VoxEU Column Frontiers of economic research Macroeconomic policy

What is the size of the multiplier? An estimate one can’t refuse

Few things divide the economics profession more than this question: How much economic activity does $1 of government spending generate? This column provides a new angle. Looking at local councils in Italy between 1990 and 1999, it examines variation in budgets due to the removal of funds by central government if mafia involvement is suspected. It finds that the fiscal multiplier starts at 1.4 and rises to 2.0.

How much more demand, output, and employment can we expect from expansionary fiscal stimulus? Conversely, how much macroeconomic pain can we expect from a contraction dictated by the need to keep public debt on a sustainable path? These questions have become most compelling during different phases of the recent global crisis (see Barro and Redlick 2009, Almunia et al. 2009, Aizenman and Kaur Pasricha 2010, and Auerbach and Gorodnichencko 2010 as just a few examples from this site).

Time-series studies typically suggest that the effects of public spending are modest overall. Moving away from linear models, multipliers are instead found to be large in specific circumstances, especially during episodes of financial and banking crises. In Corsetti et al. (2010), for instance, the point estimate for the multiplier of government spending is as high as 2 at times of financial crisis (see also Ilzetzki et al. 2009).

In a recent paper (Acconcia et al. 2011), we report a new empirical estimate of the multiplier, taking yet another approach. Instead of looking at the aggregate national data over time, we look at the output effects of spending on public works at the local level, focusing on differences across administrative units, i.e. provinces, within Italy.

We find that the multiplier is as high as 1.4 on impact, and reaches 2 over two years. Overall, the multiplier is significantly larger than 1.

Two features qualify these estimates.

  • First, by the way they are derived, our estimates are net of the influence of monetary policy and the general economic cycle of the economy.

In a nutshell, they can be interpreted as the empirical counterpart of the textbook Keynesian model, whereas the multiplier is typically introduced as the measure of the horizontal shift of the IS curve.

  • Second, a key institutional aspect of Italian data gives us an instrument which, we claim, is particularly effective to identify true multiplier effects of fiscal policy.
Mafia and multipliers

In order to investigate the size of the fiscal multiplier – i.e. the effect of fiscal policy on output – it is essential to identify exogenous and sizable variations in policy, say, government spending, which are not themselves explained by changes in output. Focusing on disaggregated data, we can actually identify this type of variations in spending in our Italian sample, by relying on the effects of a law mandating compulsory administration of local municipalities by the central government on evidence of mafia infiltration.

Specifically, since the early 1990s, a law expressively targeted to fight the increasing influence of different mafias in the public administration gives the central government the power to remove elected officials in a city council whenever the police has evidence that the council’s decisions are controlled (if only indirectly) by the mafias. Upon the removal of the city council, the central government appoints three non-elected, external commissioners in charge of ruling the city for a period of up to 18 months.

One of the first acts of the external administrators consists of suspending financial flows into local public work and investment projects – payments start again only after investigations verify that the firms carrying out these projects are not involved directly or indirectly in deals with the mafia. In our sample, the average growth rate of spending at provincial level indeed turns negative when a municipality is placed under compulsory administration. The average contraction is up to 20 percentage points.

Between 1990 and 1999 (our sample), we record 109 cases of city council dismissal. Aggregating them at provincial level, we obtain 43 observations. Of course, for the purpose of our studies, it is important that these episodes of large contractions of spending are not systematically correlated to the local economic cycle. Indeed they are determined by the outcome of police investigations often carried out outside the territory of the municipality. Moreover, the time between the emergence of the case for dismissing a city council and the implementation of the government decision to do so is very short.

Our full sample includes all 95 Italian provinces, with data spanning the 10 years from 1990 and 1999, that is, 950 observations. For all provinces, we have data on changes in GDP (value added) and changes in spending on public works. The 43 observations (from the 109 episodes) with mafia-motivated dismissals provide an ideal instrument in estimating the effect of variations in public spending on local GDP. By relying on such an instrument, we can address issues in “reverse causation”, by which it is the change in output that determines higher spending rather than vice-versa, as well as anticipation effects (points emphasised by Barro and Redlick 2011 and Ramey 2011 among others).

It is of course possible that the central government allocates resources for local investment spending with the goal of fostering growth in low-growth regions. To the extent that this is true, the areas of the country with the worst economic prospects would systematically receive the highest amount of resources for local investment. Without our instrument, the estimated multipliers may actually be too low. At the same time, investment projects may be announced some time before they are realised. Thus, anticipation effects, driving consumption and investment decisions on expectations of larger government spending in the near future, might lead to overestimation or underestimation of the multiplier, depending upon the impact of the announcement on the private sector behaviour. As a matter of fact, however, in our sample the multiplier estimated with OLS (hence, ignoring our new Mafia-related instrument) is very low.

