Do government purchases a ect unemployment? · Do government purchases a ect unemployment? Steinar...

Do government purchases affect unemployment?

Steinar Holden and Victoria Sparrman∗

University of Oslo, Department of Economics

November 29, 2012

Abstract

We investigate empirically the effect of government purchases on unemployment in20 OECD countries, for the period 1980-2007. Compared to most earlier studies weuse a more extensive data set, which allows for controlling for a host of factors thatinfluence the effect of government purchases. We find that increased government pur-chases lead to lower unemployment; an increase equal to one percent of GDP reducesunemployment by about 0.3 percentage point in the same year. The effect is greaterin downturns than in booms, and also greater under a fixed exchange rate regime thanunder a floating regime. The effect on unemployment reflects a corresponding positiveeffect of increased government purchases on the employment to population rate.

Keywords: Fiscal policy, unemploymentJEL codes: E62, H3

1 Introduction

During the financial crisis, most OECD countries used fiscal measures extensively to stim-ulate the economy. More recently, increasing public debt and rising default premia onsovereign debt have led to substantial fiscal tightening in many countries. At the sametime, unemployment has soared in many OECD countries. The large changes in fiscal pol-icy and unemployment rates raise the question of how fiscal policy affects unemployment,irrespective of what the motivation for the policy is. This paper explores an importantpart of the fiscal policy, the effect of a change in government purchases of goods andservices on aggregate unemployment.

The effect of fiscal policy on the economy has been subject to considerable interestin the recent years, cf. surveys in Auerbach et al. (2010), Perotti (2007), Beetsma andGiuliodori (2010), Hall (2009) and Ramey (2011). The bulk of this literature has dealtwith the effect of fiscal policy on GDP, while the literature exploring the effect on unem-ployment is much more limited. The distinction between GDP and unemployment is notunimportant: While more output in general requires higher employment and thus lowerunemployment, there are other effects that make the link between GDP and unemploymentless clear. Fiscal actions that lead to increased labour supply may increase unemploymenteven if output grows. Alternatively, if increased public spending crowds out private sectoroutput, and productivity is higher in the private sector, unemployment may fall even if

∗We are grateful to Roel Beetsma, Francesco Furlanetto, Hasem Pesaran, Ragnar Nymoen and FredrikWulfsberg as well as participants at presentations at the BI Norwegian Business School, the GRASP-ESOP workshop and Statistics Norway for comments and discussions. The numerical results in this paperwere obtained by use of OxMetrics 6/PcGive 13 and Stata 12. The paper was started within the projectDemand, unemployment and inflation financed by the Research Council of Norway. The paper is partof the research activities at the centre of Equality, Social Organization, and Performance (ESOP) at theDepartment of Economics at the University of Oslo. ESOP is supported by the Research Council ofNorway.

1

aggregate output goes down. This ambiguity is reflected in recent research: While Mona-celli et al. (2010), IMF (2010)) and Ramey (2012) conclude that an increase in governmentpurchases lead to lower unemployment, Bruckner and Pappa (2012) find that increasedgovernment purchases lead to higher unemployment due to increased labour force partici-pation. In the current setting, when unemployment is high on the policy agenda, it seemspertinent with more research on the effect of fiscal policy.

Our study differs from most of the previous studies along several dimensions. First, weuse an extensive panel data set for 20 OECD countries for the period 1980-2007 (some ofthe analysis also includes 1960-1979), which makes it possible to explore whether the effectof fiscal policy depends on a host of other factors, like the cyclical situation of the economy,the type of fiscal impulse, etc. A number of recent papers argue that the effect of fiscalpolicy depends crucially on the possible monetary response (e.g. Eggertson and Woodford(2003), Coenen et al. (2010) and Hall (2009)); we explore this idea by considering how theeffect differs across monetary regimes. Second, we draw upon a large literature, associatedwith among others Layard et al. (1991) and Nickell et al. (2005), which has documented theimportance of labour market institutions for the evolution of the rate of unemployment.Our analysis builds on this literature, investigating the effect of fiscal policy in a regressionframework, controlling for other explanatory variables, including institutions in the labourmarket. This also allows us to explore whether the effect of fiscal policy depends on labourmarket institutions (which we find that it does).

We find that an increase in government purchases equal to one percent of GDP leadsto a reduction in unemployment of about 0.3, irrespective of whether we use fixed effectsor IV estimation. The effect increases somewhat in year 2, for then to decrease graduallyand vanish after 10 years. The size of the effect is highly dependent on other factors in theeconomy. We find that the change in unemployment due to a rise in government purchasesis countercyclical, being greater when the economy is in a downturn; this is consistent withthe recent findings of Auerbach and Gorodnichenko (2012) and Nakamura and Steinsson(2012). Furthermore, we find a strong effect of fiscal policy on unemployment in countrieswith a fixed exchange rate, but less or no effect for countries with a floating exchange rate.This is consistent with recent research mentioned above, arguing that fiscal policy mayhave a strong impact on the economy in situations when the monetary policy is constrained(see e.g. Coenen et al. (2010), Christiano and Rebelo (2011) and Woodford (2011)).We also find a positive effect of increased government purchases on the employment topopulation rate, which corresponds to the negative effect on unemployment. The effecton employment is stronger when the economy is in a recession than a boom.

The rest of the paper is organized as follows. In section 2 we provide a brief review ofthe theoretical and empirical literature on the effect of fiscal policy on GDP and unem-ployment. Section 3 presents our empirical specification, while the empirical results arelaid out in section 4. Section 5 concludes. The data is described in the data appendix.

2 The effect of fiscal policy on unemployment - a compari-son with the literature

Most of the literature on the effect of fiscal policy focuses on the effect on GDP, seesurveys referred to above. While there is large variation in the findings, most studies findthat increased government spending has a positive impact on GDP, even if the size ofthe effect may vary considerably depending on the specific circumstances. The effect onunemployment is clearly linked to the effect on GDP, as one would expect an increasein output to be associated with higher employment. However, as fiscal policy is alsolikely to affect labour supply, a rise in employment may not lead to lower unemployment.Furthermore, if increased public spending crowds out production in the private sector, and

2

productivity is higher in the private sector, employment may increase (and unemploymentfall), even if the impact on GDP is small. In this brief review we will focus on the effect onunemployment and employment to save space. However, in the discussion of the empiricalresults below, we will also compare with related studies focussing on the effect on GDP.

An early study is Holmlund and Linden (1993), who explore the effects of public em-ployment in a calibrated search model. They find ambiguous effects on unemployment,as increased wage pressure may counteract the direct unemployment-reducing effect ofincreased public employment. More recently, Michaillat (2012) analyzes a general equi-librium search model where increased public employment leads to lower unemployment.Monacelli et al. (2010) explore the effect of government consumption in a neoclassicalmodel augmented with search and matching frictions. They show that while higher gov-ernment consumption increases the hiring rate due to the negative wealth effect inducinghigher labour supply, this effect is dampened by the rise in the real interest rate. Overall,the effect of an increase in government spending equal to one percent of GDP leads to areduction in the rate of unemployment of 0.2 percentage points. Their empirical study,which is a structural VAR analysis on US data, shows a larger effect of 0.6 percentagepoints. Ramey (2012) also finds that higher public purchases leads to lower unemploy-ment, using structural and expectational VAR analyses on US data. In contrast, Brucknerand Pappa (2012), in an analysis of 10 OECD countries using structural VARs, find thathigher government expenditures increases the unemployment rate. Bruckner and Pappa(2012) explain the difference in results compared to Monacelli et al. (2010) as due todifferent sample period, arguing that the increased government spending raises unemploy-ment after 1975. Note however that Bruckner and Pappa (2012) also find that increasedgovernment spending leads to higher GDP and higher employment, so that the increasein unemployment is caused by higher participation rates due to increased labour supply.Pappa (2009) and Linnemann (2009) also find that increased public employment leads toincreased total employment.

An important methodological problem in an analysis of the effects of fiscal policy isthat the policy potentially is endogenous, by depending on the state of the economy.In the literature, this is typically handled either by focussing on the effect of specificevents that can be thought to be exogenous, such as changes in military spending ina response to political changes (e.g. Ramey and Shapiro (1999) and Eichenbaum andFisher (2005)), or by use of a structural vector autoregression (SVAR) model, where themodel explains several macroeconomic variables by their lags and exogenous shocks to thevariables in the model, see e.g. Blanchard and Perotti (2002), Fatas and Mihov (2001),Beetsma and Giuliodori (2010) and Monacelli et al. (2010). Both these methods have clearadvantages when it comes to dealing with endogenity. However, the methods also havetheir weaknesses, see Monacelli et al. (2010) and Auerbach et al. (2010). For the specificevents approach, it may be asked whether these events also affect the economy directly,implying that one does not detect the true effect of the policy. For the SVAR approach,there are concerns whether the fiscal shocks identified in the analysis might be anticipatedby the private sector in advance, which would cause problems in the interpretation. Morerecently, a number of studies analyze the effect of fiscal policy within structural modelsused for macro policymaking, typically of the Dynamic Stochastic General Equilibrium(DSGE) type, see e.g. Coenen et al. (2010); however, the findings in these studies hingeon whether the model is appropriate, which is a matter that is still open for discussion,cf. e.g. Chari et al. (2009) and Caballero (2010). This is a clear argument for also tryingother approaches.

We consider the effect of a change in government purchases on unemployment and em-ployment, building on a panel data estimation framework derived by Nymoen and Spar-rman (2012). More specifically, we add the real change in government purchases (scaledby the share of trend-GDP), to an empirical equation for aggregate unemployment as a

3

function of a number of labour market institutions. This approach has several advantages.First, an extensive literature has shown that aggregate unemployment to a large extent isdetermined by labour market institutions, see e.g. Layard et al. (1991) and Nickell et al.(2005). Thus, it seems appropriate to control for the effect of labour market variableswhen analysing the effect of fiscal policy and investigate if the effect of the fiscal policydepends on the prevailing labour market institutions. Second, with a data set covering 20countries and 27 years, there is large variation in a number of other key variables, makingit possible to explore how the effect of fiscal policy may vary depending on for instancethe monetary regime, the cyclical state of the economy, or the size of the public debt.

Our focus on the effect of government purchases (that is, government consumption andinvestment, which includes expenditure on public employment but not transfers) mitigatesthe endogeneity problem considerably. We do not consider tax revenues and expenditureson transfers, which clearly are endogenous, following changes in the economy accordingto rules and legislation. We also exclude all “passive” unemployment expenditure like un-employment benefits, and the large majority of active unemployment related expenditure,which are classified as transfers, not government purchases. In contrast, government pur-chases are not directly linked to the state of the economy. Clearly, the state of the economyalso affects purchase decisions, but also other factors come into play, like electoral cycles,party politics, lobbyism and pressure groups, media attention, etc. Furthermore, a largepart of government purchases may be subject to a lengthy bureaucratic process involvingboth the decision making and the implementation, implying that there is no clear cut orsimple relationship between the state of the economy and government purchases.

Even if we believe that the endogeneity problem is less important than for other typesof fiscal policy, we nevertheless undertake two different analyses to handle it. First, we useinstrumental variable estimation, where we treat the measure of fiscal policy as endoge-nous, and instrument with past values of the change in government purchases and pastvalues of government debt. Second, we use an omitted variables approach. The idea hereis that fiscal policy might be correlated with the error term because it is affected by otherexplanatory variables that also affect unemployment. By including the omitted variables,the potential bias will be reduced or removed, cf. discussion below.

From a theoretical perspective, one would expect the effect of an increase in governmentspending to depend on whether it is tax-financed or debt-financed. However, to distinguishbetween these two alternatives one must be able to differentiate between tax changesinduced by changes in the economy and tax changes linked to the financing of publicexpenditure. Given the identification problems that are involved, we have chosen not todo this at this stage. Thus, our results must be interpreted as an average effect, wherethe weights depend on the average method of financing over the sample period. In mostcountries both taxes and government debt have increased somewhat relative to GDP overthe sample period, suggesting that the increase in government spending is partly tax-financed and partly debt-financed.

3 Empirical Specification

We want to find the effect of a change in government purchases on the rate of unemploy-ment. However, the size of the public sector varies considerably relative to GDP, bothacross countries and over time within a single country, thus we multiply the real changein the government purchases with government purchases as a share of trend of GDP, seeappendix A for details and calculations. Clearly, if government purchases increase by 5percent, it matters whether public purchases constitute 16 percent of GDP, as in Spain in1980, or 33 percent of GDP, as in Sweden in 1982. An alternative would be to consider thechange in government purchases as a share of GDP, as is done in some studies (e.g. Alesina

4

and Ardagna (2009), Duell et al. (2009)). Yet this specification implies that a reductionin GDP caused by a negative external shock, increases the ratio of government purchasesto GDP, even if government purchases is kept constant. Thus, one might erroneouslyconclude that government purchases have a negative effect on GDP. For this reason wealso use a backward-looking measure of trend GDP, where the trend real growth is mea-sured as the moving average of the growth rate over the past ten years. With a two-sidedmeasure of trend-growth, there would be a risk that the future evolution of GDP affectsthe estimated trend-GDP, implying a possibility that the future evolution of GDP affectsthe measure of contemporaneous fiscal policy.

