worked (sample mean N ), considering that only 1/3 of a day is dedicated to work on average. Box 1 (see above) summarises the relevant model equations. Box 1: Summary of the model equations Consumption: cdt+1 = Et+1 β (1 − δ + rt+1 ) cdt Labour supply: β
wt (1 − δ + rt+1 ) (1 − nst ) = 1 − nst+1 wt+1
Pre-set wages: wt = ξwt−1 + (1 − ξ) wtopt + εw Bargained optimal wage:
1 1−α
At
wtopt = Actual employment:
θα
1 1−α
−α
α 1−α
1−δ
1−α α
nt = E t|t−1 ωndt−1 + (1 − ω) nst−1 Capital accumulation: kt+1 = (1 − δ) kt + Yt − cdt Production function of intermediaries: yit = f (kit , Ait nit ) = At kit1−α nαit Product market equilibrium: Yt = (1 − λt ) Factor prices:
1 0
yit di = cdt + it
wt = fn kt , At ndt rt = fk (kt , At nt )
7
3 3.1
Estimation of the model Estimation of structural parameters
In order to estimate the model described in the previous section, several parameters have to be determined. These include: (i) the parameters describing the process of technological progress and wage growth; (ii) the preference parameters and the depreciation rate of the capital stock and (iii) the parameters describing the rigidities on labour and product markets. While the first parameters can be estimated easily on the basis of an AR(1) process using the TFP residuals that can be derived from a standard growth accounting exercise, the preference parameters are deeply linked to the first-order conditions that result from solving the above dynamic programming problem. This fact can be used to apply GMM techniques in order to estimate these parameters. Concretely, the parameters are chosen such as to match the moments of the model described by the first-order conditions of the above model to those of the underlying data. Notice, moreover, that these parameters can be established without a concrete knowledge about the underlying labour and product market rigidities as they are supposed to be unrelated to it. Given the highly non-linear nature of the optimisation problem, the algorithm used to pick the right parameters β, δ and θ had to ensure that any local optimum of the GMM technique is to be avoided. Here, a technique called simulated annealing has been applied, that combines a grid search procedure with an objective function to assess the size of the grid jumps. The resulting parameters for our 21 countries can be found in the following table 1. As to the remaining parameters, the wage share, α, has been taken from country tables, averaging the values over the corresponding periods for these countries, while the wage persistence, ξ, has been estimated by OLS. While the time preference rates are relatively close across countries, corresponding to the standard interval for these models between 0.95 and 0.99, the country sample displays a large range of values for the capital depreciation rates, probably reflecting some country specific trends. In particular the value for Finland seems to be excessively large, implying an annual depreciation rate of 43%; this may be related to the particular events surrounding the deep economic crisis in 1993. The two parameters β and θ seems to fall into a reasonable range, although it must be conceded that no commonly accepted estimates exist regarding the substitution elasticity between consumption and leisure. Finally, the estimates for the wage persistence should be taken cum grano salis as OLS estimates of lagged dependent variables are known to be upward biased; they may nevertheless give a sense of the importance of real wage persistence across countries with small open economies in general being characterized by less persistence than bigger, in particular continental European economies. Yet, overall, as our model in section 2 postulates, our estimates reveal a strong wage persistence, ξ, and thus a very weak endogeneity of wage determination.
