Jobs Multipliers: Evidence from a Large Fiscal Stimulus in Spain — Documentos de Trabajo N.° 1922 (2019)

Summary

A 2019 Banco de Espana working paper, Documentos de Trabajo N.° 1922, by Mario Alloza and Carlos Sanz, titled Jobs Multipliers: Evidence from a Large Fiscal Stimulus in Spain. The abstract states that the authors estimate the employment effect of PlanE, in which the national government transferred funds to municipalities for local investment projects. Using a difference-in-difference approach, they find that 100,000 euros of stimulus reduced unemployment by 0.62 jobs per year. They find spatial effects representing 8.4% of the local effect and state that their estimate of the multiplier falls in the lower range of previous work.

Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used

Full text

JOBS MULTIPLIERS: EVIDENCE FROM A LARGE FISCAL STIMULUS IN SPAIN
JOBS MULTIPLIERS: EVIDENCE FROM A LARGE FISCAL
STIMULUS IN SPAIN

Mario Alloza and Carlos Sanz

BANCO DE ESPANA

(*) We thank Henrique Basso, Jesus Fernandez-Villaverde, Enrique Moral-Benito, Omar Rachedi, Roberto Ramos,
Tano Santos, Jan Stuhler, Javier Vallés, Ernesto Villanueva, and seminar participants at the Banco de Espana for
helpful comments. The views expressed in this paper are those of the authors and do not necessarily coincide with
the views of the Banco de Espafa or the Eurosystem. Alloza: Banco de Espana, Calle Alcala 48, 28014 Madrid
(Spain), m.alloza@bde.es. Sanz: Banco de Espana, Calle Alcala 48, 28014 Madrid (Spain), carlossanz@bde.es.

Documentos de Trabajo. N.° 1922
2019
The Working Paper Series seeks to disseminate original research in economics and finance. All papers
have been anonymously refereed. By publishing these papers, the Banco de Espana aims to contribute
to economic analysis and, in particular, to Knowledge of the Spanish economy and its international
environment.

The opinions and analyses in the Working Paper Series are the responsibility of the authors and, therefore,

do not necessarily coincide with those of the Banco de Espana or the Eurosystem.

The Banco de Espana disseminates its main reports and most of its publications via the Internet at the
following website: http://www.bde.es.

Reproduction for educational and non-commercial purposes is permitted provided that the source is
acknowledged.
© BANCO DE ESPANA, Madrid, 2019

ISSN: 1579-8666 (on line)
Abstract

We estimate the employment effect of a large fiscal stimulus in Spain (PlanE), in which
the national government transferred funds to municipalities to carry out local investment
projects. Using a difference-in-difference approach by exploiting variation in the timing of the
execution of projects across municipalities, we find that 100,000 euros of stimulus reduced
unemployment by 0.62 jobs per year. We allow for possible spatial effects, i.e. the propagation
of the stimulus to neighboring municipalities, and find that these are sizable, representing
8.4% of the “local” effect. We also present evidence on the transmission mechanism, finding
that the effect was: (i) initially concentrated in the construction and industrial sectors, but
later spilled over to the broader economy, (ii) larger for males than females, (iii) larger when
the shock represented a higher share of the budget, and (iv) not larger for municipalities
headed by more educated mayors. Our estimate of the multiplier falls in the lower range of
previous work.

Keywords: fiscal policy, local fiscal multipliers, spillovers.

JEL classification: E24, E62, H380, H72.
Resumen

En este trabajo, estimamos los efectos en el empleo de un gran estimulo fiscal en Espana
(conocido como «Plan E»), por el cual el Gobierno central transfirid fondos a los municipios
para llevar a cabo proyectos de inversion. Explotando la variacién temporal en la ejecucidn de
los proyectos en los distintos municipios mediante un andlisis de diferencias en diferencias,
encontramos que 100.000 euros de estimulo redujeron el desempleo en 0,62 empleos-ano.
Consideramos la posibilidad de que existan efectos espaciales, esto es, la propagacién
del estimulo a municipios vecinos, y encontramos que son considerables: suponen un
8,4 % del efecto «local». También presentamos evidencia sobre la transmision del estimulo, y
vemos que el efecto: i) estuvo inicialmente concentrado en la construccidn y en la industria,
pero se extendid luego al resto de la economia; ii) fue mayor para los hombres que para
las mujeres; iii) fue mayor cuando el estimulo fue mas grande en proporcidn al presupuesto
municipal, y iv) no fue mayor en municipios con alcaldes de un mayor nivel educativo.
Nuestra estimacién del multiplicador se encuentra en el rango inferior de lo hallado por
estudios anteriores.

Palabras clave: politica fiscal, multiplicadores fiscales locales, efectos de desbordamiento.

Cédigos JEL: E24, E62, H30, H72.
1 Introduction

In the last decade, there has been a renewed attention to fiscal policy. Given a macroeconomic
environment characterized by a constrained monetary policy, understanding how effective
government initiatives are in stimulating the economy has become an important topic of
discussion among academics and policy makers. This is particularly relevant when analyzing
the expansionary fiscal plans that were enacted in recent years to stimulate employment
growth during recessions.! However, it is challenging to identify the causal effects of such
plans.”

In this paper, we estimate the employment effects of a large, unanticipated fiscal stimulus
in Spain (Plan Espanol para el Estémulo de la Economta y el Empleo, the Spanish Plan for
the Stimulus of the Economy and Employment, commonly know as Plan EF). This stimulus,
approved in 2008, channeled almost 13 billion euros (around 1.2% of Spanish GDP) to
municipal governments to execute public investment plans.®

In our empirical analysis, we use municipality and monthly-level data on unemployment
and the stimulus. While all municipalities received the same amount of resources (in per
capita terms), there was variation in the timing of the execution of the projects. We exploit
this variation in our estimation, performing a difference-in-difference analysis to establish a
causal relationship. The key assumption is that “early” and “late” starters were on parallel
trends around the time of the stimulus. While we cannot directly test whether this was the
case, we perform a test of parallel trends through an event study, i.e. we do placebo tests
to see whether “early” and “late” starters were on parallel trends in the months before the
stimulus. Reassuringly, the results of these tests indicate that this was indeed the case.

We then extend our benchmark specification to analyze potential spatial effects, i.e.
how a fiscal stimulus in municipality i affected other municipalities close to i.+ Restricting
the estimation of the fiscal stimulus to a particular municipality ignores two important
aspects. First, it omits a direct effect, particularly relevant for the case of small and medium-
sized municipalities, due to the fact that some of the works may be contracted to firms in
neighboring (potentially larger) municipalities. Second, it ignores general equilibrium effects,
since enhanced local economic activity could spillover to close areas.

Our estimate of the jobs multiplier (the number of jobs created per million euros of

public spending) is 5.7 jobs at the peak (significant at conventional levels). Regarding the

‘For example, in the case of the United States, the 2009 American Recovery and Reinvestment Act
(ARRA) states employment protection and expansion as its primary objectives.

?When using variation from aggregate time series data, fiscal policies originated by a response to changing
macroeconomic conditions cannot be used to established a causal relationship. See for example Nakamura
and Steinsson (2014) and Suarez Serrato and Wingender (2016) as examples of using disaggregated data to
circumvent this empirical challenge.

3To the best of our knowledge, only two other research papers have used Plan E data. Montolio (2018)
studies the effects of the plan on crime using data from the region of Catalonia. Carozzi and Repetto (2017)
study how local governments allocated investment projects across neighborhoods.

“For this, we follow the empirical design of Huber (2018), which is based on adding to the baseline
equation the shock in municipalities that fall within a given radius of 7.

BANCO DE ESPANA 7 DOCUMENTO DE TRABAJO N.° 1922
cumulative multiplier (the sum of the point multipliers over the considered horizon), we find
that 100,000 euros reduced unemployment by 0.62 job-years. Our results also point to the
presence of substantial spatial effects, representing 8.4% of the local effect (and about 20%
during the first 6 months).

While jobs multipliers are an object of interest in their own right (see, e.g., Wilson
(2012)), they offer an incomplete view of the overall effects of the stimulus. To (partially)
address this issue, we follow Chodorow-Reich (2019) and translate the employment multiplier
to an output multiplier. Our estimates are consistent with an output multiplier of 0.3-0.5.
This figure represents a strict lower bound, since our approximation ignores that new public
capital could enhance economic activity in the medium and long run.

Finally, we provide evidence on the transmission channel of the stimulus. We find that:
(i) most of the effect was initially in the construction and industrial sectors, but later spilled
over to the broader economy, (ii) the stimulus reduced male more than female unemployment,
suggesting non-neutral gender implications of the fiscal action, (iii) while all municipalities
received the same (per capita) amount of resources, the stimulus was more effective when the
received amount represented a higher percentage of the budget, and (iv) there is no evidence
that the stimulus was more effective in reducing unemployment in municipalities with more
educated mayors.

Our work contributes to the growing literature that employs sub-national units of anal-
ysis to estimate jobs multipliers.° Several papers have studied the effects of the American
Recovery and Reinvestment Act (ARRA), with a wide range of estimates. For example, while
Chodorow-Reich, Feiveson, Liscow, and Woolston (2012) find that 100K dollars increased
employment by 3.8 job-years, Conley and Dupor (2013) find that 100K dollars increased
employment by 0.5 job-years if fungibility between ARRA and lost tax revenue is imposed
(0.76 if fungibility not imposed). The estimate implied by Wilson (2012), 1.75 job-years,
falls in between the other two.° Non-ARRA papers have found (on average) larger multipli-
ers. For example, Sudrez Serrato and Wingender (2016) find that 100K spending increases
employment by 3.25 job-years while Shoag (2015) estimates this multiplier at 2.89 job-years.

We contribute to this literature in various ways. First, we exploit an arguably exogenous
variation in the timing of the implementation of the stimulus, allowing us to provide credible
estimate of causal effects. In this regard, we provide evidence supporting the identification
assumption through an event study, i.e. placebo tests on lagged outcomes. Second, we
employ a very fine unit of observation at both the cross-sectional level (municipality) and
the time level (monthly). Hence, we have a large number of observations (3,255 municipalities

over 108 months) and obtain precise estimates.’ To the best of our knowledge, we are the

°For a review, see Chodorow-Reich (2019). More broadly, our paper is related to the literature that
uses a sub-national level of analysis to estimate local income multipliers—see, for example, Nakamura and
Steinsson (2014).

°This number comes from a calculation by Chodorow-Reich (2019), as Wilson (2012) does not provide
an estimate of the cumulative multiplier.

"For example, Wilson (2012) obtains a 95% confidence interval of {1.1,15] for the one-year point multiplier,
while ours is 35% smaller ([0.1,9]).