To improve the precision of our estimates, we want to control for other variablesthat may affect unemployment. First, we include an indicator for the export market,captured by the cyclical state of the economy of the trading partners. More specifically,the indicator is calculated as a weighted average of the GDP-gap of the trading partners,where the GDP-gap is the deviation of GDP from Hodrick Prescott-trend, divided by thetrend, and the weights reflect the share of the exports from country i that goes to each ofthe trading partners j. This variable will to a large extent capture the effect of commonshocks related to the world business cycle. Second, as argued above, we include labourmarket institutions. More specifically, we include the change in government purchasesand the export market in the unemployment equation derived in Nymoen and Sparrman(2012), which in accordance with Layard et al. (1991) and Nickell et al. (2005), is a functionof labour market institutions and shocks. Thus, in our main estimations, we estimate anequation of the following form

uit =β0i + β1uit−1 + β2uit−2 + β3uit−3 + β4∆Iit−1 + β5Iit−2

+ β6dGit + β7∆XMit + β8XMit−1 + β8∆XMit−1 + εit (1)

where uit is the unemployment rate in country i in period t, dGit is the real percentagechange in government purchases, multiplied by the ratio of government purchases to trendGDP, XMit is the export market indicator, and Iit−1 is a vector of institutional labourmarket variables which includes unemployment benefits, employment protection legisla-tion, measures of coordination and centralization of wage setting. The dynamic structurefollows from the theoretical labour market framework of Nymoen and Sparrman (2012),where the institutional variables affect the wage and price setting, which subsequentlyshape the evolution of unemployment. The specification in equation (1), with the levelof unemployment and the change in government purchases, reflects how we would ex-pect these two variables to behave in growing economies: Government purchases increaseover time, while unemployment and the change in government purchases are essentiallystationary variables. This presumption is consistent with results from stationarity tests.In table 1, column 1, we report results from Dickey-Fuller tests for a unit root in thelevel of government purchases, considering both homogeneous and heterogeneous autore-gressive parameters across countries, and with several different specifications concerninglags and subtraction of cross-sectional means prior to undertaking the tests, see Matyasand Sevestre (2008). Non-stationarity is not rejected in any of the tests. Correspondingtests reject non-stationarity of the level of unemployment and the change in governmentpurchases (except for the homogeneous alternative with three lags), indicating that thesevariables are stationary, cf table 1, columns 2 and 3.1

In most of the analysis, we use a Fixed Effects (FE) estimator, allowing for permanentcountry-specific differences in unemployment that are not accounted for by the other

1For the growth in government purchases, the test of unit root with lag 3 is not significant. However,as the third lag of the growth in government purchases is not significant, this test is less important.

5

Table 1: Unit root test for government purchases, growth in government purchases andunemployment

(Git −Gt)a (dGit − dGt)

b (Uit − U t)c

Homogeneous Heterogenous Homogeneous Heterogenous Homogeneous HeterogenousH1a H1b H1a H1b H1a H1b

lag 1d 8.86 (1.00) 9.54 (1.00) -6.54 (0.00) -7.69 (0.00) -4.19 (0.00) -3.29 (0.00)lag 2d 9.69 (1.00) 8.85 (1.00) -2.22 (0.01) -4.62 (0.00) -2.43 (0.00) -2.20 (0.01)lag 3d 5.71 (1.00) 3.66 (1.00) -0.25 (0.40) -3.09 (0.00) -3.37 (0.00) -2.12 (0.01)

Obs 560 560 560 560 560 560

In these tests, the variables for Germany are prolonged with data for West-Germany before the unionization in 1991 to achieve a balanced panel.

a) Change govt. purchases (Gt) subtracted cross-sectional means (Gt). Trend and country specific constant terms

are included in all the tests.

b) Change growth in govt. purchases (dGt) subtracted cross-sectional means (dGt).

c) Change unemployment (Ut) subtracted cross-sectional means (U t).

d) Numbers in parentheses are p-values for the relevant null.

explanatory variables. A random effect model would require that there is no correlationbetween the country fixed effects and the explanatory variables in the model. However,this assumption is rejected in a Hausman test with a p-value of 1 percent. In principle,the FE estimator is biased when the regression includes a lagged endogenous variable, seeNickell (1981). However, with a long time dimension of 27 years, this bias is small, cf.Judson and Owen (1999). In addition, other estimations methods which avoid the samplebias also have their difficulties, cf. Roodman (2009).

The model is estimated on annual frequency, using data from OECD Economic Out-look, see data appendix. Annual data has the advantage of a allowing a much longertime span (we also undertake some estimations including the 1960s and 70s, and very fewcountries have quarterly data for the fiscal policy for this period). Furthermore, annualdata may capture the actual fiscal decisions better, as the fiscal impulses are likely tofollow annual budgets, as well as mitigating possible anticipation effects, see discussion inBeetsma and Giuliodori (2010).

The growth in government purchases has generally been positive in real terms in thesample period, and the unweighted decade averages in our sample vary in the interval0.52 − 0.74. There is however considerable variation within and across countries, seeappendix table A1.

4 Empirical results

We find that the change in government purchases has a highly significant negative impacton unemployment, the point estimate implying that an increase in government purchasesequal to one percent of GDP reduces unemployment by 0.27 percentage points, or 0.28 withtime dummies, cf table 2. (To focus on the novel variables, the coefficients for the labourmarket institutions are omitted; the complete results are found in the appendix, table B1).The export market variables also have a significant negative effect on unemployment. Notethat even if the level effect of the export market variable is positive, this variable is basedon GDP-gaps and thus by construction only positive in a limited number of years in arow. Simulation exercises show that the overall effect of the export market is dominated bythe strongly negative impact effect throughout the estimation period (results are availableon request). The effect of the export market variables is smaller with year dummies,and the level effect insignificant, suggesting that the export market variables without timedummies also capture the effect of common shocks that affect most or all OECD countries.Figure 1 shows the estimated residuals of the fixed effects model without time dummiesin table 2, and there is little indication of autocorrelation, even if there is some variation

6

Table 2: Estimation of equation (1) - Fixed Effects

FE FE with time dummiesCoef. Std p-value Coef. Std p-value

Unemployment previous period 1.34 0.08 0.00 1.37 0.07 0.00Unemployment two years ago -0.57 0.10 0.00 -0.61 0.10 0.00Unemployment three years ago 0.04 0.05 0.43 0.07 0.05 0.15Demand components:Export market, 1st diff. (∆XMt) -0.48 0.10 0.00 -0.39 0.14 0.01Export market, prev. period (XMt−1) 0.17 0.08 0.04 0.05 0.13 0.69Export market, 1st diff. prev. period (∆XMt−1) -0.18 0.08 0.03 -0.07 0.12 0.56Change govt. purchases (dGt) -0.27 0.08 0.00 -0.28 0.08 0.00

Obs = Country*Average groups 483 20 24.1 483 20 24.1Standard deviation of residuals 0.63 0.59χ2 of policy and exports.a 107.61 (0.00) 68.73 (0.00)1st order autocorrelationa 1.50 (0.13) 1.34 (0.18)2nd order autocorrelationa -1.74 (0.08) -1.74 (0.08)

Dependent variable: The unemployment rate.

Estimation method: Fixed effect coeffcients estimate with huber-robust standard error (Stata command xtreg

with cluster(code) is used in all the regressions. It is also controlled for labour market institutions.

a) Numbers in parentheses are p-values for the relevant null.

across countries. The tests reported in the lower part of table 2 do not indicate any firstand second autocorrelation.

Figure 2 shows the dynamic effect on unemployment from a permanent increase ingovernment purchases equal to one percent of trend GDP. The maximum impact of −0.34percentage points is reached in the second year, then the effect weakens gradually to bealmost negligible after 10 years. This effect is somewhat larger than the findings in IMF(2010), based on a study of fiscal consolidations in 15 OECD countries over the last 30years. They find that spending-based deficit cuts equal to one percent of GDP raise theunemployment rate of about 0.2 percentage points. Monacelli et al. (2010) find a largereffect on US data; an increase in government spending equal to one percent of GDP leadsto a fall in the rate of unemployment of 0.6 percentage points after ten quarters. Incontrast, Bruckner and Pappa (2012) find in an analysis of 10 OECD countries usingstructural VARs, that a typical estimate from the impulse responses implies that a 10percent increase in government expenditures increases the unemployment rate at peak(which varies from 3− 16 quarters) of around 0.2− 0.5 percent.

4.1 Endogeneity

As noted above, the estimated coefficient of government purchases will be biased if gov-ernment purchases also react to changes in the state of the economy that are correlatedwith the rate of unemployment. One way to deal with this problem is to find instrumentsthat are uncorrelated with the error term and correlated with the change in governmentpurchases. We use the lagged first difference of the change in government purchases, aswell as the lagged ratio of public debt to GDP.

As shown in Table 3, there is a fairly strong and significant correlation between thechange in government purchases and the instrumental variables. The F-test statistic foradditional instruments is equal to 24, with a p-value of 0. Note that while our results inTable 2 imply that the lagged change in government purchases is correlated with laggedunemployment, the fact that lagged unemployment is also included in the equation impliesthat lagged government purchases may well be a valid instrument. We have also triedelection year, based on the idea that governments may pursue an expansionary fiscalpolicy in connection with elections to increase the probability of reelection; see evidence

7

-2-1

01

2

1980 1990 2000 2007

AUS: Australia

Residuals FE-model robust std. errors Residuals FE-model robust std. errors and time dummies

Time

-1-.

50

.5

1980 1990 2000 2007

AUT: Austria


Time

-1-.

50

.51

1980 1990 2000 2007

BEL: Belgium


Time

-10

1

1980 1990 2000 2007

CAN: Canada


Time

-2-1

01

1980 1990 2000 2007

DNK: Denmark


Time

-2-1

01

2

1980 1990 2000 2007

FIN: Finland


Time

-1-.

50

.51

1980 1990 2000 2007

FRA: France


Time

-.5

0.5

1

1980 1990 2000 2007

DEU: Germany


Time

-2-1

01

2

1980 1990 2000 2007

IRL: Ireland


Time

-1-.

50

.51

1980 1990 2000 2007

ITA: Italy


Time

-1-.

50

.51

1980 1990 2000 2007

JPN: Japan


Time

-10

12

1980 1990 2000 2007

NLD: Netherlands


Time

-1-.

50

.51

1980 1990 2000 2007

NOR: Norway


Time

-2-1

01

1980 1990 2000 2007

NZL: New Zealand


Time

-1-.

50

.51

1980 1990 2000 2007

PRT: Portugal


Time

-20

24

1980 1990 2000 2007

ESP: Spain


Time

-10

12

1980 1990 2000 2007

SWE: Sweden


Time

-1-.

50

.51

1980 1990 2000 2007

CHE: Switzerland


Time

-10

12

1980 1990 2000 2007

GBR: United Kingdom


Time

-10

12

1980 1990 2000 2007

USA: United States


Time

Figure 1: Estimated residuals of fixed effect model in table 2

8

-.4

-.3

-.2

-.1

0

2007 2009 2011 2013 2015 2017 2019 2021 2023 2025 2027 2029year

Dynamic simulation of U, fixed effects

Simulated unemployment rate with a one percent perm. increase in govt. purchase

Figure 2: The effect of a permanent increase in government purchases, equal to one percentof GDP, from 2008, based on simulation of equation (1) with estimated coefficients fromthe FE model in table 2

in Shi and Svensson (2006). However, including election year did not affect the result, andas election year is potentially endogenous in countries where the government can choosethe time of the election, we decided to leave it out in the presented specification.

Column 1 in Table 4 shows the results of the instrumental variable estimation. Thepoint estimate indicates that an increase in government purchases equal to one percent ofGDP reduces unemployment by 0.31 percentage points, i.e. close to the effect from the FEestimates. The effect is also highly statistically significant. Thus, there is no indicationthat government purchases is endogenous. The Hausman test statistics is equal to 5.88which does not reject that the residuals are uncorrelated with the error term. The HansenJ overidentification test has a p-value of 0.31, giving no indication of invalid instruments.Note however that these tests have limited power, and the Hansen J-test assumes thatat least one instrument is valid. Thus, we choose to present both FE and IV estimationsbelow.