8
Table 1: Structural parameters Country Australia Austria Belgium Canada Denmark Finland France Germany Ireland Italy Japan Korea (Rep.) Netherlands New Zealand Norway Portugal Spain Sweden Switzerland United Kingdom United States
α 0.58 0.62 0.64 0.61 0.60 0.60 0.63 0.61 0.59 0.66 0.68 0.74 0.62 0.55 0.54 0.64 0.63 0.61 0.60 0.65 0.64
β 1.000 0.980 0.990 0.969 0.985 0.960 0.981 0.985 0.980 0.992 0.999 1.000 0.983 1.000 1.000 0.975 0.980 0.979 0.988 0.978 0.960
δ 0.024 0.007 0.007 0.021 0.015 0.094 0.017 0.017 0.015 0.012 0.008 0.025 0.022 0.045 0.052 0.032 0.024 0.014 0.014 0.011 0.053
θ 3.18 2.57 1.88 1.76 2.57 1.91 1.81 1.80 1.79 1.97 2.20 2.50 1.87 3.00 2.93 1.93 1.89 1.75 2.58 1.80 1.78
ξ 0.99 0.94 0.93 0.99 0.93 0.97 1.00 1.01 0.99 1.02 0.97 0.97 0.98 0.95 0.98 0.97 0.99 0.97 0.98 0.97 0.98
Note: The table reports the estimates of the structural parameters for the substitution elasticity between capital and labour (α), the intertemporal time preference (β), the depreciation rate of capital (δ); the substitution elasticity between labour consumption and leisure (θ) and the persistence of real wages (ξ). Source: Own calculations
3.2 3.2.1
Estimations of labour and product market rigidities Estimation methodology and results
Having determined the structural parameters relating to preferences, capital depreciation, labour share and wage persistence, the full model must now be estimated to determine the degree of product market competition and to establish the extent to which labour market frictions arise from matching frictions between labour supply and labour demand. Given that we are mainly interested in cross-country differences of product and labour market competition, 9
in the estimation procedure both parameters will be averaged across the (country-specific) time span. On the basis of the deep structural parameters estimated in the first step above, the model can be calibrated using the realized technology shocks (instead of the simulated ones), which corresponds to the standard RBC presentation. This calibration can be used to recover the theoretical aggregate demand, labour demand and labour supply decisions of firms and households at each point in time. Using a grid-search algorithm to search the parameter space (ω, λ), the parameters are set such as to minimize the residual square sum of the difference between actually observed employment and model-generated employment. In formal terms: (ω, λ) = arg min
nt − ωndt (Yt (λ)) + (1 − ω) nst
2
t
The results of this estimation procedure and the resulting parameters have been reported in Table 2 for 21 OECD countries. For further reference, the table also reports values for institutional indicators regarding the tightness of employment protection legislation (EPL), the strictness of product market regulation (PMR), the generosity of gross replacement rates of unemployment benefits (GRR) and the degree of the coordination of wage bargaining systems (WBC). All indicators have been taken from various OECD sources. As can be seen from the table, there is strong inverse correlation between the ω- and the λ-parameter, reflecting the fact that structural rigidities on one market correspond to similar rigidities on the other market. Moreover, the table shows a clear pattern that distinguishes Continental European OECD countries from the rest of the sample by having lower estimates for ω and higher estimates for λ. Indeed, the unweighted average of the ω-parameter for the EMU countries is slightly higher than half the value for the entire sample (9.1% compared to 16.3%). Conversely, the λ-parameter is 60% bigger for the EU-12 countries than for the entire OECD sample (−3.25 compared to −5.40). Taking Ireland out of the euro area sample as a country coming close to the values of the UK and the US, the difference between the euro area sample and the rest of the OECD becomes even more pronounced with estimates for ω and λ of 7.6% and 2.56 respectively. Exactly how these estimations of the labour and product market rigidities are affected by policies and institutions, however, is not immediately obvious from the models equation. In the next section we, therefore, turn to the question of relating existing indicators of labour and product market policies to our estimates of the structural rigidities.