BANCO DE ESPANA 8 DOCUMENTO DE TRABAJO N.° 1922
first to use municipality and monthly data to estimate fiscal multipliers. Third, we extend
the benchmark framework of local fiscal multipliers and estimate spatial effects, i.e. the
impact of a stimulus in a given municipality on neighboring municipalities. Fourth, we
provide evidence on the underlying transmission channel by exploiting detailed data on the
sectoral composition of unemployment, budget size, and the education level of mayors.

Our estimates of the multiplier fall in the lower range of previous estimates. Our paper
also suggests that it is important to take spatial effects into account when estimating local
fiscal multipliers.® Still, our estimate of the multiplier remains in the lower range of previous
work even after including spatial effects. Our jobs multiplier (and implied output multipliers)
are compatible in size with aggregate estimations of nation-wide effects. This evidence
suggests that there might be instances where the local multipliers are not always larger than
nation-wide multipliers. Following Farhi and Werning (2016), we argue that this might be
the case for countries belonging to a currency union, like Spain.

The rest of the paper is organized as follows. Section 2 provides some background on the
Plan E stimulus. Section 3 describes the data and the empirical strategy. Section 4 presents
the main results and robustness checks. Section 5 extends the analysis to take spatial effects
into account. Section 6 discusses the magnitude of estimates in light of previous literature.

Section 7 looks into the transmission mechanism. Section 8 concludes.

2 Background: The Plan E Stimulus

Spain was heavily affected by the Great Recession that started in 2007. GDP growth slowed
down to 1.1% in 2008 (from 4.2 and 3.8% in 2006 and 2007, respectively). The most dramatic
change at the beginning of the crisis, however, was the deterioration of the labor market.
Unemployment rate soared from 8.2% in 2007 to 11.2% in 2008.° Unemployment rate kept
on growing during the following years, reaching a peak at 26% in 2013.

In this context, the government of primer minister José Luis Rodriguez Zapatero ap-
proved, on November 28, 2008, a fiscal stimulus with the goal of “creating jobs”. Through
this stimulus, officially named the Fondo Estatal de Inversién Local and popularly known
as Plan E,'° the national government would transfer 8 billion euros to municipalities pro-
portionally to their population size: each municipality would receive up to 177 euros per
capita, which had to be spent on local public investment projects, of up to 5 million euros

per project. Importantly, the stimulus was approved when the budgets for the next year

8 Admittedly, the size of spatial effects may be contingent on the size of the unit of analysis, i.e. it may
be more relevant at the municipality level than at a larger level.

°It has been argued that this was due to the rigidity of the labor market, which made firms to adjust
through employment rather than wages, and to the large gap between the firing costs of permanent and
temporary contracts (Bentolila, Cahuc, Dolado, and Le Barbanchon (2012)).

10When initially conceived, the fiscal stimulus was given the official name of Fondo Estatal de Inversion
Local. A year later, a second wave was approved under the name of Fondo Estatal para el Empleo y la
Sostenibilidad Local. Both waves are commonly referred to as Plan E.

BANCO DE ESPANA 9 DOCUMENTO DE TRABAJO N.° 1922
(2009) had already been approved, and hence funding could not be devoted to local invest-
ment included in the 2009 budget.!' Rather, to obtain the funds, municipalities had to
submit project proposals to the national government. As can be seen in Panel (a) of Figure
A2, almost all municipalities received the maximum possible amount 177 euros per capita—
given that this was free money from a municipality’s perspective, there was no incentive
not to ask for it. Projects were submitted from December 10, 2008 to January 24, 2009,
and approved (or, in very few cases, rejected) by the national government from January to
March 2009. After the approval of projects, local governments had to follow a tender proce-
dure to award the contracts. This procedure lasted a maximum of four months, after which
construction had to start. An important feature in the design of the Plan E is the special
emphasis placed on reducing the gap between the submission and the approval of projects
(through a novel electronic system for project submissions) and between the approval and
the execution of projects (special provisions allowed the tender procedures to be conducted
through an expedited process). Upon approval of the projects, municipalities received up
to 70% of the budget cost of the works, with the remaining funds being reimbursed upon
completion. In this first wave, 8,108 out of 8,112 Spanish municipalities participated in the
Plan E, presenting a total of 30,700 projects with an approval ratio above 99.5%.'?

In 2010, a second wave of the stimulus (Fondo Estatal para el Empleo y la Sostenibilidad
Local), was implemented. The procedure was very similar to the one of the first wave, but
the size was smaller: municipalities could now receive up to 108 euros per capita, with the
total stimulus amounting to 5 billion euros. In this wave, there was a special provision that
allowed municipalities to use up to 20% of the stimulus in funding municipal government
consumption (as reflected in the local budget). Because of this reason, not all municipalities
allocated the resources in public investment projects, as can be seen in in Panel (b) of
Figure A2.'° In the second wave, there were 25,362 submissions regarding public investment
projects, with an approval rate of 99.9%.'4

Overall, the fiscal plan had two characteristics that make it an appealing setting to es-
timate local fiscal multipliers. First, the stimulus was very large quantitatively. Around
13,000 million euros were spent, amounting to 1.2% of GDP, or twice the total yearly mu-
nicipal budget in public investment. PM Zapatero mentioned that the Plan E implied “an
unprecedented mobilization of public resources”. The large size of the shock will allow us

to obtain precise estimates. Second, the stimulus was mostly unanticipated. The approval

This can be clearly gathered from Figure Al, which plots capital transfers over time. Capital transfers
are the funds received by municipalities from higher-order public administrations (including the national
government) to undertake capital expenditures, such as infrastructure work. Figure Al shows that, when
the budgets were drafted (blue line), funds received to execute capital expenditures were expected to remain
roughly stable. The actual, executed budget (red line) reflects that municipalities received an unexpected
amount of resources both in 2009 and 2010 to fund public investment.

12See Ministerio de Polftica Territorial y Administracién Publica (2010).

'3On average, 14.8 of the stimulus resources were allocated to local government consumption, with the
remaining 85.2% being allocated to public investment projects. We account for this special feature in our
empirical strategy described in Section 3.2.

14See Ministerio de Politica Territorial y Administracién Piiblica (2011).

BANCO DE ESPANA 10 DOCUMENTO DE TRABAJO N.° 1922
of the Plan E on 28 November 2008 came as a surprise.'°? After the approval, the process
followed until the beginning of construction was extremely fast. As can be seen in Figure
1, which shows the starting dates of the projects, construction had started in 75% of the
projects five months after the announcement, and in 90% after six months. The absence of
long anticipation will help us validate our empirical strategy with pre-trend placebo tests,

as discussed in the next section.

3 Data and Empirical Strategy

3.1 Data

We use data on the fiscal stimulus and unemployment at the municipality-month level. There
are 8,112 municipalities in Spain. Given that many of these municipalities are very small
and employment statistics are measured with more error, we will focus on municipalities of
more than 1,000 inhabitants (3,255 municipalities).1° Our sample period covers all months
from January 2006 to December 2014.

Our data for the fiscal stimulus comes from the Ministry of Public Administration (Minis-
terio de Politica Territorial y Administracién Publica). For each project, we can observe the
dates in which construction started and finished, its size (i.e. the amount of money spent on
it), and asummary description of the project (e.g. industrial promotion vs. refurbishment of
an existing public building). Our sample consists of a total of about 56,000 projects (30,700
in the first wave). The average size of projects was 215,290 euros (median: 75,000), ranging
from 3,000 to 5,800,000 euros. In our analysis, we focus on the starting date of construc-
tion since it proxies for the moment when firms take economic decisions (particularly, hiring
workers). Given that our unit of observation will be the municipality-month, we aggregate
the total amount spent by municipality and month. Hence, our treatment variable, x;,, will
be the (per capita) amount spent by municipality i on projects that started during month t.

Our unemployment data comes from the Ministry of Employment and Social Security
(Ministerio de Empleo y Seguridad Social). We can observe the number of registered unem-
ployed individuals by municipality and month. Our outcome variable, y;z, is the logarithm of
the number of unemployed individuals in municipality 7 in month t. In addition to the total
number of unemployed, we can observe some demographic and sectorial characteristics: the
number of unemployed by gender, by age group (less than 25, 25-45, and more than 45 years

old), and by sector (construction, industry, services, and agriculture). To shed light on the

'5The first public announcement of the approval was on November 22, when PM Zapatero announced
“by surprise” that he would be laying out some “fast, municipality-based” actions to stimulate the economy
(Diez (2008)).

'6More precisely, we focus on municipalities whose population was larger than 1,000 in 2007 (the year
before the first wave of stimulus). For very small municipalities, registered employment statistics only report
aggregate figures.

BANCO DE ESPANA 1 1 DOCUMENTO DE TRABAJO N.° 1922
transmission mechanism, we will estimate the heterogeneous effect of the stimulus by these
demographic and sectorial characteristics in Section 7. As an alternative outcome variable,
we will consider the logarithm of the number of workers affiliated to Social Security (formal
employment). We prefer, however, to focus on unemployment for two reasons. First, from
a statistical point of view, registered unemployment is a more reliable measure of dynamics
in the labor market.!” Second, it facilitates relating our estimates of the multiplier to those
of previous literature, which generally uses unemployment as the outcome variable.

Figure 2 plots the evolution of unemployment over our sample period (2006-2014), along
with the amount spent in the two waves of stimulus. We can see that unemployment almost
doubled in the year leading up to the Plan E, from 2 to 3.6 million unemployed. Unemploy-
ment continued growing, although at a slower pace, until February 2013, in which it started
to decrease.

In our analysis, we exclude outliers in both the dependent and independent variables.'®

In Section 4.2, we show that the results do not change when outliers are included.

3.2. Empirical Strategy

To estimate the dynamic impact of the stimulus on unemployment, we follow a local projec-

tion approach (Jorda (2005)), estimating a series of h regressions:

M

Yitth = PYit—-1 + Ss" Bn gjCit—j +O + Ve + itn, (1)
j=0

where yi2+n is the log of the unemployment in municipality 7 at month t+h (h € |—24, 24]),
Xiz—; is the size of the fiscal stimulus in municipality 7 at month t—j, a; is a municipality fixed
effect, 7 is a date (month) fixed effect, and €;, is an idiosyncratic error term. The parameter
p captures the persistence of the outcome variable, and the {,,; capture the dynamic impact
of the fiscal stimulus. In the baseline specification, we take WM = 12, but we will show the
robustness of the results to different values of IZ. Note that we estimate one regression for

each value of h: if h = 0, 89 measures the contemporaneous effect; if h > 0, Gp,9 measures

'7Contrary to unemployment data, affiliation data suffer from two drawbacks. First, it does not include
workers affiliated to other welfare regimes different to the Social Security. Second, unemployed workers have
strong incentives to register themselves as unemployed, since it is a pre-requisite to obtain unemployment
benefits. In contrast, Social Security records may infraestimate the size of the shadow economy.

