Growth forecast errors and government investment and consumption multipliers

Abstract

We compare government investment and consumption multipliers in developed economies during the initial years of the ongoing fiscal consolidation. We find that, in countries with high public debt, the investment multiplier is likely to be higher than what has been assumed by policy-makers and higher than the consumption multiplier. This leads to the conclusion that the consolidation should be accompanied by increased public investment.

Keywords:

• Fiscal consolidation
• fiscal multiplier
• public consumption
• public investment

JEL Classification:

• E62

Acknowledgements

The author would like to thank Sabrina Auci, Emanuele Brancati, Giovanni Callegari, Maddalena Cavicchioli, Germana Corrado, Gerardo Manzo, Paolo Paesani, Sabina Silajdzic, Alberto Zazzaro and other participants at the 54th Annual Conference of the Italian Economic Association and the 12th Researcher’s Club of the National Bank of the Republic of Macedonia, for very useful suggestions, and three anonymous referees for comments on earlier version of the paper.

Notes

No potential conflict of interest was reported by the authors.

1 If the two outliers are excluded (Bulgaria and Romania, where consumption has fallen by 9% of GDP), the average increase in public consumption is even higher, 0.6%.

2 As Perotti (2004) states, an early, though not necessarily the earliest mention of the Golden Rule is in Musgrave (1939).

3 We take 2010 as the starting year of the fiscal consolidation. This is in accordance with Blanchard and Leigh (2013).

4 Since the forecasts from the models are a result of many different factors, it is not entirely correct to speak about certain values of multipliers assumed in the models. We will, nevertheless, use this word, for ease of exposition.

5 The third component of public spending, the public transfers, are excluded from the analysis, due to data unavailability.

6 We take the gross debt, instead of the net, since the latter is available for fewer countries.

7 The bootstrapping exercise was done using 3000 replications. Higher number of replications gave similar results. The seed used for the simulation in Stata was 26011982, the date of birth of the author.

8 By including additional controls, we also, in a certain way, control for possible errors in the forecasts regarding the effects of the other variables on the GDP.

9 The spring forecasts from the European Commission are used EC (2011, 2012a), and the April editions of the Consensus Economics forecasts for G7 and Western Europe and Eastern Europe (ConsensusEconomics 2011a, b, 2012a, b).

10 The seed that was used for generating the random samples in Stata is 26011982.

11 The elements of which do not appear in the model are set to zero.

12 The BMA analysis has been implemented in R, using the BMS library, developed by Feldkircher and Zeugner (2009).

13 The index covers three aspects of the infrastructure – transport, telephony and energy. It covers around 140 countries. The highest possible value of the index is 7, the lowest 1. The average value of the index in 2010–11 for the whole world is 4.3. However, as only Italy has a value lower than this average, we choose 5 as the cut-off point for high vs. low public capital. The US, the lowest ranked high-capital country has a value of 5.7 and is ranked 24th in the world. Ireland, the highest ranked low-capital country, has a value of 4.6 and is ranked 53rd.

14 Bertola and Drazen (1993) also develop a model in which the relationship between fiscal policy (government consumption as a share of GDP) and private consumption (as a share of GDP) is non-linear, depending on the level of debt. However, in their model the relationship is negative when the debt is low and positive when the debt is high.

15 There is a third channel through which non-Keynesian effects can arise, as Alesina and Ardagna (2010) note, thorugh the labour supply. This channel is more applicable to longer horizons, so we leave it aside.