10
Table 2: Structural parameters and institutions country Australia Austria Belgium Canada Denmark Finland France Germany Ireland Italy Japan Korea (Rep.) Netherlands New Zealand Norway Portugal Spain Sweden Switzerland United Kingdom United States Correlation with Omega (P-value) Correlation with Lambda (P-value)
ω 0.25 0.31 0.00 0.18 0.02 0.11 0.01 0.12 0.23 0.01 0.15 0.12 0.03 0.17 0.15 0.37 0.00 0.13 0.02 0.58 0.47
λ EPL PMR GRR WBC -7.21 1.10 1.26 0.27 3.3 -6.01 2.39 1.81 0.26 4.3 -0.97 2.09 2.05 0.39 4.0 -10.16 0.64 1.37 0.27 1.4 -1.02 1.49 1.46 0.41 3.8 -1.06 2.09 2.07 0.44 4.8 -1.08 3.08 2.48 0.37 2.0 -8.14 2.78 1.90 0.27 4.0 -8.74 1.01 1.47 0.26 3.1 -1.03 3.28 2.78 0.05 2.9 -8.52 2.65 1.86 0.10 4.0 -3.64 2.00 2.47 N/A 1.0 -1.13 2.36 1.79 0.46 4.1 -7.17 1.03 1.43 0.27 2.8 -11.34 2.89 1.83 0.39 4.3 -10.39 3.75 2.13 0.35 3.6 -1.06 3.21 2.33 0.32 3.5 -10.75 2.43 1.80 0.29 3.3 -0.98 1.27 2.24 0.30 4.0 -9.15 0.51 1.15 0.18 1.6 -3.88 0.22 1.30 0.12 1.0 -0.56*** -0.45** -0.61*** -0.40* -0.40* (0.008) (0.043) (0.003) (0.081) (0.075) 0.10 0.49** 0.23 0.10 (0.682) (0.024) (0.322) (0.667)
Note: The table reports the estimates of the -parameter following the estimation procedure described in the text. In addition the table reports the OECD-indicators for unemployment benefits gross replacement rates (column GRR), the level of wage bargaining coordination (WBC), the strictness of employment protection legislation (EPL) and the tightness of product market regulation (PMR). The two rows indicate the correlation of the institutional variables with both the ω− and λ− parameter and their statistical significance, where the significance levels are indicated by asterisks ***: 1% level, **: 5% level, *: 10% level. Source: Own calculations, Conway et al. (2005), OECD (2004)
3.2.2
Identification of structural rigidities
The estimation of the model so far has provided a first picture on the structural variety among OECD countries. However, when trying to match the estimated parameters with particular 11
variables measuring labour and product market policies, the theoretical model gives only very scarce information. This is particularly relevant for the ω-parameter reflecting frictions on the labour market, as λ and ξ can be easily identified with the retail margin and real wage rigidity respectively. In the following discussion, we, therefore, concentrate in particular on the interpretation of ω. In order to assess the influence of certain policy indicators on ω, we discuss in more detail the different transmission mechanisms of policies on employment dynamics. In particular, four hypotheses can be put forward: • Replacement rates: The stronger the wage bargaining power of firms is, the more they will be able to adjust employment according to their labour demand. Higher replacement rates, however, will increase workers’ outside option and thereby limit firms’ bargaining power, hence higher replacement rates should lead to lower estimates of the ω-parameter; • Bargaining coordination and nominal wage stickiness: The labour market matching process may be influenced by nominal wage stickiness. While no immediate measure of nominal wage stickiness is available, a first proxy may be the level of coordination of wage bargaining systems: the higher the level of wage bargaining coordination, the more sticky nominal wages are across the economy as relative wage adjustments between jobs and occupations will be less determined by market forces. Here too, a negative correlation with ω would be expected; • Employment protection and labour adjustment: labour supply may determine employment the higher quantitative labour adjustment costs - induced, for instance, by strict employment protection legislation - are since labour demand has increasing difficulties to adjust. Here too, a lower ω is predicted with higher employment protection; • Speed of entry and exits of firms: labour adjustment costs may also arise from market entry and exit as firm turnover is an important element in the determination of overall labour demand. Consequently, the estimate of may be negatively correlated with rigid product market structures and regulations, since the latter may provide more job security. In order to assess which of these four variables best reflects our measure of the labour market matching process, four indicators have been taken from OECD studies. In particular, we have compared the cross-country variation of our ω-estimates with (i) gross replacement rates of unemployment benefit schemes, (ii) the degree of wage bargaining coordination, (iii) the strictness of employment protection legislation and (iv) the degree of product market regulation (see table 2 above). Following standard statistical significance measures, only the last two measures provide significant (negative) correlations with our estimated values of the ω-parameter, vindicating 12
the interpretation of measuring quantitative adjustment costs on the labour market. However, the fact that no further differentiation can be made on the ground of these correlations between labour and product market rigidities must be considered a limitation that should be addressed in a follow-up to this work. On the other hand, on the basis that the λ-parameter - that arises from imperfect competition on the product market - is only correlated with product market regulation, we will identify (the inverse of) λ with the degree of product market competition and ω with dismissal costs in the following policy experiments.