'8Regarding outliers in the dependent variable, we drop all municipality-months in which the unemploy-
ment rate (as a fraction of total population) was above 20%. The 99th percentile in our sample is 16%.
Regarding outliers in the independent variable, we exclude those municipalities that obtained (significantly)
more or less than the amount stipulated by the law (179 and 108 euros per capita, respectively )—see Figure
A2. More specifically, we drop those municipalities that spent more than 179 or less than 100 euros per
capita in the first wave, and those that spent more than 110 or less than 106 euros per capita in the second
wave. We show that our benchmark results are robust to the inclusion of these outliers.

BANCO DE ESPANA 12 DOCUMENTO DE TRABAJO N.° 1922
the effect of the stimulus h periods in the future; and, if h < 0, 6n,9 captures the effects h
periods in the past—hence, they are placebo tests, as we discuss below.

Note that, given that we include municipality and year fixed effects, this is in fact a
difference-in-difference specification, where the dynamic response is calculated through local
projections.!? Local projections are a natural way to identify the Average Treatment Effect
over time.?? An alternative method to estimate the dynamic impact of the shock would be
to construct the moving average (MA) representation of equation (1). As noted by Ramey
(2016), the local approach presents several advantages (at the cost of efficiency) over the
traditional method based on MA. Mainly, local projections is a more robust method to
the presence of misspecification, while MA estimation compounds misspecification errors at
each horizon. Additionally, local projections offers a simple and transparent way to test the
parallel-trend assumption with placebo tests.7!

The identification assumption is that there are no omitted factors that affect both the
timing of the projects and the unemployment rate, conditional on the municipality and
year fixed effects. As a first approximation to the validity of this assumption, in Figure
A3 we plot the starting dates of projects by municipality. We do not see any pattern
suggesting that early-starters cluster in any given region of the country. Of course, this
figure is far from conclusive about the validity of the approach, as there can be other relevant
factors that cannot be gathered from the map. For example, municipalities that are smaller,
poorer, or have a high unemployment rate may arguably be faster in requesting the funds,
or in the tender procedure, and hence might start the projects earlier. Importantly, in our
specification, the municipality fixed effects will absorb any time-invariant characteristic of
municipalities that might correlate with the beginning of the projects, e.g. as long as the
size, income, and unemployment rate of municipalities are moving in parallel around the
time of the stimulus, they will be captured by the fixed effects.

Hence, what is crucial for our identification is that the starting dates of projects do not
correlate with other time-varying characteristics of municipalities, e.g. that there is no shock
to population or the unemployment rate at the same time of the stimulus that differentially
affects early and late starters. Regarding this point, first note that, as argued by Montolio
(2018), the narrow time window in the execution of projects implies that the timing of the
approval, tender, and implementation phases of the projects is unlikely to have coincided
with another simultaneous shock that might have influenced the outcome variables. Second,
and more importantly, while it is not possible to directly test whether some omitted factor is

changing at precisely the same month of starting the project, we can do a standard difference-

19Since we have fixed effects and a lagged dependent variable, one could be concerned about the existence
of a Nickel bias (Nickell (1981)). However, given that we have a large-T panel, this bias should be negligible
in our context. In particular, the size of the bias diminishes with the number of periods T considered in the
sample: —(1+ p)/(T — 1) suggesting that for our sample size and a true parameter p = 0.8, the bias could
be at most -0.015 (around 1-2% of the true value of the parameter).

20See Jorda and Taylor (2016) and Angrist, Jorda, and Kuersteiner (2018) for further details.

?1Our benchmark results are similar when we construct impulse responses using the MA representation of
equation 1 (not shown).

BANCO DE ESPANA 13 DOCUMENTO DE TRABAJO N.° 1922
in-difference test for parallel pre-treatment trends, i.e. estimate whether unemployment of
early-starters and late-starters moved in parallel in the months before the beginning of the
treatment. We see this test as an important validitation of our empirical strategy.

We do this through an event study, in which we test whether the stimulus is correlated
with past values of the outcome variable (unemployment). In terms of Equation 1, we
explore the effects when h < 0. There are two reasons why we could find a non-zero effect:
(i) anticipation, and (ii) a violation of the parallel-trend assumption. However, note that,
for values h < —12, only (ii) is possible, so the regressions with h < —12 serve as “pure”
placebo tests. The reason is that the maximum possible anticipation was 12 months, the
time elapsed between the approval of the Plan E on November 28 2008 and the beginning
of the very last projects in December 2009. For values h € (—12,0), we cannot rule out that
some anticipation is happening, potentially complicating our pre-trend analysis. However,
note that, as explained in Section 3, 90% of the projects had started by mid May 2008. This
implies that, for the vast majority of cases, we should expect an anticipation of at most 6

months. Hence, regressions for h € [6,12] arguably serve as placebo tests as well.”

4 Results

4.1 Main Results

The main results are presented in Figure 3. For each h, it plots the point estimates and 95%
confidence intervals of G;,9. Reassuringly, we find no “effect” on unemployment for h < 0,
validating the parallel-trend assumption. We do observe significant effects for h > 0. In
particular, the starting of a public investment project has a negative effect on unemployment
on impact. This effects builds up over the first year and moderates towards the end of the
second year. At the peak (reached 8 months after the shock), a stimulus of 100 euros per
capita reduces unemployment by 0.8%.

Table 1 translates these effects into economically interpretable objects: the jobs multipli-
ers, defined as employment creation per million euros.?? Our estimates of the jobs multipliers
are significantly different from zero for all horizons considered at reasonable levels of signif-
icance.2* The peak effect is 5.7 jobs created per million euros.

Alternatively, one can look at the cumulative multiplier, i.e. the sum of the point multi-

pliers. We estimate a cumulative multiplier of 68.2, i.e. one million euros of stimulus reduce

22This is assuming the worst-case scenario in which anticipation starts from the date of the announcement.
In the, perhaps more realistic, case in which effects are anticipated from the approval of projects, anticipation
would be significantly shorter.

23 To convert our estimates from Figure 3 into the jobs multipliers of Table 1, we first translate the per
capita shock into an absolute shock, evaluated at the municipality with the mean population size, and then
translate the percent effect on unemployment into an absolute effect, evaluated at the municipality with the
mean unemployment.

24 Al] multipliers are significantly different from zero at the 68% level and all but the impact are significantly
different from zero at the 95% level.

BANCO DE ESPANA 14 DOCUMENTO DE TRABAJO N.° 1922
unemployment by 68.2 job-months, or 100,000 euros do so by 0.57 job-years. Given that
spatial effects may be present in our context, we will postpone putting these numbers into

perspective within the literature until Section 6, after we have estimated those.

4.2 Robustness

Alternative dynamic structure. We study the robustness of the results to changing
the dynamic structure. In particular, in our baseline results, we included one lag of the
dependent variable and 12 lags of the shock. Here, we consider instead 0, 6, or 12 lags of
the dependent variable, and 12 or 24 lags of the shock. The results (Figure 4) show that the

estimates remain very similar to those of the baseline.

Alternative sample selection. We analyze how changes in the sample choice may affect
our benchmark estimates. Figure 5 shows alternative specifications that include municipali-
ties considered as outliers in the benchmark sample (either because of unusually high or low
values of the dependent or independent variables). Results suggest that these alternative

specifications do not significantly alter our baseline estimations.

Effect on employment. As an alternative dependent variable, we consider the number
of workers affiliated to Social Security.”° The results, which are shown in Figure A4, suggest
that the parallel-trend assumption holds and that employment increases after the stimulus.

The effects, however, are imprecisely estimated and non-significant.?°

5 Spatial Effects

While a growing number of studies use disaggregated geographical areas to estimate the
effect of fiscal policy (see, for example, Nakamura and Steinsson (2014)), most ignore that
estimates at the local level omit spatial effects arising from general equilibrium aggrega-
tions.?” Restricting the estimation of the fiscal stimulus to a particular municipality ignores
two important aspects. First, it omits a direct effect, particularly relevant for the case of
small and medium-sized municipalities, due to the fact that some of the works may be con-
tracted to firms in neighboring (potentially larger) municipalities. Second, it ignores general
equilibrium effects, since enhanced local economic activity could spill over to close areas.
In this section, we shed light on how large these spatial effects are. To do so, we estimate
the effects that the fiscal stimulus has on nearby municipalities. In particular, we follow
Moretti (2010) and Huber (2018) and enlarge our main specification in equation (1) by

adding an additional regressor that contains the average shock occurring in municipalities

25 As explained in Section 3, there are reasons to prefer unemployment as the main dependent variable.

?6By contrast, the spatial effect on employment is significant, as we will discuss in Section 6.2.

27Two notable exceptions are Sudrez Serrato and Wingender (2016) and Dupor and McCrory (2018).
While the former do not find significant spatial effects, the latter find them to be economically relevant.

BANCO DE ESPANA 15 DOCUMENTO DE TRABAJO N.° 1922
within a radius of R kilometers of municipality 7 in period t (2/4). Hence, our specification

becomes:
M M
* * R * * *
Yitth = P'Yitart >” Brytis-p + Yo ong2h_y tol +f + Gian (2)
j=0 j=0

where op; captures the spatial effect.

5.1 Results

Figure 6 reports the estimates of coefficients o,,; in equation (2), considering a radius of
R = 20 kilometers.*® The results show that the stimulus has a sizable impact on nearby
municipalities. An increase in 100 euros per capita in municipalities within a 20 km radius of
a given municipality reduces unemployment in the latter by 2%.?? This effect is persistent,
being more noticeable during the first 16 months, and becomes close to zero after that period.

When translating this percentage effect into jobs multipliers (Table 2), we observe that, at
the peak (reached in the 14th month), a fiscal stimulus of one million euros in municipalities
within a 20-kilometer radius of a given municipality reduces unemployment in the latter by
0.6 persons.*° The spatial effect is statistically significant for reasonable levels of confidence
throughout the response horizon.*! This spillover effect. represents about 20% of the direct
effect during the first 6 months and slightly above 10% during the rest of the first year. In
Figure 7, we represent graphically the jobs multipliers of the local and spillover effects.*?
We can see that both components follow a similar dynamic pattern during the first year and
a half. Spillovers, although smaller in size, are a non-negligible contribution to the overall
impact of the fiscal plan. The total combined effect reaches its peak at the 8th month, when
an initial fiscal stimulus of one million euros is associated with a decrease of unemployment
by 5.9 workers.

Our two-year spatial cumulative multiplier is 5.7, which is approximately 8.4% of the
local cumulative multiplier (which, recall, was 68.2). Hence, our estimated total cumulative
multiplier is 73.9, i.e. 100,000 euros of stimulus reduced unemployment by 0.62 job-years.

