**1. Introduction**

The effect of an increase in governmen<sup>t</sup> spending is a central issue in macroeconomics. In this regard, different macroeconomic models have achieved contrasting conclusions about the response of private consumption to governmen<sup>t</sup> spending shocks (see, among others: Baxter and King 1993; Ambler and Paquet 1996; Linnemann and Schabert 2006; Forni et al. 2009; Leeper et al. 2010; Enders et al. 2011; Coenen et al. 2012; Corsetti et al. 2012; Kormilitsina and Zubairy 2018; Beidas-Strom and Lorusso 2019).

It is also well known that around the early 1980s, the transmission of fiscal policy shocks actually changed (see, for example: Fatás and Mihov 2001; Blanchard and Perotti 2002; Perotti 2005; Galí et al. 2007). Such a change is related to the increased asset market participation by households (Bilbiie et al. 2008). During the 1960s and the 1970s, a large fraction of households was prevented access to financial markets due to significant restrictions. Starting from the early 1980s, financial liberalisation widened private access to financial markets. In turn, such a change had an important influence on the private consumption response to governmen<sup>t</sup> spending shocks.

In this paper, we develop and estimate a Dynamic Stochastic General Equilibrium (DSGE) model, which includes two different components of governmen<sup>t</sup> spending, namely civilian and military expenditures. In line with the so-called "military Keynesianism" (see Pieroni et al. 2008), we assume

that spending decisions for these two different governmen<sup>t</sup> components are independent. According to the proponents of this view, defence expenditure satisfies two particular conditions: Firstly, it is financed independently of the other public spending categories (such as education and health). Secondly, decisions about the defence sector are taken from institutions that are independent of the other governmen<sup>t</sup> sectors (the Department of Defence in the U.S.).

Our DSGE model tries to explain the possible sources of crowding in/out effects in consumption observed in the data. To do so, we take into consideration heterogeneous households as in Galí et al. (2007) and Lorusso and Pieroni (2017). A share of households does not have access to the bond market and consumes their current disposable income at each date. On the other hand, a share of households has access to financial markets, smoothing their consumption in the desired way. Firms that produce differentiated goods decide on labour input and set prices according to the model of Calvo (1983). The fiscal policy authority purchases consumption goods, which are divided into spending for the military and non-military sectors, and raises lump-sum taxes and income taxes and issues nominal debt. Finally, our model encompasses a central bank, which sets its policy instrument, the nominal interest rate, by the rule of Taylor (1993).

We estimate our model with Bayesian techniques using U.S. data for two sub-samples: 1954:3–1979:2 (S1) and 1983:1–2008:2 (S2). This sample split allows us to analyse the changes in fiscal shocks before and after the potentially important changes to the financial markets mentioned above.

The main contribution of this paper with respect to previous literature is twofold. Firstly, we include the disaggregated components of civilian and military spending in a DSGE theoretical framework. This allows us to assess the effects of these two public spending components on several macroeconomic aggregates and, in particular, on private consumption. Secondly, we use the Bayesian approach to estimate the effects of fiscal policy shocks on the economy. This allows us to avoid the well-known shortcomings in the identification of military shocks that are associated with the neoclassical literature (Ramey and Shapiro 1999; Ramey 2011) based on the so-called "narrative approach".<sup>1</sup>

Our estimated results show that, in the U.S., the share of asset holders increased after the financial liberalisation that occurred in the early 1980s. Such an increase has important consequences on the effects of public spending shocks on the economy. In particular, we find that total government, non-military and military spending shocks affect the U.S. economy differently. An increase in total governmen<sup>t</sup> expenditure has a positive effect on output, but it induces a fall in private consumption (the so-called crowding-out effect). This occurs because the negative wealth effect generated by the increase in taxation leads both non-asset and asset-holders to increase their labour supply. Accordingly, the fall in the aggregate wage lowers households' disposable income, and in turn, private consumption decreases.

On the other hand, an increase in non-military spending induced a crowding-in effect on consumption in the pre-1980s period. Such an outcome occurred because the lower persistence of the non-military spending shock implies a lower wealth effect on asset holders, and subsequently, the shift in labour demand dominates the shift in labour supply. Accordingly, the real wage increases enough to raise aggregate consumption. Our results also indicate that military spending shocks have a less positive effect on output than civilian spending shocks in both sub-samples.

Finally, we analyse the role of monetary policy in the presence of several public spending shocks. We find that a higher nominal interest rate associated with a more aggressive monetary policy tends to strengthen household incentives to postpone consumption, inducing a negative effect on output.

The rest of the paper is structured as follows. The model is presented in Section 2. In Section 3, we describe the data used for our analysis, discuss the parameters of the model, and report the estimated

<sup>1</sup> For a detailed discussion about the criticism of the narrative approach, refer to Perotti (2005).

results. Section 4 presents the impulse response analysis of our estimated models, and Section 5 provides the robustness analysis. Finally, Section 6 concludes.

## **2. The Model**

In this section, we present our DSGE model, which is in line with the theoretical framework developed by Bilbiie et al. (2008).<sup>2</sup>