4
Consumption smoothing and structural policies
Having identified our estimates of structural rigidities with policy indicators regarding the strictness of employment protection legislation and product market regulation, we will now turn to carry out policy experiments that should give us some indication as to how these rigidities affect the short-term macroeconomic behaviour. In particular, we are interested in analysing the impact of structural rigidities on employment and consumption volatility as well as on the capacity of households to smooth consumption over the cycle, i.e. the correlation between consumption on the one hand and production, employment and wages on the other. In the following, we concentrate on a subsample of countries that can be taken as representing a larger group of countries out of the total sample with similar characteristics regarding the degree of labour and product market rigidities. Specifically, we retain Austria, France, Germany, Italy and Japan for our policy simulations.
4.1
Changes in volatility
All five countries display important differences in the variance and covariance of their main macroeconomic aggregates, in particular regarding the wage series (see table 3). Nevertheless, roughly the same picture emerges across countries as to the impact a policy change has on employment volatility, with the notable exception of Austria regarding changes in dismissal costs. The following table summarises the main results of changes in structural policies on cyclical variation of the main macroeconomic aggregates: • An increase of ω, i.e. a decrease in dismissal costs as it is interpreted here, will raise the volatility of the employment series. On the other hand, however, it leads to a decline in the volatility of consumption and production across the board. As to capital and wages, no clear pattern emerges as in particular wages remain largely unaffected by this change in employment and production volatility. • A similar observation can be made for more competitive product markets - as measured by a decrease of λ, albeit with a less marked effect than for dismissal costs. Again, and for all five countries, employment and production volatility raises (slightly), while consumption volatility declines. 13
• When both structural reforms are adjusted such as to make labour and product markets more competitive, the combined effect exceeds the sum of the individual effects. It will strongly increase employment volatility and contribute to lower consumption volatility. The effect on production volatility is ambiguous, however, with France displaying increasing volatility in contrast to the other countries in the sample.
Japan
Italy
Germany
France
Austria
Table 3: Cyclical volatility and structural reforms
Consumption Capital Employment Production Wages Consumption Capital Employment Production Wages Consumption Capital Employment Production Wages Consumption Capital Employment Production Wages Consumption Capital Employment Production Wages
Lowest ω, highest λ
Improvement of ω
Improvement of λ
6.46 6.74 22.44 36.96 94.54 3.55 2.09 3.20 7.81 92.59 3.87 3.47 7.95 14.61 163.66 3.85 2.42 5.85 11.96 364.27 4.96 2.92 10.26 19.44 94.62
5.60 6.05 20.73 27.92 94.53 3.48 2.70 9.30 7.56 92.59 3.47 3.25 8.47 10.99 163.77 3.64 2.39 8.59 9.93 454.23 4.72 2.76 12.39 15.73 94.62
6.44 6.75 22.72 37.21 94.61 3.51 2.08 3.25 7.89 92.59 3.84 3.45 8.03 14.68 163.62 3.79 2.40 5.91 12.00 364.46 4.89 2.92 10.32 19.45 94.61
Simultaneous improvement of ω and λ 5.43 6.03 21.95 27.90 94.57 3.28 2.55 11.78 8.69 92.58 3.34 3.17 10.94 11.55 163.25 3.48 2.37 10.54 10.58 357.75 4.55 2.79 14.19 16.25 94.61
Note: The table shows the volatility (variance) of the main macroeconomic aggregates resulting from 5000 simulations when different values for ω and λ are implemented. Four policy situations have been selected: (i) ω and λ are set at their most restrictive estimated values; (ii) ω is improved to the least restrictive estimate, λ is left unchanged, (iii) λ is improved to the least restrictive estimate, ω is left unchanged, and finally (iv) both ω and λ are improved to their least restrictive estimated values.
