Structural Reforms in DSGE Models : A Plead for Sensitivity Analysis

The evaluation of fiscal and structural reforms has become not only a standard but indispensable exercise in the DSGE literature and in the policy-making publications and reports. Institutions such as the IMF, the European Commission, the OECD, the ECB, and many central banks have now developed and refined their own tools and are capable of conducting such analyses in different contexts. The effects of structural reforms have been documented by D'Auria et al. (2009) for EU member states and for Italy by Annicchiarico et al. (2013) both in the R&D version of the Quest III model. The IMF or the OECD have also conducted their own evaluations for Europe (Bayoumi et al., 2004; Everaert and Schule, 2006,2008; Cacciatore et al., 2012). Fiscal reforms or consolidation have also been assessed through DSGE models. In the European context some work were conducted on the Quest III model (Vogel, 2012). Coenen et al. (2008) investigate labor tax reforms in the New Area Wide Model (NAWM). Clinton et al. (2011) provide similar insights in the case of an international model (GIMF). Coenen et al. (2012) give an extensive review of the size of fiscal multipliers in the main institutional models. The recurrence and the systematic use of DSGEs today therefore raises the question of their actual capabilities. Whereas their qualitative behaviours have largely improved and now properly describe economic data, their quantitative accuracy is still debated among economists (see for instance Schorfheide (2011) for a summary of current DSGE weaknesses). We use the two country DSGE model of the Euro area MELEZE developed at Insee to shed a new light on two standard exercises: structural and fiscal reforms evaluations. The main features of the model compare with standard tools developed in international institutions and central banks: nominal and wage rigidities, capital adjustment cost, and both Ricardian and non-Ricardian consumers. We study the dependency of fiscal and structural simulations' results to various specifications in our DSGE model. Within a range of feasible calibrations for the elasticities in the utility function, the share of non Ricardian consumers, and among other sensitivity tests, the analysis focuses on short and long term multipliers of both fiscal and structural reforms. We also rank policy schemes based on welfare analyses along the transitional paths. The model MELEZE used in this paper features the standard modeling choices of the two country monetary union literature. The core of the model for each country is inspired by Christiano et al. (2005) and Smets and Wouter (2003, 2005, 2007): firms and consumers maximize their objective (utility or profit) by interacting on the goods, labor and capital markets with both prices and wages rigidities introducing neo-Keynesian features in the model à la Erceg et al. (2000). The model also integrates risk free assets to ensure an intertemporal trade-off and real rigidities on the capital market. In addition, our model builds on academic works studying monetary and fiscal policies in monetary unions Gali and Monacelli (2008), Benigno (2004) by introducing capital markets. We also introduce non Ricardian households as advocated by Mankiw (2000), a feature which is crucial for the reaction of private consumption to public spending (Gali et al. 2007), and therefore a priori crucial to the size of fiscal multipliers. We compare this mechanism with Edgeworth complementarity as advocated by Fève and Sahuc (2013). Moreover, we introduce in our model public and private debts exchanged on a union wide financial market both at steady state and out of equilibrium. Holding debt or asset is motivated by agents' preferences for the present and comes at a financial intermediation cost embodied through a debt elastic premium. We explicit and micro-found this financial intermediation service by introducing a financial intermediation sector. Beyond public debt, the government uses public spending to stimulate and monitor economic activity. It can also exogenously modify its fiscal policy along different axes: lump-sum transfers and taxes on consumption, labor, capital income or dividends. As detailed below, we depart from traditional budget rules behaviors used in the literature, and derive a forward-looking optimizing behavior for the government. All these modeling elements are generally embedded in large scale models developed in central banks and international institutions among which are GEM at the IMF (Bayoumi et al., 2004), NAWM at the ECB (Coenen et al., 2008) or in open economy EAGLE (Gomes et al., 2012), QUEST III at the European Commission (Ratto et al., 2009) and its R&D version (Roeger et al., 2008). Whereas these models sometimes also consider both tradable and non-tradable goods, heterogeneous agents on the labor market, or endogenous growth, we choose to simplify our model and do not consider these additions. The outcome is a model tractable enough to be fully linearized by hand. We are also able to solve for the steady state for the real variables in levels and carefully account for all the steady state restrictions imposed on the parameters of the model. We replicate three different settings: France against the rest of the Eurozone, Italy against the rest of the Eurozone, and a symmetric calibration for the Euro area as a closed economy. In a first section, we study the long-term impact of mark-up reforms in both the labor and goods markets. Even in the absence of entry costs, wage bargaining and an endogenous determination of the number of firms as in Blanchard and Giavazzi (2003), our results compare with stylized facts obtained in their model. Moreover and numerically, reforms simulation as conducted in Everaert and Schule (2006) indicates that the absence of additional rigidities and of a distinction between tradable and non-tradable goods may overestimate the long-term gains from pro-competitive reforms. More importantly, even though the stylized facts behind such reforms are robust, they increase output level at steady state, their quantification is uncertain. Within a range of feasible calibrations for the elasticities in the utility function, the effect of a structural reform can be magnified threefold. Similarly, the introduction of non Ricardian agents amplifies the gains from deregulation up to a doubling factor. In a second section, we study the effect of temporary or permanent fiscal reforms. We simulate increases in public spending, transfers or decreases in various tax rates calibrated to 1% of pre-stimulus output. The resulting fiscal multipliers are compared to the main existing DSGE models based on the results provided in Coenen et al. (2012), and to the French macroeconometric model Mésange developed at Insee (Klein and Simon, 2010). We find that our model gives comparable multipliers for temporary shocks but highlight that these measures of the fiscal multipliers crucially depend on their timing and the way both fiscal and monetary authorities commit or react to the stimulus. In particular, the modeling of government spending, usually introduced through an ad hoc spending rule, can imply fiscal multipliers larger or smaller than one. We compare these results with an alternative modeling of governments' behavior. Actually, we depart from ad hoc fiscal or budget rules traditionally introduced in quantitative models to endogenise public spending and tax rates to ensure governments' solvency (Bayoumi et al., 2004; Coenen et al., 2008; Ratto et al. 2009 ; Corsetti et al., 2009). We consider governments that maximize their stream of spending in a forward-looking way, closely equivalent to a Euler equation for households. In the end, public spending fiscal multipliers can range from 0.7 to 1.3 depending on the specification of the governments' spending rule and of the monetary environment. Cuts on distorting tax rates provides lower multipliers, that turn out to be even negative in the absence of government commitment for cuts in corporate income taxes and labor income taxes. Coordination across countries leads to increased fiscal multipliers. In response to permanent spending shocks financed though lump-sum transfers, our model provides weaker long-term multipliers yet comparable to Coenen et al. (2012) results. This weaker response stems from the negative wealth effect implied by the necessary financing fall in transfers. In all, our results raise questions on the ability for current quantitative DSGE models to provide accurate quantitative estimates for economic policies. In the conduct of policy analysis, one should therefore be very cautious to properly assess the dependency of the results to the specification of the model, and provide detailed sensitivity tests. Ongoing developments to be included in this paper include: a. Studying the transitional dynamic of structural reforms b. Ranking policy schemes based on welfare analyses along the transitional paths c. Stronger justification of the government’s behavior by the introduction of government spending in households’ utility function.