In the next section, we will put these numbers in perspective within previous literature.

?8Other radii are considered in the next subsection.

2° This is approximately 2.5 times larger than the local effect estimated in Section 5. Note that this is
reasonable: given that the stimulus is measured in per capita terms, and that there is in most cases more
population living in municipalities in a radius of 20 kilometers than in the municipality itself, 100 euros per
capita amounts to a larger total stimulus here than in the local effect from Section 5.

3°To convert our estimates to multipliers, we proceed as we did for the local effect—see footnote 23. Here
we use the average population in the 20-kilometer radius to pin down the total stimulus amount.

3!Tt remains significant at 95% during the first 6 months, and remains significantly different from zero
throughout the response horizon when considering confidence bands of 68%.

32Hence, the represented spillover effects are those from Table 2. The represented own effects are the ones
obtained from estimating equation 2. As will be discussed in Section 6.2, the graph would be very similar if
we considered the own effects obtained from equation 1 (shown in Table 1).

BANCO DE ESPANA 16 DOCUMENTO DE TRABAJO N.° 1922
5.2 Robustness

Local effect under alternative specifications. In addition to providing an estimation
of the spatial effects of the stimulus, the specification of equation 2 allows us to test the
robustness of the local-effect results. If the local effects are well identified, they should not
change when the shock of nearby municipalities is added to the estimating equation, i.e. the
coefficients 3;,,; in equation (1) and 4; ; in (2) should be similar. In Figure A5, we show that

this is indeed the case.

Alternative spatial distances. So far, we have focused on a radius of R = 20 kilometers.
In Figure A6, we consider alternative radii of 10 and 30 kilometers. The effect of the stimulus
in a radius of 30 kilometers is similar to that in the baseline of 20 kilometers. The effects
considering R = 10 follow a similar dynamic pattern but are considerably smaller in size, as

expected.°?

Alternative dynamic structure. We test the robustness of the spatial results to alterna-
tive dynamic structures. We consider the same specifications as when testing the robustness
of the direct effects (see Section 4.2). The results (Figure A7) with these alternative speci-

fications are very similar to those of the baseline.

Alternative sample selection. We test if the results are affected when outliers in the
dependent or independent variables are added. The results (Figure A8) show that the results
barely change.

Spatial effect on employment. We estimate the spatial effect on employment. The
results (Figure A9) show that there is a significant spatial effect on employment in the first
six months after the shock. In particular, our estimates imply that spending 100 euros per
capita in municipalities within a 20 km radius of a given municipality increases employment
in the latter by 1% in the peak, which is reached six months after the shock. The effect

becomes non-significant after the sixth month.

6 Discussion of Results

Ours results from the previous sections suggest significant jobs multipliers both at the local
and spatial levels. In particular, we find that the peak jobs multiplier for each million euros
spend is reached in the 8th month after the stimulus and is estimated at 5.9. This translates

into a cost per job of around 170,000 euros. This figure is close to Wilson (2012), which

33This is because the total shock increases with the radius—see footnote 29.

BANCO DE ESPANA 1 7 DOCUMENTO DE TRABAJO N.° 1922
studies the employment effect of the ARRA program in the US, finding a jobs multiplier
after a year of between 4.3 and 8.1, depending on the specification.**

Further to this, Chodorow-Reich (2019) summarizes recent work on local fiscal multipli-
ers. This comparison is made in terms of the cumulative multiplier. Chodorow-Reich (2019)
finds that this statistic varies within wide bounds, ranging from 3.8 job-years in Chodorow-
Reich, Feiveson, Liscow, and Woolston (2012) to 0.5 in Conley and Dupor (2013). The latter
report a cost of maintaining a job during a year of about 200,000 dollars. Our estimated
cumulative multiplier of 0.62 job-years sits at the bottom of this wide interval.

Our estimates of the jobs multiplier reflect both the cost of hiring workers to carry out
the public investment projects, but also other costs, e.g. the cost of materials. Given that
the Plan E was exclusively a local infrastructure stimulus, the composition of the labor-to-
capital mix may be different from that in the ARRA (which included, among others, health
care provisions for Medicaid expenditure). This difference may explain why our estimates
are in the lower range of those from the ARRA.

Our estimates of the employment effects of the Plan FE suggest that the stimulus helped
to prevent employment destruction during the recession. In particular, they imply that the
stimulus impeded a destruction of around one million job-months during two years, which
would be approximately equivalent to providing full-time employment to around 40,000
workers during that period.*°

While the employment effect is a very relevant dimension in analyzing the impact of the
stimulus, it does not reflect the whole economic impact of this policy. Such evaluation would
require an estimation of i) the effect on output and ii) an evaluation of the welfare effects in
the long-run. Both dimensions present important challenges. First, there is no availability
of data on municipal output that enables a direct estimation of this effect. Second, a long
run evaluation of the effects of public investment requires a different methodology that is
beyond the scope of this paper. However, we now show how we can find an approximate
calculation of the effect of the stimulus on output in the short run.

For this calculation, we follow Chodorow-Reich (2019), who shows how to map the em-
ployment multipliers into output multipliers. The main limitation of this approximation is
that it is valid only for the short run, since it implicitly assumes that capital (either private
or public) will not adjust, effectively measuring the output effect of the stimulus that is
channeled through changes in employment. Given that Plan E' was a stimulus with a special
focus on the installation of public capital, our calculations are plausible for the short run, but
are likely to represent a strict lower bound on longer periods, once capital becomes usable

for production.*©

34These numbers refer to OLS and IV specifications when the stimulus is measured as the ARRA announced
funds to be received by each state (see Table 5 in Wilson (2012)). The jobs multipliers of 4.3 and 8.1 are
associated to cost per job of 233,000 and 123,000 dollars, respectively.

%5The Plan E provided a total stimulus of 8,000 and 5,000 million euros in the first and second waves,
respectively. We use our total estimation of 73.9 jobs-months per million euros during two years.

°6Because of general equilibrium effects, the total effect of output through employment could be even
larger if we allow the marginal productivity of labor to be increasing in capital.

BANCO DE ESPANA 18 DOCUMENTO DE TRABAJO N.° 1922
Our calculation of the short-run output multiplier is 0.3-0.5, depending on the calibration

37 This magnitude is comparable to recent aggregate estimates of

of elasticity parameters.
the nation-wide output multiplier of public investment. For example, Alloza, Burriel, and
Pérez (2018) use exogenous time series variation to show that public investment is associated
to an impact multiplier (in the first quarter) of 0.3. This figures rises to 1.1 by the end of
the first year.

Our results suggest that estimates of the local and nationwide multipliers are not that
different. This observation contrasts with previous work that finds local multipliers to be sub-
stantially higher than nation-wide ones (see Nakamura and Steinsson (2014) and Chodorow-
Reich (2019)), while supports the view that the effect of sub-national stimuli like ARRA had
an impact on output similar to nation-wide estimates (see, for example Ramey (2018)). We
argue that the conceptual differences between local and nation-wide multipliers are likely to
be small in countries belonging to currency unions such as Ssain.

The reaction of monetary policy is argued to be the main source of differences between
local and nation-wide multipliers. At the aggregate level, an expansionary fiscal policy could
trigger a tightening in the stance of monetary policy. This reaction of monetary policy is
less likely to be found in local fiscal expansions such as the ARRA or the Plan E. Hence, in
the estimations of local fiscal multipliers, dubbed as open-economy multipliers (Nakamura
and Steinsson (2014)), the sub-national entities in fact act as small open economies.

This local-national differences have similarities to the Spanish context within its currency
area (the euro area). To the extent that monetary policy in the euro area is not exclusively
set according to Spanish conditions, aggregate fiscal policy in Spain is less likely to provoke
union-wide monetary policy responses.** Since neither idiosyncratic changes in the nation-
wide economic conditions of Spain, nor idiosyncratic differences in the relative economic
performance of its municipalities are likely to have a one-to-one translation to the response
of the currency-wide monetary policy, the Spanish setting may arguably be a context where
the differences between nation-wide and local multipliers are attenuated.

7 Evidence on the Transmission of the Stimulus

To understand the channel of transmission of the stimulus, here we explore potential het-
erogeneity in the responses due to economic and demographic factors and by classifying

municipalities according to their per capita budget and the education level of their mayors.

37Explanations of this calculation and further discussions on the imposed assumptions are detailed in the
Appendix.

38Farhi and Werning (2016) discuss the implications of currency unions for the size of fiscal multipliers in
a general equilibrium model.

BANCO DE ESPANA 19 DOCUMENTO DE TRABAJO N.° 1922
7.1 Economic Sector

We estimate a version of equation (2) where the dependent variable becomes unemployment
in a given economic sector: construction, industry, services, or agriculture.*? The results
are shown in Figure 8 for the local effect (left column) and the spatial effect (right column).
Given that the fiscal stimulus aimed at promoting public infrastructure, we would expect
the construction and, to some extent, the industry sector to be the ones more affected upon
impact. However, general equilibrium effects resulting from rising disposable incomes in a
given sector may spill over to the rest of the economy.

Starting with the dynamics of the construction sector, we observe a significant, con-
temporaneous, and large effect at the local level (larger than the benchmark results for
unemployment in all sectors) that slowly converges to zero over the considered horizon. In-
terestingly, when considering the spatial effect, we again observe a substantially large and
contemporaneous response of unemployment (more than twice as large as the benchmark
results from Figure 6).*°

The effect of unemployment on the industry sector grows over time during the first 8
months and is quantitatively large: an increase of 100 euros per capita reduces unemployment
in the industrial sector by around 1% after six months. The spatial effect is also quantitatively
relevant but does not response on impact and is more transitory.

The dynamics of the service sector are slightly different: at the local level, unemployment
does not significantly respond upon impact but only later. At the spatial level, the effect
builds slowly towards a magnitude similar to the benchmark spatial effects shown in Figure 6.

Unemployment in the agriculture sector does not respond significantly to the fiscal stim-
ulus, probably reflecting that just few projects directly involved workers from this sector.
At the local level, the only significant effect occurs upon impact. At the spatial level, the
response fluctuates around zero, possible reflecting the relatively low weight of agricultural
unemployment over total (around 3% in 2008).

In sum, we observe that both the construction and industrial sectors are the most re-
sponsive to the fiscal stimulus. However, we find evidence that this effect spills over later to

the broader economy.

7.2 Demographics

Given that the gender and age distributions may vary across economic sectors, we now
investigate how the fiscal stimulus affected different demographic groups. We estimate a
new version of equation (2) where the dependent variable reflects unemployment classified

by gender (male and female) and age (younger than 25, 25-45, and older than 45 years).