14
4.2
Macroeconomic correlations and consumption smoothing
At least as important as the impact on volatility is the reaction of the correlations of macroeconomic variables with respect to changes in structural policies and market rigidities. Phase differences of macroeconomic variables allow households to hedge income risk by diversifying their supply decisions (savings and labour supply) across different activities and over time. The correlations of these variables hence measure to what extent such a risk diversification is possible and are a potential indicator of the welfare loss of labour and product market rigidities due to the cyclical impact these rigidities have. In this regard, a noticeable drop in the consumption-employment correlation occurs for our country sample when structural policies (either ω or λ) are geared from most to least restrictive. Similarly, the correlation between consumption and production falls, except for changes of ω in the case of Austria. Finally, employment and production are somewhat less correlated when in particular labour market rigidities become less prevalent, indicating that shocks to production may have a less pervasive effect on employment in a more flexible labour market. Such a disconnection between consumption and employment and production can be taken as an indication of increased intertemporal smoothing where households manage to improve the intertemporal arbitrage of risk arising from macroeconomic (systemic) shocks. Put differently, households are better able to hedge against employment risk when the rigidities on the labour market are reduced, even though they have to face higher employment volatility. Taken together the two preceding tables allow to draw some preliminary conclusions regarding the impact of structural policies on the adjustment speed and the possibilities to hedge against exogenous shocks: • On the one hand, relaxing adjustment costs increases the reaction with respect to shocks, leading to higher volatility of the underlying variables in the case of a very rigid economy (according to the understanding of this model), in particular when no other mechanism is present that provides risk sharing or allows for a flattening of quantitative reactions. • On the other hand, the reduced correlation between employment and consumption allows for a smoother adjustment path of the economy. This has a positive first-order effect for the overall economy as it helps production volatility to decline. At the same time, this de-linking of employment and consumption causes negative second-order effects as households are more exposed to employment risk. The issue of intertemporally arbitraging risk will be further discussed in the following section.
15
Table 4: Policy experiments and macroeconomic correlations Austria Germany 1.000
E
0.733
-0.135
1.000
Y
0.808
-0.017
0.993
1.000
W 0.043
0.042
0.014
0.019
C
1.000
K
0.353
E
0.312
0.305
1.000
Y
0.839
-0.013
0.587
1.000
W 0.184
-0.047
0.214
0.110
1.000
C
1.000 0.586
1.000
E
0.720
-0.136
1.000
Y
0.798
-0.016
0.993
1.000
W 0.062
0.068
0.015
0.023
1.000
λ
0.391
E
0.276
0.297
1.000
Y
0.810
-0.013
0.579
1.000
W 0.142
-0.031
0.305
0.161
1.000
0.781
-0.127
Y
1.000
0.873
0.036
0.986
1.000
0.005
0.007
0.023
1.000
0.186
1.000
-0.015
0.478
1.000
0.043
-0.100
0.220
0.102
1.000
1.000
fo tn e m ev or p m I
fo tn e d m ev an or p m I
ω
0.087 0.830
1.000 -0.129
1.000
0.035
0.986
1.000
0.023
0.032
0.002
0.007
λ
0.774 0.868
0.063
1.000
-0.076
0.550
1.000
1.000
0.069
-0.145
0.375
0.215
1.000
W
C
K
Y
W
ω
0.424 0.736
Japan E
C
1.000