For robustness, we also try another method, where we control for omitted variables.The idea behind this method is that fiscal policy might be correlated with the error termbecause it is affected by other explanatory variables that are excluded from the regressionand that also are correlated with unemployment. This would be the case if fiscal policy iseither pro- or counter-cyclical. Fiscal policy might be pro-cyclical if higher tax revenues ina boom cause politicians to spend more money; this effect is termed the voracity effect byTornell and Lane (1999). At the same time, the increase in tax revenues during the boommight be correlated with a fall in unemployment. In this case, including tax revenues asa regressor in the unemployment equation would lend government purchases uncorrelatedwith the error term, removing the bias in the coefficient. As the government purchasesare typically decided in the budget process in the fall the year prior to the budget year, itwould be the expectations that prevail when the budget is decided that might affect thebudget. We use predicted tax revenues as a proxy for expected tax revenues, where theprediction is based on a regression with two lags of tax revenues, two lags of the change inreal GDP as well as the output gap, as explanatory variables. Thus, in table 4 we include

9

Table 3: First stage estimation

First Stage regressionCoef. Std p-value

Unemployment previous period -0.11 0.04 0.01Unemployment two years ago 0.10 0.07 0.12Unemployment three years ago 0.02 0.04 0.51Demand components:Export market, 1st diff. (∆XMt) -0.04 0.10 0.67Export market, prev. period (XMt−1) 0.18 0.07 0.01Export market, 1st diff. prev. period (∆XMt−1) -0.08 0.10 0.42Instruments:Change govt. spending, 1st diff. prev. period (∆dGt−1) 0.13 0.06 0.04Debt previous period -0.01 0.00 0.00

Observations 440F-test of additional instruments, p-value in parenthesis 24 (0.00)

Dependent variable: The change in goverment spending. It is also controlled for labour market institutions.

First stage IV-regression of change in govt. spending (Stata command xtivreg2, see Baum et al. (2002))

the predicted change in tax revenues as a share of trend GDP to capture that higherrevenues might lead to increased government purchases. Alternatively, fiscal policy mightbe countercyclical if the government attempts to use fiscal policy to stabilize the economy.In this case one would expect an increase in government purchases in downturns, whenGDP growth is low, or the output gap is negative. To control for this, we also includeGDP growth and the change in the output gap, both lagged, in table 4.

We observe that the estimated effect of government purchases is somewhat smallerwhen we include the additional explanatory variables in model 2 in table 4, but it is stillstatistically significant. This lends considerable support to the robustness of this effect, asboth the lagged GDP growth and the output gap are variables that are strongly correlatedwith unemployment. Note however that some of the export market variables are no longersignificant. This emphasizes that including lagged GDP growth and lagged output gapentails a strong test of the explanatory power of the variables.

In model 3 in table 4, we control for the possible endogeneity of government purchasesin a somewhat different way, by also including consensus forecast for GDP growth, un-employment and the output gap. (In this regression, the sample is reduced because theforecasts are only available from 1997.) Again, one might conjecture that governmentpurchases would respond to such forecasts, and that the correlation we find between gov-ernment purchases and unemployment is due to both variables being correlated with theforecasts. However, we see that the change in government purchases has a significantnegative impact on unemployment, even when controlling for forecasts.

10

Tab

le4:

Equ

atio

n(1

):IV

and

FE

wit

hco

ntr

olfo

rom

itte

dva

riab

les

IVa

Model

2b

Model

3b

Coef

.S

tdp-v

alue

Coef

.Std

p-v

alue

Coef

.Std

p-v

alu

e

Changegovt.

purchases,

(dG

t)-0

.32

0.1

50.

04-0

.20

0.08

0.01

-0.2

00.

100.

05U

nem

plo

ym

ent

pre

vio

us

per

iod

1.37

0.07

0.0

01.

160.

090.

001.

060.

100.0

0U

nem

plo

ym

ent

two

year

sag

o-0

.62

0.09

0.00

-0.3

80.

110.0

0-0

.36

0.1

20.0

0U

nem

plo

ym

ent

thre

eye

ars

ago

0.06

0.04

0.1

50.

030.

050.

470.

090.

060.1

5Export

mark

et:

Exp

ort

mar

ket,

1st

diff

.(∆XM

t)-0

.45

0.08

0.0

0-0

.50

0.10

0.0

0-0

.40

0.1

40.

00E

xp

ort

mar

ket,

pre

v.

per

iod

(XM

t−1)

0.22

0.09

0.01

0.10

0.08

0.25

-0.0

20.

180.9

0E

xp

ort

mar

ket,

1st

diff

.pre

v.

per

iod

(∆XM

t−1)

-0.2

00.

100.

04-0

.05

0.12

0.6

50.

050.

140.

72Controls:

Log

GD

P1s

td

iff.

pre

v.

per

iod

-25.

756.7

40.

00

-22.

9113.8

00.1

0O

utp

utg

ap1s

td

iff.

pre

v.

per

iod

0.13

0.09

0.17

0.03

0.1

30.

81P

redic

ted

Dir

ect

and

indir

ect

taxes

div

ided

by

tren

dG

DP

,1s

tdiff

.0.0

10.

030.7

90.

050.1

60.7

6P

redic

ted

Dir

ect

and

indir

ect

taxes

div

ided

by

tren

dG

DP

,1s

tdiff

.p

rev.

per

iod

-0.0

20.0

70.

810.1

10.1

70.5

3Y

ear t−1

fore

cast

ofG

DP

grow

thyea

rt

0.01

0.08

0.8

8Y

ear t−1

fore

cast

ofou

tput

gap

yeart

-0.0

40.1

00.6

7Y

ear t−1

fore

cast

ofu

nem

plo

ym

ent

yea

rt

-0.1

40.0

70.0

4Y

ear t−1

fore

cast

ofG

DP

grow

thyea

rt

+1

0.00

0.10

0.9

8Y

ear t−1

fore

cast

ofou

tput

gap

yeart

+1

0.10

0.1

20.

40Y

ear t−1

fore

cast

ofu

nem

plo

ym

ent

yea

rt

+1

0.14

0.0

70.0

4

Obs

=C

ountr

y*A

vera

ge

grou

ps

440

2022.

046

420

23.2

207

20

10.

4Sta

ndar

ddev

iati

onof

resi

du

als

0.58

0.60

0.39

F-t

est

ofad

dit

ional

inst

rum

ents

.24

(0.0

0)χ2

ofp

olic

yan

dex

port

s.c

69.

45(0

.00)

25.

74(0

.00)

1st

order

auto

corr

elat

ionc

2.13

(0.0

3)

-2.9

3(0

.00)

2nd

order

auto

corr

elat

ionc

-1.4

7(0

.14)

1.31

(0.1

9)

Han

sen

JO

veri

den

tifi

cati

onte

stc

1.02

(0.3

1)

Dep

enden

tva

riable

:T

he

unem

plo

ym

ent

rate

.In

all

equati

ons

itis

als

oco

ntr

olled

for

lab

our

mark

etin

stit

uti

ons.

a)

Sta

taco

mm

and

xti

vre

g2,

see

Baum

etal.

(2002).

Change

gov

t.purc

hase

s(dG

t)

istr

eate

das

endogen

ous.

Inst

rum

ents

are

:∆dG

t−1

anddebt t

−1.

b)

Fix

edeff

ect

coeffi

cien

tses

tim

ate

.H

ub

er-r

obust

standard

erro

r(Sta

taco

mm

and

xtr

egw

ith

clust

er(c

ode)

.

c)N

um

ber

sin

pare

nth

esis

are

p-v

alu

esfo

rth

ere

leva

nt

null.

11

4.2 Cyclical effects and variation over time

An important question from a policy perspective is whether the effect of governmentpurchases varies over the business cycle. We measure the cyclical situation of the economyby use of the output gap as measured by the OECD.

In the first column in table 5, we extend equation (1) by including the interactionbetween the output gap and the change in government purchases, the latter measured asdeviation from the country-specific mean, to ensure that the variable has a mean of zero.The interaction term is strongly significant, with positive sign, implying that an increase ingovernment purchases leads to a larger reduction in unemployment in bad times when theoutput gap is negative than in good times. The estimated coefficient is equal to 0.05, whichimplies that if the output gap is negative and equal to −3 percentage point, an increasein government purchases equal to one percent of GDP will decrease unemployment by−0.17 + (−3)∗0.05 = −0.32 percentage points at impact. Note that we control for outputgap in the equation, which may reduce the estimated effect somewhat. The result isconsistent with several recent studies: Auerbach and Gorodnichenko (2012) find a largerfiscal multiplier in recessions than in expansions, using a regime-switching model on U.S.aggregate data; Nakamura and Steinsson (2012) find the same result exploiting differencesacross U.S. regions, while the IMF Fiscal Monitor 2012 reports consistent evidence forsome G7- countries. For theoretical motivation, see the discussion of Michaillat (2012) insection 4.3 below.

The systematic link between the output gap and the effect of government purchaseshas potentially vast policy interest, in particular for countries where fiscal tightening isrequired. The results in table 5 show that it matters when the fiscal tightening takes place,as the same fiscal tightening has a stronger effect on unemployment in a downturn of theeconomy. Ceteris paribus, this suggests that fiscal tightening should be postponed until theeconomy is in better shape. Taken at face value, the result suggests that countercyclicalfiscal policy may reduce the average unemployment rate over time: if the governmentpursued a countercyclical policy that were neutral over the cycle, where the increase ingovernment purchases in downturns were matched by the decrease in booms, the net effectwould, according to table 5, be a decrease in unemployment.

We then consider whether the effect of government purchases on unemployment variesover time, by allowing for a different effect for each decade, cf. model 2 in table 5: In the1960s and 70s, we find essentially no effect of government purchases on unemployment, insharp contrast to the later period. As the relationship clearly has changed significantlyafter 1980, it would be misleading to put so different decades together by imposing thesame relationship. This is the motivation for restricting attention to the time after 1980in the other regressions.

One might speculate that the absence of any effect in the 1960s reflects that unemploy-ment in almost all countries was very stable and low, not giving much room for an effectof government purchases. In contrast, in the 1970s, unemployment rose quite sharply inmost countries, and some countries tried to counteract this rise by use of expansionaryfiscal policy. Thus, there could be a downward bias in the estimate reflecting that therise in unemployment induced increased government spending, suggesting the use of IV.However, the IV results in the third model in table 5 are rather consistent with the FEresults; no effect of government purchases in the 1960s and 70s, and a negative effect forthe last three decades, although statistically significant only in the 2000s.

12

Tab

le5:

Th

eeff

ect

ofth

ecy

clic

alsi

tuat

ion

and

chan

geov

erti

me

Model

1a

Model

2aIV

b

Coef

.Std

p-v

alu

eC

oef

.S

tdp-v

alue

Coef

.Std

p-v

alu

e

Unem

plo

ym

ent

pre

vio

us

per

iod

1.16

0.09

0.00

1.2

40.

14

0.00

1.2

80.

060.0

0U

nem

plo

ym

ent

two

year

sag

o-0

.45

0.10

0.0

0-0

.38

0.23

0.1

0-0

.46

0.09

0.0

0U

nem

plo

ym

ent

thre

eye

ars

ago

0.06

0.05

0.2

0-0

.02

0.10

0.86

-0.0

10.0

50.

82Demand

components:

Exp

ort

mar

ket,

1st

diff

.(∆XM

t)-0

.28

0.07

0.0

0-0

.48

0.08

0.00

-0.4

60.0

80.

00E

xp

ort

mar

ket,

pre

v.

per

iod

(XM

t−1)

0.28

0.07

0.00

0.16

0.0

80.0

40.

170.0

90.

04E

xp

ort

mar

ket,

1st

diff

.pre

v.

per

iod

(∆XM

t−1)

-0.2

00.

070.0

1-0

.28

0.1

20.0

2-0

.17

0.09

0.0

6C

han

gego

vt.

pu

rch

ases

(dG

t)-0

.17

0.0

70.

02

Inte

ract

ion

chan

ge

govt.

purc

has

es(dG

t−dG

t)an

dou

tput

gapYt

0.05

0.02

0.0

1O

utp

utg

apfr

omO

EC

D(Y

t)-0

.14

0.03

0.00

Chan

gego

vt.

purc

has

es(dG

t),

1960

s0.0

40.

07

0.51

0.1

20.

100.2

4C

han

gego

vt.

purc

has

es(dG

t),

1970

s0.

040.

05

0.4

00.

280.3

00.3

6C

han

gego

vt.

purc

has

es(dG

t),

1980

s-0

.33

0.15

0.03

-0.4

20.3

20.

20C

han

gego

vt.

pu

rch

ases

(dG

t),

1990

s-0

.45

0.18

0.01

-0.3

80.2

60.1

5C

han

gego

vt.

pu

rch

ases

(dG

t),

2000

s-0

.28

0.09

0.00

-0.8

10.3

50.0

2D

um

my

for

1970

s0.

250.

15

0.1

00.

230.3

00.

45D

um

my

for

1980

s0.

770.

15

0.0

01.

130.2

40.

00D

um

my

for

1990

s0.

890.

21

0.0

01.

220.2

30.

00D

um

my

for

2000

s0.

650.

20

0.0

01.

250.2

80.

00

Obs

=C

ountr

y*A

vera

ge

grou

ps

483

20

24.

180

120

40.

0626

20

31.