°°The classification of unemployment in different sectors is based on the activity from the last previous
employment. Those without previous experience are not assigned to any sector.

40 As explained above, a municipality receiving a stimulus aimed at developing local public infrastructure
may hire a construction company in a different municipality.

BANCO DE ESPANA 20 DOCUMENTO DE TRABAJO N.° 1922
Figure 9 displays the differential response of unemployment by gender, both at the local
(left column) and spatial levels (right column). At the local level, male unemployment
responds significantly upon impact. The effect becomes close to zero by the end of the
first year. Looking at the effect arising from stimulus in surrounding municipalities, we
also observe that male unemployment responds contemporaneously and in a more persistent
manner (the unemployment contraction associated to the stimulus is still significant after 16
months). Female unemployment, by contrast, does not respond contemporaneously to the
shock. This is consistent with the results from Figure 8, since the sectors that responded
more clearly upon impact (construction and industry) have relatively low female presence.
The effect at the local level is quantitatively large towards the end of the fist year, with female
unemployment lowering by more than 1% after a fiscal stimulus of 100 euros per capita. At
the spatial level, the effect of the stimulus on female unemployment is not significant.

We next analyze the dynamic response of unemployment by age (Figure A10). When
considering young unemployed persons (< 25 years), we do not observe a generally significant
response to the fiscal stimulus (neither at the local nor at the spatial level). The effect is
however quantitatively relevant when considering adults aged 25-45 and older adults (above
45 years). In those cases, we observe that the stimulus reduced unemployment both at the
local and the spatial levels.

We conclude that it is important to notice that the fiscal stimulus has a heterogeneous
impact both in quantitative and dynamic terms for different types of workers (male versus fe-
male, young versus old). These differences are likely tied to the performance of the economic
sectors more directly affected by the fiscal stimulus. We find that these sources of hetero-
geneity are particularly relevant for the design of policies that aim at fostering employment

in a gender-balanced way.

7.3 Size of Budget

The Plan E stimulus transferred the same nominal amounts (in per capita terms) to each
municipality. The stimulus, however, may represent different relative shocks to municipali-
ties, depending on their level of spending. Figure 10 displays the distribution of per capita
spending in municipal budgets in 2009.1 For the median municipality, the first wave of the
stimulus, which transferred around 178 euros per capita to each municipality, implies a fiscal
shock of around 30% of their budget. For the 5% municipalities with the highest budget, the
stimulus amounted to 10% of their annual spending. For the bottom 5%, this figure rises
above 80%—for some of them, the stimulus was higher than their annual budgeted spending.

We investigate how different the response to the stimulus was depending on the size of

the relative shocks. Hence, we estimate equation (2) for two different samples: one with per

“1The median spending per capita in 2009 municipal budgets is around 560 euros, with an average of 170
euros and a standard deviation of 1,275. The fact that not all municipalities run the same budget may reflect
political differences, income levels, or geographical factors (e.g. the cost of providing public infrastructure is
higher in towns with difficult access).

BANCO DE ESPANA 21 DOCUMENTO DE TRABAJO N.° 1922
capita spending below the 2009 median (which we label low spending municipalities) and
another for those with a per capita spending above the median (high spending).

Results from these estimations for the local and spatial effects (left and right panels, re-
spectively) are shown in Figure 11. Starting with the local effect, we observe a quantitatively
different response of unemployment for municipalities with low spending and those with large
spending. For the former, the stimulus triggers a significant contraction upon impact that
builds up until the end of the first year, with a 100-euro per capita stimulus estimated to
reduce unemployment by around 1.2%. For the latter, the effect is not significant.

When considering the spatial dimension, the effects are relatively similar for both types
of municipalities in terms of size, although the dynamics are quite different. Municipalities
with lower spending benefit from stimulus in surrounding municipalities during the first year.
In the case of municipalities with high level of spending, this beneficial effect occurs later
on, with a peak effect 14 months after the stimulus.

These results suggest that the relative size of the shock is an important factor driving
the results of the fiscal stimulus. When the stimulus represents a higher share of the budget,
it is more effective in reducing unemployment. However, it is important to note that these
heterogeneous effects may reflect other factors that cause the differential levels of spending

(e.g., political preferences, income levels, geographical factors).

7.4 Education of Mayors

There is evidence that more educated politicians implement better policies.42 In our con-
text, it was municipal governments who selected the projects. Hence, we expect that more
educated municipal governments come up with projects that are more effective in reducing
unemployment.

To test this hypothesis, we use data on the education of mayors (heads of the local
governments) and split the sample based on whether the mayor had (at least) a university
undergraduate degree or not.**“4 We then estimate equations (1) and (2) separately for the
two subsamples. The results (Figure 12) reveal no heterogeneous response by the educational
level of the mayor: the effect of the stimulus is remarkably similar in the two subsamples.

We can think of three possible reasons why we do not observe any differential effects by
the educational level of the government. First, while the education of political leaders may

be beneficial in other domains, it may not have an impact in our context of a fiscal stimulus.

42For example, Besley, Montalvo, and Reynal-Querol (2011) find that having leaders who are more highly
educated increases economic growth.

4398% of the municipalities in our sample belong to the first group, i.e. had a college-educated mayor in
2009.

441 Spain, power at the municipal level is shared by the mayor and a city council, but the mayor plays
a more influential role (for a detailed description of their duties, see Fujiwara and Sanz (2018)). Hence, we
focus on the education of the mayor. In any case, the results are very similar if we consider instead the
education of the city councilors.

BANCO DE ESPANA 22 DOCUMENTO DE TRABAJO N.° 1922
Second, it may be that the education of mayors does have an impact in our context but only
in the long run (e.g. more educated mayors choose more productivity-enhancing projects),
and hence we cannot observe the differential effects. Third, it may be that only postgraduate
education (masters or PhD) matters. In our sample, we only observe 2.5% of mayors with
postgraduate level and we do not have enough statistical power to estimate the effects in

such a small subsample.

8 Conclusion

During the last recession, governments across the world embarked in fiscal expansionary
programs to prevent job destruction. In Spain, the Plan E channeled almost 13 billion
euros (around 1.2% of GDP) to local authorities to execute public investment plans. The
institutional design of this program, along with the availability of detailed municipality and
monthly unemployment data, allows us to implement a difference-in-difference strategy to
identify the causal effect of the fiscal stimulus on unemployment.

Our estimate of the multiplier is comparable to those at the bottom range of similar
work. This suggests that local fiscal multipliers are not necessarily larger than nation-wide
(aggregate) multipliers. We have argued that the difference between both statistics is likely
to be attenuated in countries that belong to currency unions, as is the case in Spain.

Our work opens up avenues for future research. While we have focused on the effect
of the fiscal stimulus on unemployment, there are other relevant dimensions that should be
taken into consideration when evaluating the overall impact of the program. For example,
it would be interesting to see how the stimulus affected welfare in the medium and long run.

We leave this question for future research.

BANCO DE ESPANA 23 DOCUMENTO DE TRABAJO N.° 1922
References

Alloza, M., P. Burriel, and J. J. Pérez (2018). Fiscal policies in the euro area: revisiting

the size of spillovers. Banco de Espana Working Paper (1820).
Angrist, J. D., O. Jorda, and G. M. Kuersteiner (2018). Semiparametric estimates of

monetary policy effects: string theory revisited. Journal of Business & Economic
Statistics 36(3), 371-387.

Baxter, M. and R. G. King (1993). Fiscal policy in general equilibrium. American Eco-
nomic Review 83(3), 315-334.

Bentolila, S., P. Cahuc, J. J. Dolado, and T. Le Barbanchon (2012). Two-tier labour
markets in the Great Recession: France versus Spain. Economic Journal 122(562),
F155-F187.

Besley, T., J. G. Montalvo, and M. Reynal-Querol (2011). Do educated leaders matter?
Economic Journal 121(554), F205-227.

Blanchard, O. and R. Perotti (2002). An empirical characterization of the dynamic ef-
fects of changes in government spending and taxes on output. Quarterly Journal of
Economics 117(4), 1329-1368.

Carozzi, F. and L. Repetto (2017). Distributive politics inside the city? The political

economy of Spain’ s Plan E. Mimeo.

Chodorow-Reich, G. (2019). Geographic cross-sectional fiscal spending multipliers: What

have we learned? American Economic Journal: Economic Policy, forthcoming.

Chodorow-Reich, G., L. Feiveson, Z. Liscow, and W. G. Woolston (2012). Does state fiscal
relief during recessions increase employment? Evidence from the American Recovery

and Reinvestment Act. American Economic Journal: Economic Policy 4(3), 118-45.

Conley, T. G. and B. Dupor (2013). The American Recovery and Reinvestment Act: Solely

a government jobs program? Journal of monetary Economics 60(5), 535-549.

Diez, A. (2008). Clima social tras las elecciones del 24-M. https://elpais.com/elpais/
2008/11/22/actualidad/1227345420_850215. html.

Dupor, B. and P. B. McCrory (2018). A cup runneth over: Fiscal policy spillovers from
the 2009 Recovery Act. Economic Journal 128(611), 1476-1508.

Farhi, E. and I. Werning (2016). Fiscal multipliers: Liquidity traps and currency unions.
Volume 2 of Handbook of Macroeconomics, pp. 2417 — 2492. Elsevier.

Fujiwara, T. and C. Sanz (2018). Rank effects in bargaining: Evidence from government

formation. Review of Economic Studies forthcoming.

Huber, K. (2018). Disentangling the effects of a banking crisis: Evidence from German

firms and counties. American Economic Review 108(3), 868-98.

BANCO DE ESPANA 24 DOCUMENTO DE TRABAJO N.° 1922
Jorda, O. (2005). Estimation and inference of impulse responses by local projections.
American Economic Review 95(1), 161-182.

Jorda, O. and A. M. Taylor (2016). The time for austerity: estimating the average treat-
ment effect of fiscal policy. Economic Journal 126(590), 219-255.

Ministerio de Politica Territorial y Administracién Ptiblica (2010). Informe de Seguimiento
del Fondo Estatal de Inversién Local (FEIL). Secretaria General Técnica, Madrid,
Spain.

Ministerio de Politica Territorial y Administracién Publica (2011). Impacto Terrirtorial del
Fondo Estatal para el Empleo y la Sostenibilidad Local (FEESL). Secretaria General
Técnica, Madrid, Spain.

Montolio, D. (2018). The effects of local infrastructure investment on crime. Labour Eco-

nomics forthcoming.
Moretti, E. (2010). Local multipliers. American Economic Review 100(2), 373-77.

Nakamura, E. and J. Steinsson (2014). Fiscal stimulus in a monetary union: Evidence

from US regions. American Economic Review 104(3), 753-92.