K
0.322
1.000
E
0.898
-0.122
Y
0.951
0.018
0.990
1.000
W 0.016
0.038
-0.001
0.027
E
1.000 1.000
C
1.000
K
0.092
1.000
E
0.522
-0.070
1.000
Y
0.779
-0.163
0.594
1.000
W 0.029
0.065
-0.015
0.226
1.000
0.247
1.000
0.930
-0.121
1.000
0.961
-0.029
0.996
1.000
0.539
-0.006
0.007
0.006
1.000
C
1.000
K
0.332
1.000
E
0.893
-0.125
1.000
Y
0.948
0.017
0.990
1.000
W 0.735
0.011
0.003
0.343
1.000
1.000
1.000
fo tn e m ev rop m I
1.000
fo tn e d m ev an rop m I
1.000
0.131
0.068
1.000
0.833
-0.124
0.627
1.000
0.063
-0.227
0.539
0.060
1.000
C
1.000
λ
0.257
K
0.109
1.000
E
0.047
-0.142
1.000
Y
0.691
-0.217
0.655
1.000
W 0.784
0.010
-0.027
0.005
1.000
1.000
0.926
-0.122
1.000
0.958
-0.029
0.995
1.000
0.013
0.025
0.003
0.064
1.000 0.025
1.000
ts eh gi h
fo tn e m ev rop m I
1.000 0.839
, ts e w oL
λ
K
1.000
λ
0.312
ts eh gi h
fo tn e m ev or p m I
1.000 1.000
ω
fo tn e d m ev an rop m I
1.000
1.000
0.518
, ts e w oL
ω
λ
ω
ω λ
fo tn e m ev rop m I
1.000
0.011
Italy
fo tn e m ev rop m I
W
1.000
C
ts eh gi h
Y
ω
0.509
0.307
K
, ts e w oL
E
1.000 1.000
K
C
K
λ
0.571
C 1.000
ω
fo tn e d m ev an or p m I
K
W
ω
λ
fo tn e m ev or p m I
Y
λ
ω
fo tn e m ev or p m I
E
1.000
λ
ts eh gi h
λ
ω
, ts e w oL
K
1.000
0.077
0.007
1.000
0.780
-0.156
0.653
1.000
0.062
-0.221
0.617
0.064
ω
C C
Note: The table shows the calibration results of 5000 simulations when implementing different values for the ω and λ are implemented. Four policy situations have been selected: (i) ω and λ are set at their most restrictive estimated values; (ii) ω is improved to the least restrictive estimate, λ is left unchanged, (iii) λ is improved to the least restrictive estimate, ω is left unchanged, and finally (iv) both ω and λ are improved to their least restrictive estimated values. Capital letters indicate the macroeconomic aggregates: C consumption, K capital, E employment, Y production, W wages.
16
In order to further test the importance of consumption smoothing, the following table provides an overview of the relative volatility of real private consumption with respect to both employment and production volatility. If the above conjecture is correct, i.e. that a decrease of the consumption-employment correlation entails an increasing intertemporal smoothing, the rise in savings should be reflected as a decrease of the relative volatility of consumption with respect to either production or employment or both as it would help to smoothen consumption relative to these two aggregates. Table 5: Intertemporal smoothing (relative consumption volatility)
with respect to production
with respect to employment
Relative consumption volatility Austria France Germany Italy Japan Austria France Germany Italy Japan
Lowest ω, highest λ
Improvement of ω
Improvement of λ
0.288 1.110 0.488 0.658 0.484 0.175 0.454 0.265 0.322 0.255
0.270 0.374 0.409 0.424 0.381 0.201 0.461 0.316 0.367 0.300
0.283 1.081 0.479 0.641 0.474 0.173 0.445 0.262 0.316 0.251
Simultaneous improvement of ω and λ 0.247 0.279 0.305 0.331 0.321 0.195 0.377 0.289 0.329 0.280
Note: The table presents the relative volatility of consumption both with respect to employment and wages as a measure of intertemporal risk sharing when different values for ω and λ are implemented. Four policy situations have been selected: (i) ω and λ are set at their most restrictive estimated values; (ii) ω is improved to the least restrictive estimate, λ is left unchanged, (iii) λ is improved to the least restrictive estimate, ω is left unchanged, and finally (iv) both ω and λ are improved to their least restrictive estimated values.