3Sta

ndar

ddev

iati

onof

resi

du

als

0.58

0.65

0.63

F-t

est

ofad

dit

ional

inst

rum

ents

,19

60s.cd

4(0

.00)

F-t

est

ofad

dit

ional

inst

rum

ents

,19

70s.cd

2(0

.09)

F-t

est

ofad

dit

ional

inst

rum

ents

,19

80s.cd

2(0

.07)

F-t

est

ofad

dit

ional

inst

rum

ents

,19

90s.cd

6(0

.00)

F-t

est

ofad

dit

ional

inst

rum

ents

,20

00s.cd

3(0

.00)

χ2

ofp

olic

yan

dex

por

ts.d

101

.79

(0.0

0)16

2.75

(0.0

0)

97.5

0(0

.00)

1st

order

auto

corr

elat

iond

2.81

(0.0

1)1.

15(0

.25)

2nd

order

auto

corr

elati

ond

1.57

(0.1

2)0.

44(0

.66)

Han

sen

JO

veri

den

tifi

cati

onte

std

15.

98(0

.10)

Dep

enden

tva

riable

:T

he

unem

plo

ym

ent

rate

.In

all

equati

ons

itis

als

oco

ntr

olled

for

lab

our

mark

etin

stit

uti

ons.

a)

Fix

edeff

ect

coeffi

cien

tses

tim

ate

.H

ub

er-r

obust

standard

erro

r(Sta

taco

mm

and

xtr

egw

ith

clust

er(c

ode)

.

b)

Sta

taco

mm

and

xti

vre

g2,

see

Baum

etal.

(2002).

IV:

Change

gov

t.purc

hase

s(dG

t)

istr

eate

das

endogen

ous.

Inst

rum

ents

are

:(∆

dG

t−1),

(dG

t−2),

(∆debt t

−1)

and

(debt t

−1).

c)F

irst

regre

ssio

nre

sult

sof

stata

com

mand

xti

vre

g2,

see

Baum

etal.

(2002).

d)

Num

ber

sin

pare

nth

eses

are

p-v

alu

esfo

rth

ere

leva

nt

null.

13

4.3 Distinguishing between types of government purchases: investment,wage consumption and non-wage consumption

Both from a theoretical and policy perspective it is of considerable interest to explorewhether the effect of a change in government purchases differs depending of the type ofpurchase. In statistical sources, one typically distinguishes between three categories, whichwe also use in our analysis: government wage consumption, which is essentially expen-ditures on public employment, ie. employees in the public sector (dCGW ), governmentnon-wage consumption (dCGNW ) and government real investments (dIG). In our sam-ple, government wage consumption constitutes 54 percent of total government purchases,government non-wage consumption 29 percent, and government investments 17 percent(unweighted average across countries, from 1960-2007). We consider the same form of theleft hand side variable as before, i.e. the change in each of this categories (indicated by thed in the variable name), in real terms, and measured as share of trend-GDP, see appendixA for a detailed explanation.

The results are presented in table 6. Model 1 shows that increased government wageconsumption (increased public employment) has a significant negative impact on the un-employment rate, while the estimated effect of government investment is negative but notstatistically significant, and the effect of government non-wage consumption is negativebut small. Ramey (2012) also find a stronger effect on unemployment from governmentwage consumption - hiring workers - than from other parts of government purchases, whileIlzetzki et al. (2010) find about the same point estimates for the effect on GDP of con-sumption and investment for advanced countries, with multiplier estimates (the effect ofgovernment purchases on GDP) of 0.4 at impact and 0.8 in the long run. Ilzetzki et al.(2010) do not distinguish between wage and non-wage consumption.

In model 2 in table 6, we explore whether the effect of the different types of governmentpurchase depends on the cyclical state of the economy. We find a strong positive andstatistically significant interaction term for government wage consumption, implying thatincreased public employment has a stronger dampening effect on unemployment when theoutput gap is negative, consistent with our prior results. The effect is large: with an outputgap of minus 3, the coefficient is −0.55+(−3)∗0.13 = −0.94, implying that an increase ingovernment wage consumption equal to one percent of GDP reduces unemployment withalmost one percentage point at impact. The finding that an increase in government wageconsumption has a much stronger effect on unemployment in a recession is consistent withthe predictions from the dynamic stochastic general equilibrium search model of Michaillat(2012). In the search model of Michaillat (2012), increased public employment has a muchstronger effect on total employment in recessions, because there is much less crowding outof private employment. When unemployment is high, there is no shortage of unemployedworkers, implying that higher public employment has little impact on the hiring of privatefirms.

The third model in table 6 presents the result of model 1 using IV; we find that theeffect of government investment is much stronger, with a point estimate of −0.78, and ap-value of 0.06. The point estimate of wage consumption is about as in the FE estimation,−0.61, but not statistically significant.

14

Tab

le6:

Th

eeff

ect

ofd

iffer

ent

typ

esof

gove

rnm

ent

pu

rch

ases

Model

1a

Model

2a

IV3b

Coef

.Std

p-v

alue

Coef

.Std

p-v

alue

Coef

.Std

p-v

alue

Unem

plo

ym

ent

pre

vio

us

per

iod

1.27

0.07

0.0

01.

040.

080.

00

1.27

0.07

0.00

Unem

plo

ym

ent

two

year

sag

o-0

.51

0.10

0.0

0-0

.36

0.08

0.0

0-0

.53

0.1

00.0

0U

nem

plo

ym

ent

thre

eye

ars

ago

0.01

0.05

0.7

80.

050.

040.1

90.

020.

050.

66Demand

components:

Exp

ort

mar

ket,

1st

diff

.(∆XM

t)-0

.43

0.11

0.0

0-0

.24

0.0

70.

00-0

.42

0.0

90.0

0E

xp

ort

mar

ket,

pre

v.

per

iod

(XM

t−1)

0.17

0.12

0.15

0.2

80.

120.0

20.

250.1

00.0

1E

xp

ort

mar

ket,

1st

diff

.pre

v.

per

iod

(∆XM

t−1)

-0.2

30.

100.0

2-0

.28

0.08

0.0

0-0

.24

0.1

10.

03C

han

gego

vt.

inve

stm

ents

,(dIG

t)-0

.13

0.08

0.1

2-0

.07

0.08

0.43

-0.7

80.

420.0

6C

han

gego

vt.

non

-wag

eco

nsu

mpti

on,

(dCGNW

t)-0

.05

0.02

0.01

-0.0

30.

020.0

9-0

.10

0.1

90.6

2C

han

gego

vt.

wag

eco

nsu

mpti

on,

(dCGW

t)-0

.68

0.20

0.0

0-0

.55

0.15

0.0

0-0

.61

0.4

80.

20In

tera

ctio

ndGI t

andYt

-0.0

10.0

40.9

0In

tera

ctio

ndCGNW

tan

dYt

0.0

10.

010.3

5In

tera

ctio

ndCGW

tan

dYt

0.13

0.05

0.0

1O

utp

utg

apfr

omO

EC

D(Y

t)-0

.17

0.03

0.0

0

Obs

=C

ountr

y*A

vera

ge

grou

ps

375

1623

.4375

16

23.

4357

16

22.

3Sta

ndar

ddev

iati

onof

resi

du

als

0.5

90.

530.

60F

-tes

tof

addit

ional

inst

rum

ents

,go

vt.

inve

stm

ents

.cd

5.03

(0.0

0)F

-tes

tof

addit

ional

inst

rum

ents

,go

vt.

non

-wag

eco

nsu

mp

tion

.cd

1.87

(0.0

6)F

-tes

tof

addit

ional

inst

rum

ents

,go

vt.

wage

consu

mp

tion

.cd

3.13

(0.0

0)χ2

ofp

olic

yan

dex

por

ts.b

229

.65

(0.0

0)24

1.64

(0.0

0)1st

order

auto

corr

elat

ionb

1.73

(0.0

8)2.

82(0

.01)

2nd

order

auto

corr

elat

ionb

-1.7

0(0

.09)

0.82

(0.4

1)H

anse

nJ

Ove

riden

tifi

cati

onte

st9.

08(0

.11)

Dep

enden

tva

riable

:T

he

unem

plo

ym

ent

rate

.In

all

equati

ons

itis

als

oco

ntr

olled

for

lab

our

mark

etin

stit

uti

ons.

a)

Fix

edeff

ect

coeff

cien

tses

tim

ate

,ro

bust

standard

erro

rs(x

treg

wit

h,r

obust

)

b)

Sta

taco

mm

and

xti

vre

g2

wit

hopti

on

robust

,se

eB

aum

etal.

(2002).

Change

gov

t.purc

hase

s(dG

t)

istr

eate

das

endogen

ous.

Inst

rum

ents

are

:(∆

dG

t−1),

(dG

t−2)

and

(∆debt t

−1).

c)F

irst

regre

ssio

nre

sult

sof

stata

com

mand

xti

vre

g2,

see

Baum

etal.

(2002).

d)

Num

ber

sin

pare

nth

eses

are

p-v

alu

esfo

rth

ere

leva

nt

null.

15

4.4 Monetary regime, labour market institutions, debt and openness

In this subsection we first explore whether the effect of government purchases depends onthe monetary regime, as implied by standard text book macro models like the MundellFleming model, and also emphasized in much of the recent literature, e.g. Coenen et al.(2010). Under an inflation target, an expansionary effect of increased government pur-chases will be counteracted by a rise in the interest rate, partly offsetting the effect onunemployment. Also with other types of floating exchange rates, one would expect anexpansionary effect from fiscal policy to be counteracted by changes in the exchange rateand the interest rate. In contrast, if the nominal interest rate is unaffected, as it will bewith a fixed exchange rate and for a small country in a monetary union, and inflationincreases so that the real interest falls, the government multiplier might be considerablyabove unity. Note however that this effect depends on private sector expectations. Aspointed out by Nakamura and Steinsson (2012), if one imposes purchasing power parityin the long run, a short run increase in inflation will be compensated by lower inflation inlater periods, offsetting the effect on the long run real interest rate. Yet to what extentprivate agents in fact take such effects into account remains an empirical question.

We use three dummies to capture the different monetary regimes within the sampleperiod; fixed exchange rate regimes, floating exchange rate regimes (in recent years in-cluding inflation targeting), and membership in the European Monetary Union (EMU).Countries that took part in the European Exchange Rate Mechanism ERM are defined ashaving a fixed exchange rate regime, except for Germany, which we define as floating inlight of Germany’s dominating position and the independent status of the Bundesbank.We also tried to distinguish between credible and non-credible fixed exchange rate regimesdepending on the interest rate differential relative to the anchor country (in most casesGermany), defining the regime as non-credible if the interest rate differential exceeds 1percentage point in annual terms. The idea here is that if the fixed exchange rate lackscredibility, a fiscal expansion could have a further negative effect on credibility, thus rais-ing devaluation expectations and push up interest rates. However, the point estimateswere essentially the same for credible and non-credible fixed exchange rate regimes, so wedecided to drop this distinction in the results we report.

Model 1 in table 7 shows that the effect of government purchases differs sharply acrossmonetary regimes. The point estimate is −0.32 in the EMU and −0.43 with a fixedexchange rate regime, both statistically significant. In contrast, the point estimate issmaller and imprecisely determined with a floating exchange rate. The difference acrossregimes is in line with our theoretical expectations, where fiscal policy is effective undera fixed exchange rate regime, but not under float. The difference across exchange rateregimes is consistent with those of Ilzetzki et al. (2010); they find a significant positiveeffect of increased government consumption on GDP for fixed exchange rate regimes, whilethe effect is significant and negative at impact for floating regimes. Model 2 displays theIV results: here the coefficient in the EMU is statistically significant and negative, whilethe other coefficients are imprecisely determined and thus not statistically significant.

In model 3, we explore whether the interaction with the cyclical situation of the econ-omy depends on the monetary regime. We do this by including an interaction term betweenthe output gap, a dummy for monetary regime and the change in government purchases,measured as a deviation from the country-specific mean. The interaction terms are allpositive, consistent with our results above the fiscal policy has a stronger effect on unem-ployment during a downturn, but the coefficient is only significant for the fixed exchangerate.

16

Tab

le7:

Th

eeff

ect

ofm

onet

ary

regi

me

and

lab

our

mar

ket

inst

itu

tion

s

Model

1a

IVb

Model

3a

Model

4a

Model

5a

Coef

.Std

p-v

alu

eC

oef

.Std

p-v

alue

Coef

.Std

p-v

alu

eC

oef

.Std

p-v

alu

eC

oef

.Std

p-v

alu

e

Unem

plo

ym

ent

pre

vio

us

per

iod

1.32

0.07

0.00

1.37

0.0

60.0

01.

13

0.08

0.0

01.

32

0.08

0.0

01.

30

0.0

70.0

0U

nem

plo

ym

ent

two

year

sag

o-0

.56

0.10

0.00

-0.6

30.

090.0

0-0

.43

0.0

90.0

0-0

.53

0.1

00.

00

-0.5

30.1

00.0

0U

nem

plo

ym

ent

thre

eye

ars

ago

0.04

0.04

0.36

0.0

60.

040.1

90.0

60.

04

0.1

90.0

20.0

50.6

50.

02

0.05

0.6

8Export

mark

et:

Exp

ort

mar

ket,

1st

diff

.(∆XM

t)-0

.47

0.09

0.00

-0.4

30.0

80.0

0-0

.27

0.0

70.0

0-0

.48

0.10

0.00

-0.4

70.

