This paper extends the standard New Keynesian DSGE model by considering that some agents may not be able to smooth consumption and may have consumption habits. Although both assumptions are not new in the literature, the joint consideration of these two features of consumption behavior strongly affects the results addressed by the recent literature, both theoretically and empirically.

Evidence on the existence of heterogeneous consumers has been first provided, nearly fifteen years ago, by Campbell and Mankiw (1989, 1990, 1991). According to them, only a fraction of households (savers) is able to plan consumption along with the standard Hall’s consumption function, while a relevant fraction of households (spenders) equates current consumption to current income period by period (violating the permanent income hypothesis).

The policy implications of the introduction of spenders in the model are important. Considering fiscal policy, if consumers not able to smooth consumption, the Barro- Ricardo equivalence does not hold. For this reason, savers are often referred to as Ricardian consumers and spenders as non-Ricardian consumers.

Recently, economists have instead focused on the effects of spenders on monetary policy by considering agents that consume according to a rule-of-thumb behavior within the New Keynesian theoretical apparatus. They find that the presence of rule-of-thumb consumers may overturn some of the conventional policy prescriptions addressed by the literature.

Galì et al. (2004), e.g., explore the Taylor rule properties when considering rule-of- thumb households and show that the Taylor principle 1) may be not a sufficient criterion for stability when there are many rule-of-thumb consumers; 2) becomes a sufficient but non necessary condition for stability when monetary policy is set according to a standard (feedback) Taylor rule. Instead, in the case of forward-looking interest rate rules, conditions for a unique equilibrium are somewhat different from the usual ones.

Amato and Laubach (2003) explore the optimal monetary rule when considering rule- of-thumb households and firms. By modeling consumers’ rule-of-thumb behavior as a consumption habit, households’ current decisions mimic past behavior of all agents (including optimizing agents). They show that, while the monetary policy implications of rule-of-thumb firms are minimal, the presence of rule-of-thumb consumers alters the determination of the optimal interest rate. As their fraction increases, higher inertial monetary policy is required.

Similar results are found by Di Bartolomeo and Rossi (2007), who however focus on the effectiveness of monetary policy. They find that, although an increase in consumers who cannot access to the financial markets (and thus cannot smooth consumption) reduces effects of interest rate policies via the consumption inter-temporal allocation (according to the permanent income effect), monetary policy becomes more effective as the degree of financial markets participation falls. In fact, after a change in the interest rate, spenders and savers revise their consumption plans in the same direction, because the fall of the interest rate supports the increase of current output also by affecting spenders’ consumption through higher real wages (staggered prices in fact imply a decline in the mark-up after an initial increase in economic activity; this allows real wages to increase, leads to a boom in rule-of-thumb consumption, generates inflation and improves the effectiveness of monetary policy).

By using a simplified version of Galì et al. (2004), Bilbiie (2005) and Di Bartolomeo and Rossi (2005) find that, for high fractions of rule-of-thumb (ROT) consumers, the interest rate increase becomes expansionary, thus showing that two-demand regimes can emerge (according to the “slope” of IS curve). On the empirical side, rule-of-thumb consumption has been considered consistent with the puzzling result of a weak or positive relationship between expected consumption growth and real interest rates (Ahmad, 2005, Bilbiie 2006, Canzoneri et al., 2006). The empirical relevance of the New Keynesian DSGE theoretical predictions is however still ambiguous, since its evaluation has been generally obtained by estimating reduced-form forward-looking IS curves, whose coefficients are only a convolution of the deep parameters.

The contribution of this paper is to extend the aforementioned literature both theoretically and empirically. We firstly derive an analytical closed-form solution of the model and study its stability regions, and then we evaluate the model by stochastic simulations, obtained from Bayesian parameters estimates for the G7 economies.

In order to derive a closed-form solution of the model, as in the standard New- Keynesian models, we do not consider capital accumulation. This also allows us to simply discuss the dynamic properties of the model. According to the fraction of spenders, we analytically discriminate between two demand regimes (i.e. two IS- curves), defined according to the response of the aggregate demand to nominal interest rate movements. We show that the possibility of a demand regime shift has a dramatic importance for the analysis of monetary policy effectiveness (discussed by Amato and Laubach, 2003; Di Bartolomeo and Rossi, 2005, 2007). We then show that the consideration of external habits reduces the probability of obtaining a regime shift in the demand schedule, as it increases the threshold fraction of spenders above which an inversion of the slope of the IS-curve is obtained.

The possibility of a demand regime shift has remarkable implications for the analysis of equilibrium determinacy, as discussed in Bilbiie (2005, 2006) and Di Bartolomeo and Rossi (2005). On this respect we show that, the unconventional results stressed by Galì et al. (2004) hold only if the relationship between the nominal interest rate and the aggregate demand is positive, i.e. when the IS-curve is positively sloped.

The second stage of the analysis focuses on the empirical evaluation of the theoretical predictions of the model. Our investigation aims to evaluate the empirical relevance of the regime inversion from a direct estimate of the structural parameters of the model. Moreover, our analysis aims to providing an assessment of the heterogeneous effects of monetary policy. The values of the structural parameters are not calibrated or fixed on the basis of previous evidence, as in the standard practice. Because of our strong empirical bearing, we estimate the structural coefficients employing quarterly data for the seven most industrialized economies (G7) for the 1963-2003 period. Differently from the common practice emerging in recent studies (see Smets and Wouters, 2003; Coenen and Straub, 2005), we consider country-level data separately in order to stress the cross-country heterogeneity. The complexity and nonlinearity of the resulting structure of the model suggests the implementation of a Bayesian Monte-Carlo Markov Chain estimation procedure (MCMC).