In this regard, table 5 confirms our initial conjecture. For both labour and product market reforms the relative volatility of consumption with respect to employment declines, albeit more strongly for decreases in the ω-parameter. Interestingly to note is that consumption smoothing even can be observed for Austria for which consumption volatility has not declined with more flexible labour markets (as documented by table 3). As regards consumption smoothing relative to production, reforms on the labour do not seem to produce the expected results, despite the fact that table 4 indicated a reduction in the correlation between consumption and production even for an increase in the ω-parameter. On the other hand, product market reforms unambiguously allow to reduce consumption volatility relative to the variance of production. 17
5
The trade-off between volatility and efficiency
The previous section has shown that the improved risk insurance implied by a change in employment adjustment costs and higher product market competition results in an improvement in the relative dynamics of the different macroeconomic aggregates. It, therefore, helps households to smoothen their consumption path across time, implying a reduced overall volatility of the consumption path. However, a privatisation of employment risks occurs where households are forced to insure individually against employment risks (assuming that the capital market allows to do so). In the model here, this may nevertheless beneficial as it decreases both overall production and consumption risk and with a CRRA utility function, this will decrease the risk premium for the economy and thereby improve its welfare. Depending on the relative importance of reduced consumption volatility versus increased employment volatility, however, a trade-off between the two may arise with consequences for the optimal setting of structural reforms on product and labour markets. In order to evaluate how this trade-off affects the optimal stance of structural reforms from the point of view of economic welfare, we follow Lucas (1987) by estimating the welfare costs of consumption and employment volatility on the basis of: 1 1 2 Welfare Costs = σc2 + θ σN 2 2 2 2 with σc : consumption variance and σN : employment variance, corresponding to the CRRA preferences of households in our model. The following figure 1 presents the log-inverse of the welfare costs as a measure of economic welfare for four countries of our sample and at different values of the ω- and λparameters in order to account for the effect of structural reforms on households’ welfare. The welfare function is generally well-behaved, yielding a unique optimum, and no indication as to the presence of alternative, local optima. The optimal parameter combinations are indicated for each country in the note to the figure and allow to draw the following conclusions: • For either Austria and Germany, the trade-off between consumption smoothing and employment volatility does not develop strong effects on economic welfare. In both countries, the consumption smoothing effect is strong enough to make households prefer fully flexible labour (high ω) and product (low λ) markets. • In the case of Italy and Japan, however, the trade-off does play a role and pushes the optimal stance for structural reforms towards the interior of the graph, with more regulation on both product and labour markets. For Japan, this result has to be interpreted carefully, though, as the welfare function is relatively flat at the top, leading to welfare estimates that are very close to each other for a wide range of policy parameters. Together, these results point to the possibility - but not the necessity - that the impact of structural reforms on employment volatility may be strong enough to have countries 18
prefer a less ambitious reform program. This, however, depends on the initial regulatory position, the relative weight of consumption versus employment volatility in the welfare function (measured by θ in our model) and the increase of employment volatility relative to the change in the variance of consumption and may evolve over time. Figure 1: Structural reforms and economic welfare Austria
Germany -0.20
0.00 -0.20 -0.40 -0.60 -0.80 -1.00 -1.20
-0.22
er fal e W ic m on coE
-0.24 -0.26 -0.28 -0.30
-1.40 -1.60 4 3 . 1
7 .9 8 -
1 -
8 .1 7 -
λ
1 1 . 1 -
2 .0 1 -
8 5 . 0
5 .3 0
4 .2 0
7 1 . 0
5 .1 0
2 .1 0
7 0 . 0
2 .0 0
-0.32
4 3 . 1 1 -
0 .0 0
1 0 . 0
7 .9 8 -
ω
8 .1 7 -
λ
1 .1 1 -
2 .0 1 -
8 5 . 0
5 .3 0
Italy
4 .2 0
7 1 . 0
5 .1 0
2 1 . 0
7 .0 0
2 .0 0
1 0 . 0
0 .0 0
ω
Japan -0.20
-0.200 -0.205 -0.210 -0.215 -0.220 -0.225 -0.230