10

0.0

0E

xp

ort

mar

ket,

pre

v.

per

iod

(XM

t−1)

0.15

0.08

0.06

0.19

0.09

0.0

30.

28

0.07

0.0

00.2

10.

08

0.0

10.1

80.

09

0.0

4E

xp

ort

mar

ket,

1st

diff

.pre

v.

per

iod

(∆XM

t−1)

-0.1

90.

090.

03-0

.19

0.10

0.0

6-0

.21

0.0

70.

00

-0.2

20.

09

0.0

2-0

.22

0.1

00.

02

Govt.

purchasesand

moneta

ryre

gim

e:

Chan

gego

vt.

purc

has

es(dG

t)M

onet

ary

unio

n(E

MU

)-0

.32

0.16

0.05

-0.4

00.

210.0

6-0

.55

0.2

40.

02

-0.4

50.

13

0.0

0C

han

gego

vt.

purc

has

es(dG

t),

Fix

edex

chan

gera

te-0

.43

0.13

0.00

-0.1

10.

250.6

7-0

.21

0.1

00.0

4-0

.31

0.11

0.01

Chan

gego

vt.

purc

has

es(dG

t),

Flo

atin

gex

chan

gera

te-0

.05

0.09

0.56

-0.2

50.

240.3

00.0

40.0

80.

62

-0.2

30.

08

0.0

1D

um

my

for

EM

U0.

010.

270.

97-0

.14

0.26

0.58

0.0

60.2

50.8

10.0

00.

21

1.00

Dum

my

for

Fix

edex

chan

gera

te0.

470.

270.0

90.

080.

340.

82

0.1

60.2

80.

56

0.3

30.2

10.

11

Inte

ract

ion

chan

gego

vt.

purc

has

es(dG

t)an

dou

tput

gapY

,M

onet

ary

unio

n(E

MU

)0.1

10.0

90.

21

Inte

ract

ion

chan

gego

vt.

purc

has

es(dG

t)an

dou

tput

gapY

,F

ixed

exch

ange

rate

,cr

edib

le0.0

60.0

20.

02

Inte

ract

ion

chan

gego

vt.

purc

has

es(dG

t)an

dou

tput

gapY

,F

loat

ing

exch

ange

rate

0.0

20.0

50.

65

Outp

utg

apfr

omO

EC

D(Y

t)-0

.16

0.03

0.0

0Govt.

purchasesand

labourmark

et:

Chan

gego

vt.

purc

has

es,

(dG

t)-0

.31

0.0

70.0

0In

tera

ctio

nch

ange

govt.

purc

has

esan

dth

epre

dic

ted

effec

tof

lab

our

mar

ket

inst

ituti

ons

-0.2

60.0

70.0

0-0

.27

0.0

60.0

0

Obs

=C

ountr

y*A

vera

gegr

oups

483

2024.

144

020

22.

0483

2024.1

483

20

24.

1483

2024

.1Sta

ndar

ddev

iati

onof

resi

dual

s0.

620.

580.5

70.6

20.6

1F

-tes

tof

addit

ional

inst

rum

ents

,E

MU

.cd

6.6

3(0

.00)

F-t

est

ofad

dit

ional

inst

rum

ents

,F

ixed

exch

ange

rate

.cd

4.84

(0.0

0)F

-tes

tof

addit

ional

inst

rum

ents

,F

loat

ing

exch

ange

rate

.cd

5.98

(0.0

0)χ2

ofp

olic

yan

dex

por

ts.d

135.

87(0

.00)

127

.24

(0.0

0)

103.7

9(0

.00)

138.

21

(0.0

0)

1st

order

auto

corr

elat

iond

1.26

(0.2

1)2.

73

(0.0

1)1.

51

(0.1

3)0.9

9(0

.32)

2nd

order

auto

corr

elat

iond

-1.5

9(0

.11)

1.1

2(0

.26)

-1.4

8(0

.14)

-1.7

1(0

.09)

Han

sen

Jov

erid

enti

fica

tion

test

8.10

(0.2

3)

Dep

enden

tva

riable

:T

he

unem

plo

ym

ent

rate

.In

all

equati

ons

itis

als

oco

ntr

olled

for

lab

our

mark

etin

stit

uti

ons.

a)

Fix

edeff

ect

coeff

cien

tses

tim

ate

,ro

bust

standard

erro

rs(x

treg

wit

hro

bust

).

Hub

er-r

obust

standard

erro

r(Sta

taco

mm

and

xtr

egw

ith

clust

er(c

ode)

.

b)

Sta

taco

mm

and

xti

vre

g2,

see

Baum

etal.

(2002).

Change

gov

t.purc

hase

s(dG

t)

istr

eate

das

endogen

ous.

Inst

rum

ents

,se

para

tely

for

each

regim

e,are

:(∆

dG

t−1),

(dG

t−2)

and

(debt t

−1).

xti

vre

g2

wit

hro

bust

c)F

irst

regre

ssio

nre

sult

sof

stata

com

mand

xti

vre

g2,

see

Baum

etal.

(2002).

d)

Num

ber

sin

pare

nth

esis

are

p-v

alu

esfo

rth

ere

leva

nt

null.

17

We then turn to the importance of labour market institutions. The theoretical resultsare mixed: Ardagna (2007) finds that an increase in government purchases leads to in-creased unemployment in a monopoly union model, while Furlanetto (2011) shows thatreal wage rigidity may play an important role to preserve the traditional effect of fiscalpolicy in New Keynesian models, i.e. that increased purchases leads to higher employment.

First, as noted above we include institutions in all the regressions. This turns out tobe of limited importance, as the effect of fiscal policy is only slightly larger in a regressionwithout the labour market institutions (results available on request). Second, we wantto explore whether the effect of fiscal policy varies across countries with different labourmarket institutions. To this end, we first construct a summary index on the basis ofour regression results. Specifically, we detect the estimated effect of the labour marketinstitutions in the appendix, model 3 in table B1, by calculating the index as the productof the estimated coefficients and the actual values of the labour market institutions. Wecompute the deviation of the index from its sample mean to obtain an index with zeromean. We then interact the change in government purchases with the index of labourmarket institutions. The results in Table 7, column 4, shows that an increase in governmentpurchases has a stronger negative impact on unemployment in country-years with labourmarket institutions that induce higher unemployment. The effect is highly significantstatistically, and numerically rather strong: In Australia, labour market institutions are”employment-friendly” with mean index value −0.77, and the effect on unemployment ofan increase in government purchases equal to one percent of GDP is equal to −0.31 +(−0.77) ∗ (−0.26) = −0.11, In contrast, in Sweden, institutions are more conducive tounemployment with mean index of 0.38, implying that the overall coefficient for an increasein government purchases is−0.31+0.38∗(−0.26) = −0.41. Thus, fiscal policy seems to havea stronger impact on unemployment in countries with adverse labour market institutions.One possible interpretation of this is that there is less crowding out of private employmentin countries with rigid labor markets.

One possible concern with these results is that may be caused by spurious correlation,as labour market institutions are generally more rigid in Continental Europe, where wealso find most of the EMU countries. Thus, in column 5, we also include the interactionof fiscal policy and monetary regime, and we find that the results still hold - fiscal policyhas a larger effect with adverse labour market institutions. In fact, the coefficient forgovernment purchases is now negative and statistically significant for all three monetaryregimes, suggesting that an increase in government purchases has a negative impact alsounder a floating regime, as long as the labour market institutions are at the average.

We have also tried other possible interaction effects. Giavazzi and Pagano (1990) arguethat a severe fiscal contraction might be expansionary in situations with concern for therisks of high public debt. This suggests that the effect of government purchases maydepend on the level of public debt. In a recent study using structural VARs on quarterlydata for 44 countries, both advanced and developing countries, Ilzetzki et al. (2010) findthat the fiscal multiplier depends on the level of government debt, and that the fiscalmultiplier is zero in high debt countries. To explore the possible importance of publicdebt, we interact the change in government purchases with lagged public debt as a ratioto GDP (measured as deviation from sample mean, which is equal to 0.65). We interactwith debt above and below the mean separately, to allow for non-linear effects, and wealso include debt as a separate explanatory variable, as the levels of debt might well becorrelated with the level of unemployment, cf. Bertola (2010). However, the interactionterms have no explanatory power, with coefficients of about 0.001 and fairly small standarderrors (0.006 and 0.004), implying that these effects are fairly precisely estimated to zero(results available on request).

We also explore whether the effect of government purchases depends on the opennessof the country. According to traditional Keynesian analysis, the government expenditure

18

multiplier is smaller in an open economy. In line with this, Beetsma and Giuliodori (2010)find in an analysis of 14 EU countries a clear positive effect of a rise in governmentpurchases on GDP in “closed economies” (defined as countries where the ratio of exportplus import to GDP is below sample average), and no significant effect in the remaining“open economies”. Ilzetzki et al. (2010) also find a stronger expansionary effect in closedeconomies than in open. To analyse the effect of openness, we interact the change ingovernment purchases with an indicator of openness, based on the ratio of export plusimport to GDP. As the degree of openness has increased over time, we consider twodifferent specifications of this indicator, one where the indicator measures the deviationof the export plus import ratio from the overall sample mean, implying that the indicatoralso captures the increase in openness over time, and one where the indicator is measuredas deviation from year mean, thus omitting the change in openness over time. However,in both cases the interaction term is close to zero when we control for monetary regime(results available on request).

4.5 The effect on the employment rate

In this section we explore whether our findings of a clear effect on the unemployment rate isreflected in a corresponding effect on the employment rate. Generally, we find effects whichare very similar as those reported above, with coefficients of the opposite sign. For example,when we estimate equation 1 with the employment rate as the dependent variable, thecoefficient for the change in government purchases is 0.23, with a p-value of 0.00, cf. table8. Models 2 and 3 show that it is essentially government wage consumption which has apositive effect on employment, and for this variable the effect is much stronger in recessions(when the output gap is negative), consistent with our unemployment findings above. Thecorrespondance between the effects on employment and on unemployment indicates thatvariation in government purchases usually has limited effect on labour supply. In table 9we repeat the regressions from table 7, but with employment rather than unemploymentas the dependent variable. We observe that in the regressions in column 1, the change ingovernment purchases is only significant for countries with a fixed exchange rate. The IVresults in column 2 are not significant. In column 3, which includes an interaction withthe output gap, government purchases has a significant effect in EMU countries, while theinteraction term is significant for both EMU and countries with a fixed exchange rate.For EMU countries, both point estimates are quite large, implying a substantial impactof fiscal policy in a downturn.

As above, we find that the effect of government purchases on the labour market isstronger in countries with institutions that are conducive to unemployment. Furthermore,when we control for labour market institutions, we find that an increase in governmentpurchases leads to higher employment also in countries with a floating exchange rate(Model 5 in table 9), even if the point estimate is still considerably lower than under afixed exchange rate.

19

Tab

le8:

Th

eeff

ect

onth

eem

plo

ym

ent

rate

Model

1M

od

el1

wit

hM

odel

2M

odel

3ti

me

dum

mie

sC

oef

.Std

p-v

alu

eC

oef

.S

tdp-v

alu

eC

oef

.Std

p-v

alu

eC

oef

.Std

p-v

alu

e

Em

plo

ym

ent

pre

vio

us

per

iod

1.54

0.09

0.00

1.53

0.09

0.00

1.56

0.0

90.0

01.

300.1

30.

00E

mplo

ym

ent

two

year

sag

o-0

.76

0.12

0.00

-0.7

40.

110.

00

-0.8

20.1

30.0

0-0

.64

0.12

0.0

0E

mplo

ym

ent

thre

eye

ars

ago

0.14

0.05

0.0

00.

130.

040.

000.

150.0

50.0

00.

19

0.0

40.

00Demand

components:

Exp

ort

mar

ket,

1st

diff

.(∆XM

t)0.

400.

090.

000.

380.

160.

020.

330.1

00.0

00.

160.

070.

03E

xp

ort

mar

ket,

pre

v.

per

iod

(XM

t−1)

-0.1

60.0

70.

01-0

.03

0.16

0.85

-0.1

60.

080.0

3-0

.29

0.0

90.

00E

xp

ort

mar

ket,

1st

diff

.pre

v.

per

iod

(∆XM

t−1)

0.14

0.07

0.06

-0.0

40.

150.7

80.

140.0

80.0

90.

18

0.0

60.0

0C

han

gego

vt.

pu

rch

ases

(dG

t)0.

230.

070.0

00.2

50.

070.0

0C

han

gego

vt.

inve

stm

ents

,(dIG

t)0.

100.

050.0

50.0

80.0

60.2

1C

han

gego

vt.

non

-wag

eco

nsu

mpti

on,

(dCGNW

t)0.

030.

010.

00-0

.00

0.0

20.9

0C

han

gego

vt.

wag

eco

nsu

mpti

on,

(dCGW

t)0.4

50.1

60.

010.3

50.

130.

01In

tera

ctio

ndGI t

andYt

0.00

0.0

40.