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

Ramey, V. A. (2016). Macroeconomic shocks and their propagation. In Handbook of

Macroeconomics, Volume 2, pp. 71-162. Elsevier.

Ramey, V. A. (2018). Ten years after the financial crisis: What have we learned from
the renaissance in fiscal research? Technical report, Presented at the NBER “Global

Financial Crisis 10”.

Shoag, D. (2015). The impact of government spending shocks: Evidence on the multiplier

from state pension plan returns. Unpublished manuscript.

Suarez Serrato, J. C. and P. Wingender (2016). Estimating local fiscal multipliers. Un-

published manuscript.

Wilson, D. J. (2012). Fiscal spending jobs multipliers: Evidence from the 2009 American
Recovery and Reinvestment Act. American Economic Journal: Economic Policy 4(3),
251-82.

BANCO DE ESPANA 25 DOCUMENTO DE TRABAJO N.° 1922
Figures

Figure 1: Time distribution of starting dates

Density
.006 .008 01
l l l

.004
!

.002
!

o a=
01jan2009 01jul2009 01jan2010 01jul2010 01jan201
Starting date of project

Notes: Distribution of starting dates of projects. Bins are 10-day wide.

BANCO DE ESPANA 26 DOCUMENTO DE TRABAJO N.° 1922
Figure 2: Unemployment and fiscal stimuli

oO
S
iB 8
io)
— +
ip)
5 Q
D ©
Bo [Sd
284 Se
o 2
® =
3 ‘ E
& (oo) —
Na LO wn
r= as
fo)
ES E
n
oo —
row a)
SB
e Tit
>
= }
S
& +O
T T T T T T
2006m1 2008m1 2010m1 2012m1 2014m1 2016m1
Date
unemployment (left) __——— fiscal stimulus (right)

Notes: The blue line represent monthly unemployment between 2006m1 and 2015m12. The red line represents

the size of the fiscal stimulus.

BANCO DE ESPANA 27 DOCUMENTO DE TRABAJO N.° 1922
Figure 3: Effect of the stimulus on unemployment

| Al

ate

.00005

0

Difference in log(unemployment)
-.00005

-.0001
!

-.00015

-24 ~16 -8 0 8 16 24
Months relative to shock

Notes: Blue points represent the dynamic response of unemployment to a fiscal stimulus of one euro per
capita. Vertical lines are 95% confidence bands computed using standard errors clustered by municipality.

BANCO DE ESPANA 28 DOCUMENTO DE TRABAJO N.° 1922
Figure 4: Robustness: Alternative dynamic structure

T T T T T
-24 -16 -8 0 8 16
Months relative to shock

—e®— benchmark ——®— Olagsy, ~——® 6lags y;
—e— 12 lags y, —e— 6lagsx, —®— 24 lags x,

3
—_ Fike VAN ik AN AA T | | f
S | a WAG. :
oe
5S |
oo
5 l
0
S |
ro)
l

24

Notes: Blue points represent the dynamic response of unemployment to a fiscal stimulus of one euro per
capita, under different dynamic specifications. Vertical lines are 95% confidence bands computed using

standard errors clustered by municipality.

BANCO DE ESPANA 29 DOCUMENTO DE TRABAJO N.° 1922
Figure 5: Robustness: Alternative sample selection

55a ye ooo0s

fe)
-.0

-.0001

Difference in |

T T T
24 -16 -8 0 8 16 24
Months relative to shock

-.00015

——e®— benchmark ~—®—— with dep. var. outliers
—e— with 2009 indep. var. outliers ——e— with 2010 indep. var. outliers

Notes: Blue points represent the dynamic response of unemployment to a fiscal stimulus of one euro per
capita, for different samples. Vertical lines are 95% confidence bands computed using standard errors clus-
tered by municipality.

BANCO DE ESPANA 30 DOCUMENTO DE TRABAJO N.° 1922
Figure 6: Spatial effects (20km)

.0004

.0002

0

ANUNA NEAR Ne
TANT TrRTT NT

Difference in log(unemployment)

-.0002
!

24 ~16 -8 0 8 16 24
Months Relative to Shock

-.0004
!

Notes: Blue points represent the dynamic response of unemployment in a municipality to an increase in 1
euro per capita from the fiscal stimulus in surrounding municipalities (within 20km). Vertical lines are 95%
confidence bands computed using standard errors clustered by municipality.

BANCO DE ESPANA 31 DOCUMENTO DE TRABAJO N.° 1922
Figure 7: Total jobs multiplier effect

0

-2

Number of jobs per million euros
-4

0123 45 6 7 8 9 101112131415 1617 18 19 20 21 22 23 24
Time (months)

DE localeffect = ~~ spatial effect

Notes: Blue bars represent the total number of jobs created per million euros spent in the municipality.
Yellow bars represent the total number of jobs created per million euros spent in surrounding municipalities
(within 20km).

BANCO DE ESPANA 2 DOCUMENTO DE TRABAJO N.° 1922
Figure 8: Response of unemployment by sectors

Construction (local effect) Construction (spatial effect)

Hag il aici nif

-24 -16 8 -24 -16
~ Months relative to shock Months relative to shook

.0002
.0005
L

0001

0

0

-.0001

Difference in log(unemployment — construction)

Difference in log(unemployment - construction)

-.0002
-.0005
L

Industry (local effect) Industry (spatial effect)

te cathl dul ll |

.0002 .0004
L

0 0001
=

|,

a
a
a
a
—

ul

il

0

Difference in log(unemployment industry)
-.0001
==
Difference in log(unemployment industry)

Ane vn
. y -16 8 8 16 24 r Sa 16 8
Months relative to shock Months relative to shock
Services (local effect) Services (spatial effect)

.0002
L

.00005 = .0001

Wd cl, - sell

! meyer fl itt

ALLL LALcthL A alti Ju lh

Pam mi

24 -16 8 8 16 24 24 16

0

0

Difference in log(unemployment services)
~.0002
1

-.00015 ~-.0001 ~-.00005
Difference in log(unemployment services)

~.0004
L

Months relative to shock Months relative to shock
Agriculture (local effect) Agriculture (spatial effect)
WAM if ,
7. uh Ll Hal t tt TT 7 i - sgitttiigtth sites tu,
a, a.
7 oe nr er ‘te rr
Months relative to shock Months relative to shock

Notes: Blue points represent the dynamic response of unemployment in a municipality to an increase in 1 euro
per capita from the fiscal stimulus in the same municipality (local effect, in left column) and in surrounding
municipalities (spatial effect, in right column). Each row shows the response of registered unemployment
from different economic sectors. Vertical lines are 95% confidence bands computed using standard errors
clustered by municipality.

BANCO DE ESPANA 3 DOCUMENTO DE TRABAJO N.° 1922
Figure 9: Response of unemployment by gender

Males (local effect) Males (spatial effect)

.0001
L
.0004
L

0 00005
— :
—
Fe
=
—
=
=
_
—_
z
E

0 0002
L

—

l

t__,

=o

Po

—
—e—

-_—1

=

+

Ty

|,

a

l,

-.0002
L

Difference in log(male unemployment)

-.00015 ~-.0001 -.00005
1

—|

_—I

|

_—t

4

oe

ed

I

4

—__]

—s

—

fa
ad
————
Pdi

Difference in log(male unemployment)

-.0004
L

-24 -16 -8 0 8 16 24 -16 -8 0 8 16 24
Months relative to shock Months relative to shock

L
-.0006
L

1
nw
x

Females (local effect) Females (spatial effect)

-0001
L
-0006
L

-0004

=

=

—

=

—
—
}_
—_

=
——
I

lL

—
-—
+;

=

lL,
———
}_,
—E
—
—
} _
-—
=

Difference in log(female unemployment)

2 a wh LL. 1
Perey i |

Months relative to shock Months relative to shock

Notes: Blue points represent the dynamic response of unemployment in a municipality to an increase in
1 euro per capita from the fiscal stimulus in the same municipality (local effect, in left column) and in
surrounding municipalities (spatial effect, in right column). Upper and bottom rows show the response of
male and female registered unemployment, respectively. Vertical lines are 95% confidence bands computed
using standard errors clustered by municipality.

BANCO DE ESPANA 4 DOCUMENTO DE TRABAJO N.° 1922
Figure 10: Distribution of per capita spending in municipal budgets

DPT 1
0 1000 2000 3000 4000

Municipal expenditure (budget) per capita in 2009

0015
!

Density
.001
l

5.0e-04
l

Notes: Distribution of the ratio of spending to population from municipal budgets.

BANCO DE ESPANA 35 DOCUMENTO DE TRABAJO N.° 1922
Figure 11: Response of unemployment by spending per capita

Unemployment response (local effect) Unemployment response (spatial effect)

baba Kohat
a ie,

-0001
L
000.
L

.0002
L

0

0
1

-.0001
L
Pad
Pad

Difference in log(unemployment)
-.0002
L

Difference in log(unemployment)

g S |
o] eo
0 T T T T T oS T T T T T
-24 -16 -8 0 8 16 24 -24 -16 -8 0 8 16 24
Months relative to shock Months relative to shock
—e— lowspending ~— ~~ highspending —e— lowspending ~—= ~~ highspending

Notes: Points represent the dynamic response of unemployment in a municipality to an increase in 1 euro
per capita from the fiscal stimulus in the same municipality (local effect, in left column) and in surrounding
municipalities (spatial effect, in right column). Blue and yellow lines refer to spending per capita below
and above median, respectively. Vertical lines are 95% confidence bands computed using standard errors
clustered by municipality.

Figure 12: Response of unemployment by level of education of Mayors

Unemployment response (local effect) Unemployment response (spatial effect)

Le All HY

it

.0002

ite)
3
Ss 4
fo)

-0001
L

eee li, alta he
TG

-24 -16 -8 0 8 16 24

0

-.0001
1

Difference in log(unemployment)

-.0002
1
Difference in log(unemployment)
0
\

-.0005
L

-24 -16 -8 0 8 16 24
Months relative to shock Months relative to shock
—e— without college =~ withcollege —e— without college =~ with college

Notes: Points represent the dynamic response of unemployment in a municipality to an increase in 1 euro
per capita from the fiscal stimulus in the same municipality (local effect, in left column) and in surrounding
municipalities (spatial effect, in right column). Blue and yellow lines refer to level of education per capita
with and without college education, respectively. Vertical lines are 95% confidence bands computed using
standard errors clustered by municipality.