-0.22
er alf e W ic m on coE
-0.24 -0.26 -0.28
-0.240 7 .9 8 -
λ
8 .1 7 -
1 .1 1 -
2 0 . 1 -
8 .5 0
5 .3 0
4 .2 0
7 1 . 0
5 .1 0
2 1 . 0
7 .0 0
2 .0 0
1 0 . 0
er alf e W ic m on coE
-0.30
-0.235 4 3 . 1 1 -
er alf e W ic m on oc E
-0.32
4 3 . 1 1 -
0 .0 0
7 .9 8 -
ω λ
8 .1 7 -
1 .1 1 -
2 0 . 1 -
8 .5 0
5 .3 0
4 .2 0
7 1 . 0
5 .1 0
2 1 . 0
7 .0 0
2 .0 0
1 0 . 0
0 .0 0
ω
Note: The optimal structural reform values implied by the analysis of economic welfare are:
ω λ
Austria 0.35 -11.34
Germany 0.17 -11.34
Source: Own calculations
19
Italy 0.07 -8.27
Japan 0.07 -0.97
6
Conclusion
The paper has attempted to assess the role product and labour market policies play for the cyclical properties of OECD countries. In particular, we have aimed at evaluating the importance of structural reforms on consumption smoothing as one particular measure for a country’s resilience to shocks. A dynamic model with imperfect product market competition and labour market frictions has been set up and estimated for 21 OECD countries. Using widely used institutional indicators that measure labour and product market policies - such as employment protection legislation, product market regulation and others - we have concluded that the two main estimates for structural rigidities reflect employment adjustment costs and the degree of product market competition. Finally, we have explored the effects of structural reforms on labour and product markets by way of simulations in order to assess the likely impact of the reforms on the volatility and correlation of the main macroeconomic time series as well as on economic welfare. On the basis of these simulation results the paper allows to draw two policy-relevant conclusions: • Improving the functioning of either the labour or the product market helps to lower consumption volatility, both absolutely and relative to employment volatility. Moreover, reforms on these two markets improve the household’s capacity to disconnect its consumption stream from shocks to employment and production. • The improved consumption smoothing comes at a cost, however, as employment becomes more volatile. This may not necessarily have a strong impact on economic welfare and depends on the initial regulatory position, the relative weight of consumption versus employment volatility in the welfare function and the increase of employment volatility relative to the change in the variance of consumption. For some countries, however, this may imply that less than fully deregulated product and labour markets are optimal form an economic welfare perspective. Overall, the paper presents first evidence on the impact of structural rigidities and reforms on the dynamic properties of OECD economies. Moreover, the model gives an indication as to the potential of these reforms to improve upon economic welfare through stronger resilience. An aspect that has not been taken up in this model - and that may be relevant in particular in relation with consumption smoothing - concerns frictions on the financial market. Simple search frictions as well as asymmetric information and moral hazard have been shown to be pervasive on credit markets, restricting households in their capacity to hedge against income risk through net lending. We leave a more detailed development of these issues within the above framework to future research.
20
7
References 1. Blanchard, O. and J. Gali (2005), ”Real wage rigidities and the New Keynesian model”, Manuscript, MIT; 2. Caballero, R. J. and M. Hammour (1998), ”The Macroeconomics of Specificity”, Journal of Political Economy, 39/2, pp. 257-273; 3. Caballero, R. (2007), Specificity and the Macroeconomics of Restructuring, Cambridge: MIT-Press; 4. Conway, P., V. Janod, and G. Nicoletti (2005), ”Product Market Regulation in OECD Countries, 1998 to 2003”, OECD Economics Department Working Paper, no. 419; 5. Gong, G, and W. Semmler (2006), Stochastic Dynamic Macroeconomics-Theory and Emprical Evidence, New York: Oxford University Press; 6. Hall, R. E. (1978), ”Stochastic Implications of the Life Cycle-Permanent Income Hypothesis: Theory and Evidence ”, Journal of Political Economy, 86/6, pp. 971-987; 7. Lucas, R. (1987), Models of Business Cycles, Boston: Blackwell Publishing; 8. Mankiw, N.G. and R. Reis (2002), ”Sticky information versus sticky prices: A proposal to replace the New Keynesian Phillips Curve”, Quarterly Journal of Economics, no. 117/4, pp. 1295-1328; 9. OECD (2004), Employment Outlook, Paris;
10. Uhlig, H. (2004), ”Macroeconomics and Asset markets: Some Mutual Implications”, manuscript, Dept. of Economics, Humboldt University, Berlin.
21