92In

tera

ctio

ndCGNW

tandYt

0.00

0.0

00.

91In

tera

ctio

ndCGW

tan

dYt

-0.1

20.0

20.

00O

utp

utg

apfr

omO

EC

D(Y

t)0.1

70.0

40.0

0

Obs

=C

ountr

y*A

vera

ge

grou

ps

483

20

24.1

483

2024

.137

516

23.4

375

1623.4

Sta

ndar

ddev

iati

onof

resi

du

als

0.66

0.63

0.60

0.5

4χ2

ofp

olic

yan

dex

port

s.a

39.6

8(0

.00)

23.8

0(0

.00)

1st

order

auto

corr

elat

iona

1.5

8(0

.11)

2.59

(0.0

1)

1.73

(0.0

8)

2.66

(0.0

1)2nd

order

auto

corr

elat

iona

-0.5

2(0

.60)

-0.8

5(0

.40)

-1.7

0(0

.09)

2.0

4(0

.04)

Dep

enden

tva

riable

:T

he

emplo

ym

ent

top

opula

tion

rate

.

Est

imati

on

met

hod:

Fix

edeff

ect

coeff

cien

tses

tim

ate

wit

hhub

er-r

obust

standard

erro

r(S

tata

com

mand

xtr

eg

wit

hcl

ust

er(c

ode)

and

robust

opti

on

isuse

din

all

the

regre

ssio

ns.

Itis

als

oco

ntr

olled

for

lab

our

mark

etin

stit

uti

ons.

a)

Num

ber

sin

pare

nth

eses

are

p-v

alu

esfo

rth

ere

leva

nt

null.

20

Tab

le9:

Th

eeff

ect

onth

eem

plo

ym

ent

rate

Model

1a

Model

2IV

bM

odel

3a

Model

4a

Model

5a

Coef

.Std

p-v

alu

eC

oef

.Std

p-v

alue

Coef

.Std

p-v

alu

eC

oef

.Std

p-v

alu

eC

oef

.Std

p-v

alu

e

Em

plo

ym

ent

pre

vio

us

per

iod

1.51

0.09

0.00

1.64

0.07

0.0

01.3

10.1

00.0

01.

54

0.0

90.

00

1.5

00.

09

0.00

Em

plo

ym

ent

two

year

sag

o-0

.75

0.12

0.00

-0.9

20.

120.0

0-0

.62

0.1

10.

00

-0.7

60.1

20.

00

-0.7

40.1

20.0

0E

mplo

ym

ent

thre

eye

ars

ago

0.14

0.04

0.00

0.1

90.0

60.

00

0.1

90.0

50.0

00.

14

0.0

50.

00

0.14

0.0

40.

00

Export

mark

et:

Exp

ort

mar

ket,

1st

diff

.(∆XM

t)0.

400.

090.

000.3

10.0

90.

00

0.2

10.0

70.

00

0.40

0.0

90.

00

0.4

00.

09

0.0

0E

xp

ort

mar

ket,

pre

v.

per

iod

(XM

t−1)

-0.1

10.

060.

06-0

.11

0.10

0.29

-0.2

30.0

60.0

0-0

.18

0.07

0.0

1-0

.12

0.06

0.0

5E

xp

ort

mar

ket,

1st

diff

.pre

v.

per

iod

(∆XM

t−1)

0.14

0.07

0.04

0.13

0.09

0.15

0.1

40.

05

0.01

0.1

50.0

70.

02

0.1

50.

07

0.0

3Govt.

purchasesand

moneta

ryre

gim

e:

Chan

gego

vt.

purc

has

es(dG

t)M

onet

ary

unio

n(E

MU

)0.

250.

190.

210.

010.

170.9

70.5

70.2

30.

01

0.3

10.

17

0.06

Chan

gego

vt.

purc

has

es(dG

t),

Fix

edex

chan

gera

te0.

350.

110.

00-0

.17

0.38

0.67

0.1

00.

13

0.4

50.

28

0.1

10.

01

Chan

gego

vt.

purc

has

es(dG

t),

Flo

atin

gex

chan

gera

te0.

060.

070.

38-0

.37

0.29

0.2

1-0

.04

0.0

60.

51

0.1

50.

08

0.04

Dum

my

for

EM

U0.

250.

250.3

20.

160.

300.

600.

30

0.25

0.23

0.2

60.

23

0.2

5D

um

my

for

Fix

edex

chan

gera

te-0

.37

0.23

0.1

0-0

.28

0.45

0.53

0.0

50.2

90.

85

-0.2

90.1

90.

12

Inte

ract

ion

chan

gego

vt.

purc

has

es(dG

t)an

dou

tput

gapY

,M

onet

ary

unio

n(E

MU

)-0

.17

0.09

0.0

5In

tera

ctio

nch

ange

govt.

purc

has

es(dG

t)an

dou

tput

gapY

,F

ixed

exch

ange

rate

,cr

edib

le-0

.06

0.0

30.

01

Inte

ract

ion

chan

gego

vt.

purc

has

es(dG

t)an

dou

tput

gapY

,F

loat

ing

exch

ange

rate

-0.0

30.

05

0.5

0O

utp

utg

apfr

omO

EC

D(Y

t)0.

17

0.03

0.0

0Govt.

purchasesand

labourmark

et:

Chan

gego

vt.

purc

has

es,

(dG

t)0.

25

0.0

70.

00

Inte

ract

ion

chan

gego

vt.

purc

has

esan

dth

epre

dic

ted

effec

tof

lab

our

mar

ket

inst

ituti

ons

0.15

0.0

70.

02

0.1

40.

05

0.01

Obs

=C

ountr

y*A

vera

gegr

oups

483

2024

.144

020

22.0

483

20

24.1

483

20

24.1

483

20

24.1

Sta

ndar

ddev

iati

onof

resi

dual

s0.

650.

600.5

90.6

50.6

4F

-tes

tof

addit

ional

inst

rum

ents

,E

MU

.cd

5.92

(0.0

0)F

-tes

tof

addit

ional

inst

rum

ents

,F

ixed

exch

ange

rate

.cd

4.17

(0.0

0)F

-tes

tof

addit

ional

inst

rum

ents

,F

loat

ing

exch

ange

rate

.cd

7.16

(0.0

0)χ2

ofp

olic

yan

dex

por

ts.b

57.7

7(0

.00)

170.3

1(0

.00)

56.9

4(0

.00)

66.9

0(0

.00)

1st

order

auto

corr

elat

ionb

1.50

(0.1

3)2.5

8(0

.01)

1.32

(0.1

9)1.2

8(0

.20)

2nd

order

auto

corr

elat

ionb

-0.4

7(0

.64)

2.34

(0.0

2)

0.49

(0.6

2)

-0.4

1(0

.68)

Han

sen

JO

veri

den

tifica

tion

test

2.92

(0.4

0)

Dep

enden

tva

riable

:T

he

emplo

ym

ent

top

opula

tion

rate

.In

all

equati

ons

itis

als

oco

ntr

olled

for

lab

our

mark

etin

stit

uti

ons.

a)

Fix

edeff

ect

coeff

cien

tses

tim

ate

,ro

bust

standard

erro

rs(x

treg

wit

hro

bust

).H

ub

er-r

obust

standard

erro

r(Sta

taco

mm

and

xtr

egw

ith

clust

er(c

ode)

.

b)

Sta

taco

mm

and

xti

vre

g2,

see

Baum

etal.

(2002).

Change

gov

t.purc

hase

s(dG

t)

istr

eate

das

endogen

ous.

Inst

rum

ents

,se

para

tely

for

each

regim

e,are

:(∆

dG

t−1)

and

(debt t

−1).

c)F

irst

regre

ssio

nre

sult

sof

stata

com

mand

xti

vre

g2,

see

Baum

etal.

(2002).

d)

Num

ber

sin

pare

nth

esis

are

p-v

alu

esfo

rth

ere

leva

nt

null.

21

5 Concluding remarks

The vast policy interest of fiscal policy issues has led to a fast increasing body of researchtrying to detect the effect of a change in fiscal policy. However, few of these studies explorethe effect on unemployment, and the results in the existing studies are mixed. We inves-tigate the effect of changes in government purchases on unemployment and employmentby use of panel data estimation, building on an empirical equation where long run unem-ployment is a function of a number of labour market variables, along the lines of Layardet al. (2005) and Nickell et al. (2005). Our study exploits a fairly large data set, covering20 countries and 27 years, making it possible to explore how the effect differ according tothe circumstances, like monetary regime, cyclical situation of the economy, and whetherthe labour market institutions are ”employment-friendly”. It turns out that the effectdiffers strongly with these circumstances, underscoring the importance of exploring thedifferences.

We find that a permanent increase in government purchases equal to one percent ofGDP on average leads to a reduction in unemployment of 0.3 percentage point, using bothfixed effects and IV estimation. There is considerable variation in the effect of governmentpurchases depending on the specific circumstances. Consistent with recent studies likeAuerbach and Gorodnichenko (2012) and Nakamura and Steinsson (2012), we find thatthe effect is considerably larger when the economy is in a weak cyclical situation.

The systematic link between the output gap and the effect of government purchaseshas potentially vast policy interest. In many countries, the increasing public debt impliesthat fiscal policy has to be tightened before long, so there is little scope for using fiscalpolicy to reduce unemployment. However, taken at face value the results in table 5 suggestthat even a fiscally neutral countercyclical policy may reduce the average unemploymentrate over time. According to these results, if the increase in government purchases indownturns were matched by a decrease of the same magnitude in booms, the net effectwould be a decrease in unemployment. Such an effect would be inconsistent with standardtheories of equilibrium unemployment, but it is in line with the countercyclical effect offiscal policy in the general equilibrium search model of Michaillat (2012). The persistenthigh unemployment rates in many OECD countries, indicating long-lasting deviationsfrom long run equilibrium levels, makes the result highly relevant for policy.

The monetary regime is important for the effect. In line with the Mundell Flemingmodel, we find a strong effect of government purchases on unemployment for countrieswithin a monetary union or with a fixed exchange rate regime, and much weaker effectsof government purchases for countries with a floating exchange rate. This finding is con-sistent with the argument of among others Coenen et al. (2010), that fiscal policy hasa strong impact on the economy when the monetary policy does not respond. Consid-ering different types of government purchases, we only find a strong significant effect ofgovernment wage consumption (i.e. public employment), while the IV estimates suggestthat government investment has a strong dampening effect on unemployment. The effectof government wage consumption is strongly countercyclical, consistent with the searchmodel of Michaillat (2012). We also find that the fiscal policy has a stronger effect incountries with labour market institutions that are conducive to high unemployment.

Finally, we explore the effect of a change in government purchases on the employmentrate. For the most part, the results correspond well to the unemployment results. In-creased government purchases equal to one percent of GDP is estimated to increase theemployment rate by 0.23 percentage points. The effect essentially comes for governmentwage consumption (i.e. public employment), it is stronger in a downturn, and also strongerin countries with a fixed exchange rate.

22

A Appendix: Data definitions and sources

The data are from OECD (2008b) unless otherwise noted. The sample period is from 1980to 2007, except for Table 5, which also includes 1960-1979, with the following countries:Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy,Japan, Netherlands, Norway, New Zealand, Portugal, Spain, Sweden, Switzerland, UnitedKingdom and United States (for Germany, the data starts in 1991). The labour marketdata is also based on the OECD (2008b), but the more detailed description is given inSparrman (2010).

A.1 Government purchases

The change in government purchases (dG) is measured as the growth rate in real termsof government purchases, multiplied with government purchases as a share of trend GDP.The formula of dG is:

dGit =(CGVit + IGVit)− (CGVit−1 − IGVit−1)

(CGVit−1 + IGVit−1)∗ CGit + IGit − CFKGit

Y CTit∗ 100 (A1)

where CG is government consumption, IG government investments, CFKG is consump-tion of fixed capital, and Y CT is trend-GDP. The variables are in nominal prices, exceptthose where the last letter V indicates real terms. Note that government purchases donot include transfers like social security expenditures etc. Note also that we subtractconsumption of fixed capital (CFKG) from government consumption to obtain the ac-tual expenditure, as the consumption of fiscal capital is an imputed measure. CFKG isnot subtracted in the real growth rate for reasons of data availability, but this is unim-portant as there presumably is little variation over time in the imputed consumption offixed capital. Investment data is missing for some countries (Spain, Italy, Switzerland)and for these countries we use government consumption only. Trend-GDP is equal to thebackward looking 10 year moving average of real GDP (Y Q) multiplied with the two yearmoving average of the price deflator (PGDP ) to a variable in nominal terms. Some ofthe volume variable series are calculated on the basis of the relevant identities with valuesand deflators, as they are not published by the OECD.