BANCO DE ESPANA 36 DOCUMENTO DE TRABAJO N.° 1922
Tables

Table 1: Jobs multipliers (employment creation per million euros), local effect

| impact 3m 6m 12m 24m peak
point estim. -0.88 -2.72 -3.6 -4.46 -3.77 -5.69
CI 68% [-1.6,-0.2] [-4.0,-1.5]  [-5.1,-2.2]  [-6.8,-2.2] [-5.6,-2.0] | [-7.2,-4.3]
CI 95% [-2.3,0.5]  [-5.2,-0.3]  [-6.5,-0.8] [-9.0,-0.1]  [-7.4,-0.2]  [-8.6,-3.0]

Notes: The first row show the estimated jobs multiplier in a municipality (number of jobs created per million
euros spent in the same municipality). Second and third rows display 68 and 95% confidence intervals,
respectively.

Table 2: Jobs multipliers (employment creation per million euros), spatial effect

| impact 3m 6m 12m 24m peak
point estim -0.24 -0.44 -0.53 -0.33 -0.25 -0.59
CI 68% [-0.32,-0.16] [-0.60,-0.28] — [-0.73,-0.34] [-0.56,-0.10] [-0.48,-0.02] — {-0.81,-0.37]
CI 95% [-0.39,-0.08] [-0.76,-0.12] [-0.92,-0.15) —{-0.77,0.12] —[-0.71,0.21] — {-1.02,-0.16]

Notes: The first row show the estimated jobs multiplier in a municipality (number of jobs created per
million euros spent in the surrounding municipalities within 20km). Second and third rows display 68 and
95% confidence intervals, respectively.

BANCO DE ESPANA 37 DOCUMENTO DE TRABAJO N.° 1922
Online Appendices

Appendix Figures

Figure Al: Initial and actual capital transfers

2500 5000 7500 10000
!

total capital transfers (millions euros)

0

2004 2006 2008 2010 2012 2014 2016

initial budget _ ———— actual budget

Notes: The blue line represents the sum of capital transfers received by all municipalities in Spain, as reflected

in the initial budget. The red line represents the same variable as reflected in the actual, executed budget.

BANCO DE ESPANA 38 DOCUMENTO DE TRABAJO N.° 1922
Per capita spending on Plan E
100 150 200
i 1 1

50
1

Figure A2: Plan E Spending

Panel A: First wave

100 150 200
1 4

Per capita spending on Plan E
50
i

20000 40000 60000 80000 400000
Population

Panel B: Second wave

20000 40000 60000 80000 400000
Population

Notes: Each point represents a Spanish municipality. Panel A shows per capita spending in euros as a

function of population size of the municipality in the first wave of Plan EF (2009). Panel B shows per capita

spending during the second wave (2010). Municipalities with more than 100,000 inhabitants are not shown

in the graphs.

BANCO DE ESPANA 39 DOCUMENTO DE TRABAJO N.° 1922
Figure A3: Geographical distribution of starting dates

(19apr2009,300ct2009]

(01apr2009, 19apr2009]
(15mar2009,01apr2009] ~
[10jan2009,15mar2009]

No data

Notes: Distribution of the average starting date of projects.

BANCO DE ESPANA 40 DOCUMENTO DE TRABAJO N.° 1922
Figure A4: Robustness: Effect on employment (local effect)

.00005 .0001
|

0

Difference in log(employment)

-.00005

-24 ~16 -8 0 8 16 24
Months relative to shock

Notes: Blue points represent the dynamic response of employment (Social Security affiliations) to a fiscal
stimulus of one euro per capita. Vertical lines are 95% confidence bands computed using standard errors
clustered by municipality.

BANCO DEESPANA 41 DOCUMENTO DE TRABAJO N.° 1922
Figure A5: Robustness: Local effect with and without spatial control

.00005
!

0
|

-.00005

Difference in log(unemployment)
-.0001
!

24 ~16 -8 0 8 16 24
Months relative to shock

-.00015

——e— local effect (benchmark specification)
~ direct effect (spatial specification)

Notes: Blue and yellow points represent the dynamic response of unemployment in a municipality to an
increase in one euro per capita from the fiscal stimulus in the same municipality estimated from equations 1
and 2, respectively. Vertical lines are 95% confidence bands computed using standard errors clustered by
municipality.

BANCO DE ESPANA 42 DOCUMENTO DE TRABAJO N.° 1922
Figure A6: Robustness: Alternative radius (spatial effects)

.0004

.0002

(Ne Alp Fi Al nf T
rd Y Vk KT) WV) iy

yi

\

Difference in log(unemployment)
-.0002 0
| |

-.0004
|

—24 -16 -8 0 8 16 24
Months relative to shock

—e— R=20 — R=10  —e*#— R=30

Notes: Points represent the dynamic response of unemployment in a municipality to an increase in 1 euro
per capita from the fiscal stimulus in the surrounding municipalities for different radii (10, 20, and 30 km).
Vertical lines are 95% confidence bands computed using standard errors clustered by municipality.

BANCO DE ESPANA 43 DOCUMENTO DE TRABAJO N.° 1922
Figure A7: Robustness: Alternative dynamic structure (spatial effects)

.0004

.0002

=
a,

\ Ee?
OLA
/
ak

Difference in log(unemployment)
-.0002 0
l |
“os
a

-.0004
!

24 ~16 -8 0 8 16 24
Months relative to shock

——e®— benchmark ——®— Olagsy, ~——®— 6lags y;
—e— 12 lags y, —e*— 6lagsx, —®— 24 lags x,

Notes: Blue points represent the dynamic response of employment in a municipality to an increase in 1
euro per capita from the fiscal stimulus in the surrounding municipalities (within 20km), under different
dynamic specifications. Vertical lines are 95% confidence bands computed using standard errors clustered
by municipality.

BANCO DE ESPANA 44 DOCUMENTO DE TRABAJO N.° 1922
Figure A8: Robustness: Alternative sample selection (spatial effects)

.0004

ent)
02

m

querer Bb

TY ;
YN TY VY aia

. y b

a
ST
H
=——

Difference in lo
-.0002
l
>
NS

-.0004

T T T T T T
-24 -16 -8 0 8 16 24
Months relative to shock

—e®— benchmark ~—®-—— with dep.var. outliers

—e— with 2009 indep. var. outliers ——e— with 2010 indep. var. outliers

Notes: Blue points represent the dynamic response of unemployment in a municipality to an increase in
1 euro per capita from the fiscal stimulus in the surrounding municipalities (within 20km), for different
samples. Vertical lines are 95% confidence bands computed using standard errors clustered by municipality.

BANCO DE ESPANA 45 DOCUMENTO DE TRABAJO N.° 1922
Figure A9: Robustness: Effect on employment (spatial effect)

.0002

.0001
!

0
|
—
—_
—_
—
4

Difference in log(employment)

-.0001
!

24 ~16 -8 0 8 16 24
Months relative to shock

Notes: Blue points represent the dynamic response of employment (Social Security affiliations) in a munici-
pality to an increase in 1 euro per capita from the fiscal stimulus in the surrounding municipalities (within

20km). Vertical lines are 95% confidence bands computed using standard errors clustered by municipality.

BANCO DE ESPANA 46 DOCUMENTO DE TRABAJO N.° 1922
Figure A10: Response of unemployment by age

< 25 years (local effect)

.0001 .0002
L

0

vit ad (hil, IM K

Difference in log(unemployment - youngs)

-.0003 -.0002 -.0001
L L L

ey

vy

willl

24 -16

-8
Months relative to shock.

16

25 — 45 years (local effect)

.00005 = .0001
L L

0

qT

UAL

Difference in log(unemployment - adults)

fl

i

|

-.00015 -.0001 -.00005
L L L

24 -16

Months relative to shock

> 45 years (local effect)

.00005

0

bald I

hatte tls In

i vir

-.00005
L

Difference in log(unemployment - old)
=.0001
fi

Ni

-.00015
L

24 -16

Notes: Blue points represent the dynamic response of unemployment in a municipality to an increase in
1 euro per capita from the fiscal stimulus in the same municipality (local effect, in left column) and in
surrounding municipalities (spatial effect, in right column).
response of registered unemployment amongst young workers (up to 25 years), adults (25-45 years) and
older workers (above 45), respectively. Vertical lines are 95% confidence bands computed using standard

-8
Months relative to shock

errors clustered by municipality.

BANCO DE ESPANA 47 DOCUMENTO DE TRABAJO N.° 1922

< 25 years (spatial effect)

001
L

0005
L

Mi in {t Ty

I i

Difference in log(unemployment - youngs)
-.0005 0
1

-.001
L

pint

Tet i

bea q

-24

T
-16

-8 8
Months relative to shock

16 24

25 — 45 years (spatial effect)

0004

0002
L

a)

|

—

ald

Difference in log(unemployment - adults)
-.0002 0
fi

-.0004
L

i

=

24

T
-16

-8
Months relative to shock

> 45 years (spatial effect)

0004
L

0002
L

dfhda ufldutel

(ll) dAlTA

Difference in log(unemployment - old)
-.0002 0
fi

ae eT

ry

-.0004
L

Upper, midd

-8
Months relative to shock

le, and bottom rows show the
Appendix: Mapping Jobs Multipliers to Output Multi-
pliers

Aggregate nation-wide government spending multipliers are often expressed in term of changes
in dollars of output per dollar of public spending.*® In contrast, local fiscal multipliers are
often expressed in terms of jobs or employment multipliers due to the lack of output at a ge-
ographically disaggregated level. In the case of Spain, although there is a limited availability
of annual regional measures of output, these do not exist at the municipal level.

In this section, we follow Chodorow-Reich (2019) to establish an approximate mapping
between the employment and output multipliers.

First, consider the following relationships (geographical subindices are dropped for clarity

purposes):

AYi+h = By AG;
AEF in = BAG,

where Y;,;, is output at period t+ h and E;,, denotes employment (number of workers
in the economy). 37 measures the output multiplier to a fiscal shock (change in dollars of
output per change of dollars in government spending), which is the object that we want to
recover. BF is the jobs multipliers, as estimated in sections 4 and 5.

For convenience, we define the percentage deviations of output and employment due

. . AY, AE .
to changes in government spending as yi_p = ye and e+, = —E, respectively. Let

n= age the change in spending as a share of total output. We can then express the

percentage deviation of employment as:

Y;

Ett
Ct+h = Br Bo
t

Next, consider a production function as:
Y, = A(H)E,)" (KP)* (KE)°,

where L; = HE; is the total number of labor (measured in hours) supplied in an economy in
time t. KP and K@ represent private and public capital, respectively. The most important
simplification of this analysis is to assume that a = ¢ = 0, that is, neither private or public
capital become available for the production in time t. While this is a plausible assumption

often encountered in macroeconomic models, it makes our calculation only informative for

45See for example Blanchard and Perotti (2002) for a seminal estimation of the aggregate output multipliers
in the US and Ramey (2018) for a recent review.