In principle, it may be possible that police investigation determining the dismissal of a city council also affects local output through channels different from the multiplier effect related to the cut in public spending. For instance, the same police investigation affecting output through the public works channel may also determine (even if only temporarily) as a side effect the breakdown of other mafia activities, causing a direct negative impact on local economic activity. In this case, however, it is reasonable to expect, as a further outcome of an intense police investigation, that a number of people be reported to the judicial authority because of mafia association. This is the reason why we include such variables (together with many others) in our analysis, as controls for a possible transmission channel that does not work through the spending multiplier.

The advantage of assessing the multiplier of local spending

It is worth clarifying why our use of local information may present an advantage over time-series studies. To start with, the transmission of fiscal policy can be expected to vary extensively with the contemporaneous stance of monetary policy. The central bank can boost the output effect from a fiscal expansion if it accommodates it (i.e. it does not lean against the wind of the increase in demand); but by the same token monetary policy can also neutralise the aggregate macro effects of fiscal stimulus. After all, the popular textbook rendition of the Mundell-Fleming model, for instance, has popularised the notion that fiscal policy is effective under a fixed-exchange rate (since in this case the monetary policy automatically accommodate fiscal policy) – and ineffective under a flexible exchange rate.

The problem concerns not only monetary policy but also budget policies. Different governments may adopt different budget-consolidation measures over time. A temporary increase in spending today may be offset over the years by more taxes, less spending, or some combination of the two. It is not difficult to see that the short-run multiplier may vary substantially, depending on expectations of future budget consolidation policy. Even more so if the current stimulus generates doubts about debt sustainability, as to create tensions in the public debt market, and raise risk premia (with spillover effects on the private debt market) – see Corsetti et al. 2010. Once again, empirical models which do not account for budget consolidation policies may induce a bias in the estimate, due to the omission of debt/deficit variables.

By focusing on the cross-sectional variation in public investment spending across Italian provinces, our study addresses at least part of these issues. We control for both cyclical movements and policy impulses that are common to all provinces – hence our estimates are net of the common national cycle as well as of the common monetary stance. Moreover, since this component of spending is allocated by the central government without consequences for local taxation, the influence of budget adjustment at the local level is less of an issue.


In addition to our study, other contributions have focused on evidence across states/regions/provinces, exploiting institutional information to estimate the multiplier – see for example Nakamura and Steinsson (2010), Fishback and Kachanovskaya (2010), Serrato and Wingender (2010), and Clemens and Miran (2010).

Remarkably, the estimates of the multiplier we find across provinces in Italy – 1.4 on impact and 2 dynamically – are close to the results based on cross-sectional variations in spending across US states (see e.g. Serrato and Wingender 2010).

The message from this new strand of empirical literature is thus straightforward. Fiscal expansions and contractions do affect economic activity above and beyond their size. With fiscal stimulus or contraction in place, demand and output can, of course, move in response to more than fiscal policy, say, monetary policy, external shocks etc. But the pure fiscal effects appear to be strikingly in line with conventional wisdom.


Acconcia, A., G Corsetti and S. Simonelli, (2011) "Mafia and Public Spending: Evidence on the Fiscal Multiplier from a Quasi Experiment" CEPR Discussion Paper 8305.

Aizenman, Joshua and Gurnain Kaur Pasricha (2010), “The net fiscal expenditure stimulus in the US 2008-2009: Less than what you might think”, VoxEU.org, 3 March.

Almunia, Miguel, Agustín S Bénétrix, Barry Eichengreen, Kevin H O’Rourke, Gisela Rua (2009), “The effectiveness of fiscal and monetary stimulus in depressions”, VoxEU.org, 18 November.

Auerbach, Alan J and Yuriy Gorodnichenko (2010), “Measuring the output responses to fiscal policy”, VoxEU.org, 3 September.

Barro, RJ and CJ Redlick (2009), “Design and effectiveness of fiscal-stimulus programmes”, VoxEU.org, 30 October.

Barro, R. J., and C. J. Redlick (2011). “Macroeconomic Effects from Government Purchases and Taxes”, The Quarterly Journal of Economics CXXVI(1): 51-102.

Clemens, J, and S Miran (2010), “The Effects of State Budget Cuts on Employment and Income”, Working Paper, Harvard University.

Corsetti, G, A Meier and GJ Muller (2010). “What Determines Government Spending Multipliers?”, mimeo, Cambridge University.

Fisher, JDM and R Peters (2010), “Using Stock Returns to Identify Government Spending Shocks”, Economic Journal, 120(544):414-436.

Ilzetzki E, EG Mendoza, CA Vegh  (2009), “How big are fiscal multipliers? New evidence from new data”, VoxEU.org, 1 October.

Nakamura, E and J Steinsson (2010). “Fiscal Stimulus in a Monetary Union: Evidence from U.S. Regions”, mimeo, Columbia University.

Ramey, VA (2011), “Identifying Government Spending Shocks: It’s All in the Timing”, The Quarterly Journal of Economics, CXXVI(1):1-50.

Serrato, JCS and P Wingender (2010), “Estimating Local Fiscal Multipliers”, Working Paper, University of California at Berkeley. 

9,344 Reads