A.2 Monetary regime

We have constructed 3 dummies to account for changes in the monetary regime over thesample period; a floating exchange rate, a fixed exchange rate, and membership in theEuropean Monetary Union (EMU). The dummy Dfloat indicates a floating exchange rate:Australia, Canada, Germany, Japan, New Zealand, Switzerland, United States and UnitedKingdom (except 1990 and 1991 2). Germany is defined to have a floating regime in lightof Germany’s dominating position European Exchange Rate Mechanism ERM, and theindependent status of the Bundesbank. Later also Sweden (since 1992) and Norway (since1999, which is usually assumed to be the year with the de facto change of regime) adopteda floating exchange rate, with inflation targeting. The dummy Dfixed indicates a fixedexchange rate, and this includes the countries that took part in the ERM, except Germany.DEMU indicates EMU membership, covering Austria, Belgium, Finland, France, Germany,Ireland, Italy, Portugal and Spain since 1999.

A.3 Export market indicator

The export market (XM) indicator is calculated as a weighted average of the GDP-gap ofthe trading partners, where the GDP-gap is the deviation of GDP from Hodrick Prescott-

2The UK was a member of the European exchange rate mechanism (ERM) from October 1990 toSeptember 1992

23

trend (with smoothing parameter 100), and the weights reflect the share of the exportsfrom country i that goes to each of the trading partners j. The formulae is

XMit = Σjwijt ∗GAPjt (A2)

where wijt = xijt/Σjxijt. xijt is export from country i to county j in year t. The tradingpartners to one country in the sample are all the other countries in the sample and the restof ’the world’. The exports data is from SITC Revision 2 OECD (2010), and are used tocalculate the export shares for each country in the sample. The time series are prolongedbackwards with the exports to the world when observations are missing. The GDP-gapfor each of the twenty OECD countries is calculated using data from OECD (2008b).The world GDP-gap is constructed using data for the real GDP in The Conference Board(2010). We have used the GDPGK-series with GDP expressed in 1990 U.S. dollars, whichcovers 123 countries in the database.

A.4 Other variables

The unemployment rate - we use the standardized unemployment rate (UNR) fromEconomic Outlook OECD (2008a) .

Output gap - is defined as the actual GDP less potential GDP, as a share of potentialGDP. It is measured in percentage points and collected from OECD (2008b).

Election year - is collected from Armingeon et al. (2010), and the original data sourceis European Journal of Political Research (Political Data Yearbook, various issues); Mackieand Rose (1991); Keesing’s Archive; Parline database. The variable describes date ofelection of national parliament (lower house). The variable covers the years in the period1960 to 2008.

Gross Public Debt - is collected from Armingeon et al. (2010), and the original datasource is several versions of Oecd Economic outlook. See details regarding versions andthe mission observations in Codebook by Armingeon et al. (2010).

Openness - is total trade (export and imports ) in percentage of GDP. The variableis collected from Armingeon et al. (2010). See details regarding versions and the missionobservations in Codebook by Armingeon et al. (2010).

24

Table A1: Real growth in government purchases, multiplied by the ratio of governmentpurchases to trend GDP - country specific mean and standard deviation

Country stats 1960-69 1970-79 1980-89 1990-99 2000-07 1960-07

Australia mean 1.38 0.87 0.87 0.64 0.80 0.90sd 0.75 0.60 0.57 0.26 0.21 0.55Austria mean 0.93 0.89 0.23 0.48 0.13 0.53sd 0.48 0.35 0.24 0.29 0.30 0.46Belgium mean 1.55 1.36 0.17 0.35 0.43 0.75sd 0.52 0.40 0.53 0.37 0.29 0.70Canada mean 1.56 1.16 0.72 0.25 0.82 0.88sd 0.59 0.70 0.36 0.52 0.20 0.66Denmark mean . 0.89 0.16 0.54 0.52 0.51sd . 0.69 0.52 0.45 0.31 0.55Finland mean 1.31 1.19 0.79 0.20 0.37 0.77sd 1.07 0.72 0.32 0.85 0.35 0.81France mean 1.13 0.89 0.80 0.44 0.49 0.72sd 0.21 0.35 0.24 0.41 0.20 0.38Germany mean 1.47 1.22 0.19 0.45 0.13 0.68sd 0.94 0.58 0.51 0.48 0.21 0.78Ireland mean 1.06 1.59 -0.04 1.11 1.45 1.02sd 0.57 0.81 1.41 0.50 1.37 1.13Italy mean 0.88 0.70 0.59 0.04 0.40 0.51sd 0.15 0.19 0.24 0.37 0.17 0.37Japan mean 2.17 1.79 0.53 0.73 -0.08 1.03sd 0.85 1.44 0.47 0.62 0.21 1.14Netherlands mean 1.12 0.78 0.67 0.65 0.93 0.81sd 0.37 0.76 0.32 0.24 0.76 0.54New Zealand mean 0.73 0.97 0.28 0.54 0.93 0.68sd 1.07 1.49 0.94 1.01 1.14 1.13Norway mean 1.64 1.56 0.87 1.18 0.90 1.22sd 0.58 0.45 0.53 0.61 0.53 0.61Portugal mean . 1.80 0.93 0.94 0.21 0.80sd . 0.83 0.66 0.61 0.61 0.74Spain mean 0.49 0.99 0.96 0.79 1.18 0.91sd 0.18 0.27 0.50 0.46 0.18 0.40Sweden mean 1.86 1.10 0.50 0.41 0.24 0.80sd 0.49 0.70 0.31 0.56 0.45 0.76Switzerland mean 0.37 0.22 0.36 0.10 0.07 0.21sd 0.13 0.32 0.27 0.24 0.15 0.27United Kingdom mean 0.70 0.53 0.20 0.30 0.74 0.46sd 0.75 0.62 0.32 0.39 0.82 0.60United States mean 1.20 0.12 0.67 0.26 0.44 0.51sd 0.82 0.43 0.42 0.28 0.23 0.58Total mean 1.25 1.00 0.52 0.52 0.56 0.74sd 0.77 0.80 0.61 0.58 0.65 0.73

25

Table A2: Public debt for the countries in the panel over the sample period.

Country stats 1960-69 1970-79 1980-89 1990-99 2000-07 1960-07

Australia mean . . 0.24 0.33 0.19 0.26sd . . 0.01 0.07 0.03 0.09Austria mean . 0.23 0.48 0.65 0.70 0.50sd . 0.07 0.09 0.06 0.04 0.19Belgium mean . 0.60 1.08 1.31 1.01 1.00sd . 0.04 0.17 0.07 0.09 0.28Canada mean 0.61 0.48 0.62 0.93 0.75 0.68sd 0.06 0.04 0.10 0.08 0.06 0.17Denmark mean . . 0.66 0.73 0.48 0.63sd . . 0.10 0.07 0.09 0.14Finland mean . 0.11 0.17 0.52 0.49 0.34sd . 0.03 0.02 0.18 0.04 0.20France mean . 0.33 0.35 0.57 0.70 0.48sd . 0.04 0.04 0.12 0.04 0.17Germany mean . 0.22 0.38 0.51 0.65 0.43sd . 0.05 0.04 0.10 0.04 0.17Ireland mean . 0.61 0.94 0.80 0.34 0.70sd . 0.05 0.15 0.16 0.04 0.26Italy mean 0.37 0.75 0.91 1.18 1.18 0.91sd 0.03 0.13 0.06 0.13 0.03 0.29Japan mean . 0.24 0.65 0.87 1.59 0.80sd . 0.12 0.10 0.22 0.14 0.50Netherlands mean 0.72 0.56 0.79 0.86 0.59 0.71sd 0.03 0.05 0.11 0.07 0.04 0.13New Zealand mean . . . 0.49 0.31 0.39sd . . . 0.09 0.04 0.11Norway mean . 0.43 0.34 0.34 0.47 0.39sd . 0.05 0.04 0.05 0.10 0.08Portugal mean . . . 0.66 0.69 0.68sd . . . 0.03 0.04 0.04Spain mean . . 0.47 0.64 0.55 0.58sd . . 0.02 0.11 0.08 0.11Sweden mean . 0.30 0.62 0.74 0.59 0.56sd . 0.03 0.09 0.13 0.06 0.19Switzerland mean . . . 0.45 0.54 0.49sd . . . 0.08 0.04 0.08United Kingdom mean 0.95 0.58 0.47 0.45 0.44 0.53sd 0.05 0.08 0.05 0.08 0.03 0.16United States mean 0.56 0.44 0.52 0.68 0.59 0.56sd 0.05 0.02 0.08 0.04 0.03 0.09Total mean 0.62 0.43 0.59 0.69 0.64 0.60sd 0.16 0.19 0.26 0.27 0.31 0.28

26

B Appendix: Additional information of Table 2

Table B1 presents the complete results of the models in table 2, including the estimatesfor the labour market variables. Thus, model 1 is reestimation of the unrestricted unem-ployment equation of Nymoen and Sparrman (2012), including dummies for large outliers.Model 2 includes government purchases and the indicator for the export market, and inmodel 3, the dummies for large outliers are omitted, as including them is likely to involve adownward bias in the effect of the fiscal policy. For comparison, table B1 also includes twomodels more, 2i and 3i. Comparing models 2 and 2i shows the effect of omitting the largeoutliers, which is a fairly large increase in the coefficient values of government purchasesand the export market. Model 3i extends model 3 by including year dummies. We observethat the coefficients for the export market become much smaller, reflecting considerableco-movement of the export markets for all countries. In contrast, the coefficient for thechange in government purchases is not affected, presumably because any comovement ingovernment purchases across countries is not linked to comovement in unemployment. Theestimated autoregressive coefficients have the expected sign from a wage-bargain model,cf. Nymoen and Sparrman (2012). The dynamic specification is based on that the insti-tutional variables jointly affect the wage and price-setting in the next period, cf. Nymoenand Sparrman (2012).

27

Tab

leB

1:E

stim

atio

nof

equ

atio

n(1

)w

ith

the

chan

gego

vern

men

tp

urc

has

esan

dex

por

tm

ark

ets(

tab

le11).

Model

1M

odel

2M

odel

2i

Model

3M

odel

3ia

Coef

.Std

p-v

alue

Coef

.Std

p-v

alue

Coef

.Std

p-v

alu

eC

oef

.Std

p-v

alu

eC

oef

.Std

p-v

alu

e

Unem

plo

ym

ent

pre

vio

us

per

iod

1.44

0.07

0.00

1.35

0.06

0.00

1.35

0.0

70.0

01.3

40.0

80.

001.3

70.0

70.0

0U

nem

plo

ym

ent

two

year

sag

o-0

.66

0.08

0.00

-0.5

50.

080.

00-0

.58

0.0

90.0

0-0

.57

0.10

0.0

0-0

.61

0.1

00.

00U

nem

plo

ym

ent

thre

eye

ars

ago

0.08

0.04

0.03

0.05

0.03

0.1

20.0

40.0

40.

290.0

40.0

50.4

30.0

70.0

50.

15In

stitutions:

Em

plo

ym

ent

pro

tect

ion

(EP

L),

1st

diff

.pre

vio

us

per

iod

0.45

0.39

0.25

0.38

0.36

0.2

90.

330.4

30.4

40.3

10.4

20.4

60.0

80.3

70.8

4E

PL

,tw

oye

ars

ago

0.2

70.

120.

020.

230.

130.0

80.1

60.

160.3

10.1

60.1

60.

310.0

90.1

70.6

1B

enefi

tre

pla

cem

ent

rati

o(B

RR

),1s

tdiff

.pre

vio

us

per

iod

0.7

90.

740.

281.

090.

780.1

62.2

50.9

50.0

22.2

70.

970.0

21.

790.7

00.0

1B

RR

,tw

op

erio

ds

ago

1.48

0.68

0.03

0.94

0.58

0.1

11.6

90.6

00.0

01.7

10.6

10.0

01.5

10.6

50.0

2B

enefi

tdura

tion

(BD

),1s

tdiff

.pre

vio

us

per

iod

-0.9

50.

690.

17-0

.54

0.66

0.4

2-0

.38

0.85

0.6

6-0

.38

0.8

30.6

5-0

.18

0.5

70.

75B

D,

two

per

iods

ago

-0.7

90.

480.

10-0

.54

0.40

0.18

-1.0

20.4

00.0

1-1

.02

0.4

10.0

1-1

.03

0.4

20.0

1In

tera

ctio

n-

BR

Ran

dB

D1s

tdiff

.pre

vio

us

per

iod

-1.3

22.

730.

630.

382.

440.

883.

143.0

90.3

12.8

12.9

80.

352.6

42.6

40.3

2In

tera

ctio

n-

BR

Ran

dB

Dtw

op

erio

ds

ago

3.4

71.

940.

072.

421.

490.

103.

271.8

20.

073.2

91.8

60.0

83.6

71.

770.0

4In

tera

ctio

n-

CO

and

UD

NE

T1s

tdiff

.pre

vio

us

per

iod

-4.5

72.

430.

06-6

.42

2.88

0.03

-5.5

83.3

00.0

9-5

.29

2.7

90.0

6-4

.28

3.0

80.1

6In

tera

ctio

n-

CO

and

UD

NE

Ttw

op

erio

ds

ago

-0.7

60.

770.

33-1

.01

0.46

0.0

3-1

.14

0.83

0.1

7-1

.09

0.8

60.

21-0

.88

0.8

00.