BANCO DE ESPANA 48 DOCUMENTO DE TRABAJO N.° 1922
the short run. For medium and long-run horizons it is sensible to assume that both public
and private capital will become usable. Under the common assumption that the marginal
product of labor is increasing in both types of capital, our calculation will recover a strict
lower bound of public investment.*°

Let hy = a denote the percentage variation in hours per worker and x = he the
elasticity of the hours per worker to total employment. We can then express the output

multiplier as a function of the employment multiplier:

Ut+h Ytth Ct+h Yy E
/ Gt Ctt+h Gt ( M e, f

In the case of the US, Chodorow-Reich (2019) argues that (1 —y)(1+ yx) & 1, concluding
that the output multiplier can be approximated by just considering the employment multi-
plier (in terms of cost per job) by output per worker (Y;/£;). For the case of Spain, we find
that for reasonable parameters of y and y, (1 — y)(1+ x) could be between 0.75 and 1.3.47
The ratio i was 19,639 and 18,342 in 2010 and 2012, respectively. Following ours previous
results, we consider a jobs multiplier of 5.89 (which represents a cost of about 170,000 euros
per job). Under this calibration, we find the output multiplier to be 0.3-0.5 in the short

run.

46The final effect of public investment on output depends crucially on the parameter ¢, which measures
the efficiency of public capital. Baxter and King (1993) find that public investment can crowd in private
investment in a calibrated general equilibrium model even for low parameters of ¢, delivering an aggregate
multiplier multiplier of above 4 in the long run. Alloza, Burriel, and Pérez (2018) use aggregate time series
variation to find that the output multiplier from public investment in Spain is 0.3 upon impact raising up
to 1.1. by the end of the first year, and around 2.3 by the second year.

47Labor share y is around 0.63 using standard calculations and Spanish national accounts data. Data for
total employment, nominal output, and hours per worker are obtained from the Spanish National Statistics
Institute (INE) for years 2008-2012.

BANCO DE ESPANA 49 DOCUMENTO DE TRABAJO N.° 1922
BANCO DE ESPANA PUBLICATIONS

WORKING PAPERS

1820
1821

1822

1823

1824
1825

1826

1827

1828

1829

1830

1831

1832

1833

1834

1835

1836
1837

1838
1839

1840
1841

1842

1843
1844

1845
1846

1847

1848
1849

MARIO ALLOZA, PABLO BURRIEL and JAVIER J. PEREZ: Fiscal policies in the euro area: revisiting the size of spillovers.
MARTA MARTINEZ-MATUTE and ALBERTO URTASUN: Uncertainty, firm heterogeneity and labour adjustments.
Evidence from European countries.

GABRIELE FIORENTINI, ALESSANDRO GALESI, GABRIEL PEREZ-QUIROS and ENRIQUE SENTANA: The rise and fall
of the natural interest rate.

ALBERTO MARTIN, ENRIQUE MORAL-BENITO and TOM SCHMITZ: The financial transmission of housing bubbles:
evidence from Spain.

DOMINIK THALER: Sovereign default, domestic banks and exclusion from international capital markets.

JORGE E. GALAN and JAVIER MENCIA: Empirical assessment of alternative structural methods for identifying cyclical
systemic risk in Europe.

ROBERTO BLANCO and NOELIA JIMENEZ: Credit allocation along the business cycle: evidence from the latest boom
bust credit cycle in Spain.

ISABEL ARGIMON: The relevance of currency-denomination for the cross-border effects of monetary policy.

SANDRA GARCIA-URIBE: The effects of tax changes on economic activity: a narrative approach to frequent anticipations.
MATIAS CABRERA, GERALD P. DWYER and MARIA J. NIETO: The G-20 regulatory agenda and bank risk.

JACOPO TIMINI and MARINA CONESA: Chinese exports and non-tariff measures: testing for heterogeneous effects at the
product level.

JAVIER ANDRES, JOSE E. BOSCA, JAVIER FERRI and CRISTINA FUENTES-ALBERO: Households’ balance sheets and
the effect of fiscal policy.

OSCAR ARCE, MIGUEL GARCIA-POSADA, SERGIO MAYORDOMO and STEVEN ONGENA: Adapting lending policies
when negative interest rates hit banks’ profits.

VICENTE SALAS, LUCIO SAN JUAN and JAVIER VALLES: Corporate cost and profit shares in the euro area and the US:
the same story?

MARTIN GONZALEZ-EIRAS and CARLOS SANZ: Women’s representation in politics: voter bias, party bias, and electoral
systems.

MONICA CORREA-LOPEZ and BEATRIZ DE BLAS: Faraway, so close! Technology diffusion and firm heterogeneity in the
medium term cycle of advanced economies.

JACOPO TIMINI: The margins of trade: market entry and sector spillovers, the case of Italy (1862-1918).

HENRIQUE S. BASSO and OMAR RACHEDI: The young, the old, and the government: demographics and fiscal
multipliers.

PAU ROLDAN and SONIA GILBUKH: Firm dynamics and pricing under customer capital accumulation.

GUILHERME BANDEIRA, JORDI CABALLE and EUGENIA VELLA: Should | stay or should | go? Austerity,
unemployment and migration.

ALESSIO MORO and OMAR RACHEDI: The changing structure of government consumption spending.

GERGELY GANICS, ATSUSHI INOUE and BARBARA ROSSI: Confidence intervals for bias and size distortion in IV
and local projections — lV models.

MARIA GIL, JAVIER J. PEREZ, A. JESUS SANCHEZ and ALBERTO URTASUN: Nowcasting private consumption:
traditional indicators, uncertainty measures, credit cards and some internet data.

MATIAS LAMAS and JAVIER MENCIA: What drives sovereign debt portfolios of banks in a crisis context?

MIGUEL ALMUNIA, POL ANTRAS, DAVID LOPEZ-RODRIGUEZ and EDUARDO MORALES: Venting out: exports during
a domestic slump.

LUCA FORNARO and FEDERICA ROMEI: The paradox of global thrift.

JUAN S. MORA-SANGUINETTI and MARTA MARTINEZ-MATUTE: An economic analysis of court fees: evidence from
the Spanish civil jurisdiction.

MIKEL BEDAYO, ANGEL ESTRADA and JESUS SAURINA: Bank capital, lending booms, and busts. Evidence from
Spain in the last 150 years.

DANIEL DEJUAN and CORINNA GHIRELLI: Policy uncertainty and investment in Spain.

CRISTINA BARCELO and ERNESTO VILLANUEVA: The risk of job loss, household formation and housing demand:
evidence from differences in severance payments.
1850
1851

1901

1902

1903

1904

1905

1906
1907

1908

1909

1910
1914

1912

1913

1914

1915
1916

1917

1918

1919
1920

1924

1922

FEDERICO TAGLIATI: Welfare effects of an in-kind transfer program: evidence from Mexico.

OSCAR ARCE, GALO NUNO, DOMINIK THALER and CARLOS THOMAS: A large central bank balance sheet? Floor vs
corridor systems in a New Keynesian environment.

EDUARDO GUTIERREZ and ENRIQUE MORAL-BENITO: Trade and credit: revisiting the evidence.

LAURENT CAVENAILE and PAU ROLDAN: Advertising, innovation and economic growth.

DESISLAVA C. ANDREEVA and MIGUEL GARCIA-POSADA: The impact of the ECB’s targeted long-term refinancing
operations on banks’ lending policies: the role of competition.

ANDREA ALBANESE, CORINNA GHIRELLI and MATTEO PICCHIO: Timed to say goodbye: does unemployment
benefit eligibility affect worker layoffs?

CORINNA GHIRELLI, MARIA GIL, JAVIER J. PEREZ and ALBERTO URTASUN: Measuring economic and economic
policy uncertainty, and their macroeconomic effects: the case of Spain.

CORINNA GHIRELLI, JAVIER J. PEREZ and ALBERTO URTASUN: A new economic policy uncertainty index for Spain.
SERGIO PUENTE and ELENA VOZMEDIANO: Minimum and bargained wages: pass-through and labor market
performance.

SERGIO MAYORDOMO and OMAR RACHEDI: The China syndrome affects banks: the credit supply channel of
foreign import competition.

MONICA CORREA-LOPEZ, MATIAS PACCE and KATHI SCHLEPPER: Exploring trend inflation dynamics in Euro Area
countries.

JAMES COSTAIN, ANTON NAKOV and BORJA PETIT: Monetary policy implications of state-dependent prices and wages.
JAMES CLOYNE, CLODOMIRO FERREIRA, MAREN FROEMEL and PAOLO SURICO: Monetary policy, corporate
finance and investment.

CHRISTIAN CASTRO and JORGE E. GALAN: Drivers of productivity in the Spanish banking sector: recent evidence.
SUSANA PARRAGA RODRIGUEZ: The effects of pension-related policies on household spending.

MAXIMO CAMACHO, MARIA DOLORES GADEA and ANA GOMEZ LOSCOS: A new approach to dating the reference
cycle.

LAURA HOSPIDO, LUC LAEVEN and ANA LAMO: The gender promotion gap: evidence from Central Banking.
PABLO AGUILAR, STEPHAN FAHR, EDDIE GERBA and SAMUEL HURTADO: Quest for robust optimal
macroprudential policy.

CARMEN BROTO and MATIAS LAMAS: Is market liquidity less resilient after the financial crisis? Evidence for US
treasuries.

ESTEBAN GARCIA-MIRALLES, NEZIH GUNER and ROBERTO RAMOS: The Spanish personal income tax:

facts and parametric estimates.

SAKI BIGIO, GALO NUNO and JUAN PASSADORE: A framework for debt-maturity management.

LUIS J. ALVAREZ, MARIA DOLORES GADEA and ANA GOMEZ-LOSCOS: Inflation interdependence in advanced
economies.

DIEGO BODAS, JUAN R. GARCIA LOPEZ, JUAN MURILLO ARIAS, MATIAS J. PACCE, TOMASA RODRIGO LOPEZ,
JUAN DE DIOS ROMERO PALOP, PEP RUIZ DE AGUIRRE, CAMILO A. ULLOA and HERIBERT VALERO LAPAZ:
Measuring retail trade using card transactional data.

MARIO ALLOZA and CARLOS SANZ: Jobs multipliers: evidence from a large fiscal stimulus in Spain.

Unidad de Servicios Auxiliares

BANCO DEESPANA Alcala, 48 - 28014 Madrid

: E-mail: publicaciones@bde.es
Eurosistema www.bde.es

File and source

File

jobs-multipliers-evidence-from-a-large-fiscal-stimulus-in-spain.pdf

Size

8,936,717 bytes

SHA-256

1eb1af9f0f8b930fbf6edc1ca73644efb37493a74532d3395489720c0712762b

Our copy

jobs-multipliers-evidence-from-a-large-fiscal-stimulus-in-spain.pdf

Original

doi.org