27In

tera

ctio

n-

CO

and

TA

X1st

diff

.pre

vio

us

per

iod

-6.9

55.

230.

18-7

.10

4.01

0.0

8-8

.12

4.7

70.

09-7

.79

4.6

90.1

0-6

.66

4.8

70.1

7In

tera

ctio

n-

CO

and

TA

Xtw

op

erio

ds

ago

-1.6

31.

340.

22-0

.76

1.28

0.55

-1.9

61.6

20.2

3-1

.99

1.6

10.2

2-1

.56

1.6

40.3

4U

nio

nden

sity

(UD

NE

T),

1st

diff

.pre

vio

us

per

iod

-0.5

53.

880.

89-1

.58

3.73

0.6

70.

013.3

91.0

00.1

53.4

90.9

71.9

23.9

30.6

3U

DN

ET

,tw

op

erio

ds

ago

0.91

0.96

0.35

0.88

1.02

0.39

1.75

1.2

20.1

51.7

31.2

20.

160.9

61.2

70.4

5C

oor

din

atio

n(C

O),

1st

diff

.pre

vio

us

per

iod

0.2

10.

290.

460.

190.

230.

42-0

.44

0.3

10.1

6-0

.47

0.32

0.1

4-0

.34

0.37

0.3

6C

O,

two

per

iods

ago

-0.0

50.

100.

63-0

.07

0.11

0.5

3-0

.15

0.1

50.

33-0

.15

0.1

50.3

3-0

.10

0.1

50.4

8T

axle

vel

(TA

X),

1st

diff

.pre

vio

us

per

iod

-5.6

33.

050.

06-3

.57

2.74

0.1

9-4

.48

2.4

10.

06-4

.12

2.3

40.0

8-0

.21

2.2

40.9

2T

AX

,tw

op

erio

ds

ago

2.01

1.43

0.16

1.40

1.45

0.3

31.

961.7

20.2

52.0

01.7

10.2

42.7

01.6

50.1

0L

arg

eou

tlie

rb0.

84

0.04

0.00

0.78

0.03

0.0

0Demand

components:

Exp

ort

mar

ket,

1st

diff

.(∆XM

t)-0

.42

0.07

0.0

0-0

.49

0.1

00.0

0-0

.48

0.1

00.0

0-0

.39

0.1

40.0

1E

xp

ort

mar

ket,

pre

v.

per

iod

(XM

t−1)

0.14

0.07

0.05

0.16

0.0

80.0

40.1

70.0

80.0

40.0

50.1

30.

69

Exp

ort

mar

ket,

1st

diff

.pre

v.

per

iod

(∆XM

t−1)

-0.2

10.

060.0

0-0

.18

0.0

80.

04-0

.18

0.0

80.0

3-0

.07

0.1

20.5

6C

han

gegov

t.purc

hase

s,1s

tdiff

.(∆dG

t)-0

.20

0.07

0.01

-0.2

70.0

80.0

0C

han

gego

vt.

purc

has

es,

pre

v.

per

iod

(dG

t−1)

-0.3

00.

100.0

0-0

.31

0.11

0.0

1C

han

gegov

t.purc

hase

s,1s

tdiff

.pre

v.

per

iod

(∆dG

t−1)

0.09

0.07

0.1

70.0

70.0

80.4

3C

han

gegov

t.purc

hase

s,(dG

t)-0

.27

0.0

80.0

0-0

.28

0.0

80.0

0

Obs

=C

ountr

y*A

vera

gegro

ups

483

2024

.148

320

24.1

483

20

24.1

483

20

24.

1483

20

24.1

Sta

ndar

ddev

iati

onof

resi

duals

0.5

70.

520.6

30.6

30.5

9χ2

of

all

the

exog

enou

sva

riab

les.c

1162

82.9

3(0

.00)

1397

.08

(0.0

0)25

40.

83(0

.00)

5777.6

0(0

.00)

51711.8

0(0

.00)

χ2

ofdum

my,

fisc

alp

olicy

and

exp

orts

.c14

69.3

1(0

.00)

χ2

ofp

olic

yan

dex

por

ts.c

144.

78(0

.00)

121.

25(0

.00)

107.6

1(0

.00)

68.

73(0

.00)

1st

order

auto

corr

elat

ionc

1.01

(0.3

2)0.

99(0

.32)

1.5

4(0

.12)

1.5

0(0

.13)

1.3

4(0

.18)

2nd

order

auto

corr

elat

ionc

-2.8

8(0

.00)

-2.5

5(0

.01)

-1.8

3(0

.07)

-1.7

4(0

.08)

-1.7

4(0

.08)

Dep

enden

tva

riable

:U

nem

plo

ym

ent.

Est

imati

on

met

hod:

Fix

edeff

ect

coeff

cien

tses

tim

ate

,ro

bust

standard

erro

rs(x

treg

wit

h,r

obust

)is

use

din

all

the

regre

ssio

ns.

a)

Wit

hti

me

dum

mie

s.

b)

Bre

ak

by

larg

eoutl

ier

appro

ach

.

c)N

um

ber

sin

pare

nth

eses

are

p-v

alu

esfo

rth

ere

leva

nt

null.

28

References

Alesina, A. F. and S. Ardagna (2009, October). Large Changes in Fiscal Policy: TaxesVersus Spending. Technical Report 15438, National Bureau of Economic Research.

Ardagna, S. (2007). Fiscal policy in unionized labor markets. Journal of Economic Dy-namics and Control 31 (5), 1498–1534.

Armingeon, K., S. Engler, P. Potolidis, M. Gerber, and P. Leimgruber (2010). Codebook:Coparative political data set I 1960-2008.

Auerbach, A., W. Gale, and B. Harris (2010). Activist fiscal policy. Journal of EconomicPerspectives 24 (4), 141–164.

Auerbach, A. J. and Y. Gorodnichenko (2012). Measuring the output responses to fiscalpolicy. American Economic Journal: Economic Policy 4 (2), 1–27.

Baum, C. F., M. E. Schaffer, and S. Stillman (2002, April). Ivreg2: Stata module forextended instrumental variables/2sls and gmm estimation. Statistical Software Compo-nents, Boston College Department of Economics.

Beetsma, R. and M. Giuliodori (2010). Discretionary Fiscal Policy: Review and Estimatesfor the EU. Economic Journal 121 (1), F4–F32.

Bertola, G. (2010). Fiscal Policy and Labor Markets at Times of Public Debt. NordicEconomic Policy Review 1 , 111–148+.

Blanchard, O. and R. Perotti (2002, November). An Empirical Characterization Of TheDynamic Effects Of Changes In Government Spending And Taxes On Output. TheQuarterly Journal of Economics 117 (4), 1329–1368.

Bruckner, M. and E. Pappa (2012). Fiscal expansions, unemployment, and labor forceparticipation. International Economic Review 54 (4), 1205–1228.

Caballero, R. (2010). Macroeconomics after the crisis: Time to deal with the pretense-of-knowledge syndrome. Journal of Economic Perspectives Forthcoming.

Chari, V. V., P. J. Kehoe, and E. R. McGrattan (2009, January). New Keynesian Models:Not Yet Useful for Policy Analysis. American Economic Journal: Macroeconomics 1 (1),242–66.

Christiano, Lawrence J, E. M. and S. Rebelo (2011). When is the government spendingmultiplier large? Journal of Political Economy 119 (1), 78–121.

Coenen, G., C. Erceg, C. Freedman, D. Furceri, M. Kumhof, R. Lalonde, D. Laxton,J. Linde, A. Mourougane, D. Muir, S. Mursula, C. D. Resende, J. Roberts, W. Roeger,S. Snudden, M. Trabandt, and J. Veld (2010). Effect of Fiscal Stimulus in StructuralModels. Technical report.

Duell, N., S. Singh, and P. Tergeist (2009, February). Activation policies in norway. Techni-cal report, http://www.olis.oecd.org/olis/2009doc.nsf/linkto/delsa-elsa-wd-sem(2009)4.

Eggertson, G. B. and M. Woodford (2003). The zero bound on interest rates and optimalmonetary policy. Brookings Papers on Economic Activity 1, 139–211.

Eichenbaum, M. and J. D. M. Fisher (2005, February). Fiscal Policy in the Aftermath of9/11. Journal of Money, Credit and Banking 37 (1), 1–22.

29

Fatas, A. and I. Mihov (2001, April). The Effects of Fiscal Policy on Consumption and Em-ployment: Theory and Evidence. CEPR Discussion Papers 2760, C.E.P.R. DiscussionPapers.

Furlanetto, F. (2011). Fiscal stimulus and the role of wage rigidity. Journal of EconomicDynamics and Control 35, 512–527.

Giavazzi, F. and M. Pagano (1990, May). Can Severe Fiscal Contractions be Expansion-ary? Tales of Two Small European Countries. NBER Working Papers 3372, NationalBureau of Economic Research, Inc.

Hall, R. E. (2009, November). By How Much Does GDP Rise if the Government BuysMore Output? Technical Report 15496, National Bureau of Economic Research.

Holmlund, B. and J. Linden (1993, July). Job matching, temporary public employment,and equilibrium unemployment. Journal of Public Economics 51 (3), 329–343.

Ilzetzki, E., E. G. Mendoza, and C. A. Vegh (2010, October). How Big (Small?) are FiscalMultipliers? NBER Working Papers 16479, National Bureau of Economic Research,Inc.

IMF (2010, October). Will it hurt? Macroeconomic effects of fiscal consolidation. WorldEconomic Outlook .

Judson, R. A. and A. L. Owen (1999). Estimating dynamic panel data models: a guidefor macroeconomists. Economics Letters 65 (1).

Layard, R., S. Nickell, and R. Jackman (1991). Unemployment. Oxford: Oxford UniversityPress.

Layard, R., S. Nickell, and R. Jackman (2005). Unemployment (2 ed.). Oxford: OxfordUniversity Press. First published 1991.

Linnemann, L. (2009). Macroeconomic effects of shocks to public employment. Journalof Macroeconomics 31, 252–267.

Matyas, L. and P. Sevestre (2008). The Econometrics of Panel Data. Fundamentals andRecent Developments in Theory and Practice, Volume 46. Berlin Heidelberg: Springer.

Michaillat, P. (2012). A theory of countercyclical government-consumption multiplier.Technical report, London School of Economics.

Monacelli, T., R. Perotti, and A. Trigari (2010, April). Unemployment Fiscal Multipliers.Technical Report 15931, National Bureau of Economic Research.

Nakamura, E. and J. Steinsson (2012). Fiscal stimulus in a monetary union: Evidencefrom U.S. regions. Technical report, Columbia University.

Nickell, S. (1981). Biases in dynamic models with fixed effects. Econometrica 49 (6), pp.1417–1426.

Nickell, S., L. Nunziata, and W. Ochel (2005). Unemployment in the OECD Since the1960s. What Do We Know? Economic Journal 115 (500), 1–27.

Nymoen, R. and V. Sparrman (2012). Panel Data Evidence on the Role of Institutionsand Shocks for Unemployment Dynamics and Equilibrium. Department of Economics,University of Oslo MEMO 20.

30

OECD (2008a). Economic Outlook No. 84. Economic Outlook: Annual Projections forthe OECD Countries (database).

OECD (2008b). OECD Economic Outlook No. 83. OECD Economic Outlook: AnnualProjections for the OECD Countries (database).

OECD (2010). SITC Revision 2 (Standard International Trade Classification Revision 2,1976) - data from 1961 onwards.

Pappa, E. (2009). The effects of fiscal shocks on employment and the real wage. Interna-tional Economic Review 50 (1), 217–244.

Perotti, R. (2007, June). In Search of the Transmission Mechanism of Fiscal Policy.Technical Report 13143, National Bureau of Economic Research.

Ramey, V. (2011). Identifying government spending shocks: It’s all in the timing. Quar-terly Journal of Economics 126 (1), 1–50.

Ramey, V. A. (2012, November). Government spending and private activity. In Fiscal Pol-icy after the Financial Crisis, NBER Chapters. National Bureau of Economic Research,Inc.

Ramey, V. A. and M. D. Shapiro (1999, April). Costly Capital Reallocation and theEffects of Government Spending. Technical Report 6283, National Bureau of EconomicResearch, Inc.

Roodman, D. (2009). A Note on the Theme of Too Many Instruments. Oxford Bulletinof Economics and Statistics 71 (1), 135–158.

Shi, M. and J. Svensson (2006, September). Political budget cycles: Do they differ acrosscountries and why? Journal of Public Economics 90 (8-9), 1367–1389.

The Conference Board (2010). The Conference Board Total Economy Database, January2010. Technical report.

Tornell, A. and P. R. Lane (1999, March). The Voracity Effect. American EconomicReview 89 (1), 22–46.

Woodford, M. (2011). Simple analytics of the government expenditure multiplier. Amer-ican Economic Journal: Macroeconomics 3 (1), 1–35.

31

Do government purchases a ect unemployment? · Do government purchases a ect unemployment? Steinar...

Documents

Transcript of Do government purchases a ect unemployment? · Do government purchases a ect unemployment? Steinar...