Realizar un cuadro comparativo entre los conceptos de enseñanza y
2.1 Condiciones intrapersonales del aprendizaje escolar.
The previous chapter considered the impact of ‘exogenous’ tax changes on the macroecon- omy. Equally important are the consequences of ‘endogenous’ movements in tax policy — those taken in response to macroeconomic conditions. However, identifying the de- terminants and effects of these tax changes poses a significant challenge, especially for the structure-free VAR methods pursued in the previous chapter. This chapter estimates the endogenous component of fiscal policy using a DSGE structural model with Bayesian methods. Previously estimated DSGE models of fiscal policy had to assume that tax shocks were latent variables, with the tax data based on National Accounts measures of revenues. As movements in aggregate revenues are a combination of the automatic stabilisers, discretionary endogenous policy reactions and exogenous shocks, these sepa- rate components are assumed to be unobserved. As discussed in Chapter 2, a motivation for the narrative approach is to try and measure the unobserved discretionary policy changes. The central innovation in this chapter is therefore to directly use the narrative measures of endogenous and exogenous tax changes in an estimated DSGE model to identify better the feedback from key macroeconomic variables to the fiscal instruments. Understanding the endogenous response of fiscal instruments is important for several reasons. First, policymakers (at least in the United States and the United Kingdom) have often engaged in countercyclical, debt-management and spending-driven tax policy. Understanding the macroeconomic consequences of these actions is of particular impor- tance to policymakers, especially in the current climate of fiscal retrenchment. Second, mainstream macroeconomic models can generate a range of theoretical predictions de-
Chapter 3 3.1 Introduction
pending on how fiscal instruments finance government spending shocks and respond to debt. This theoretical point is well known (see for example Baxter and King (1993)) but Leeper et al. (2010) point out that empirical research on how fiscal instruments react to debt, for example, is “scarce”. Third, endogenous movements in tax policy suggest an important interaction between monetary and fiscal policy.
The existing literature has attempted to estimate the economic effects of changes in fiscal policy using a DSGE model with Bayesian methods; for example Coenen and Straub (2005), Forni et al. (2009), Lopez-Salido and Rabanal (2008) or Ratto et al. (2009). However, much of this literature either considers limited feedback rules, for example only from debt, or focuses entirely on exogenous fiscal shocks. The most closely related work is by Traum and Yang (2009), Zubairy (2010) and Leeper et al. (2010), who incorporate rich policy rules with feedback mechanisms from debt and output. Zubairy (2010) and Traum and Yang (2009) also employ models with monetary and fiscal policy.1 Traum and Yang (2010) examine the interaction between passive and active monetary and fiscal policy rules in a DSGE model estimated using Bayesian methods. I will touch on this issue later.2
This chapter differs from previous work by making direct use of tax changes measured by the narrative approach in the estimation of a DSGE model. Consequently, particular attention is paid to the estimated endogenous feedback coefficients from debt, output and spending to the fiscal instruments.3 Integrating narrative measures into the DSGE model is non-trivial. First, narrative measures are likely to be measured with error, as discussed by Mertens and Ravn (2011a). Moreover, there are likely to be expectational errors associated with deviations from the endogenous tax policy rule, in addition to exogenous tax changes of the type identified by Romer and Romer (2010) and in Chapter 2. It is unclear that the expectational errors are separately identified from the narrative measured ‘exogenous’ tax changes and their measurement errors. I use new estimates by Mertens and Ravn (2011a) of the degree of measurement error in the Romer–Romer data to calibrate the measurement error in the DSGE model.
The model includes sticky prices and the non-fiscal elements resemble New Keynesian models such as that of Smets and Wouters (2003, 2007).4 The model is also based on recent work by Mertens and Ravn (2011b) who study the effect of anticipated versus unanticipated ‘exogenous’ shocks identified by Romer and Romer. To specifically focus on the endogenous policy reactions, a rich fiscal policy description is employed following Traum and Yang (2009), Zubairy (2010) and Leeper et al. (2010). So that the estimates
1
The wider empirical literature jointly considering monetary and fiscal policy also tends to lack a rich description of the feedback to fiscal instruments.
2In the literature on active/passive policy regimes, various papers are not based on estimation of a DSGE model; see for example Davig and Leeper (2006, 2011).
3Chapter 2 discusses that these are the main categories of feedback found in the narrative record. 4
The Smets–Wouters model does not include a rich fiscal policy specification for analysing the questions of interest here.
Chapter 3 3.1 Introduction
remain as comparable as possible to the existing literature, I focus on the United States. I therefore use the Romer–Romer narrative dataset in my estimation.
Having estimated the model’s parameters making direct use of the narrative tax measures, I focus on several aspects of the results. First, this chapter is particularly concerned with new estimates of the endogenous feedback from output, debt and spending to the fiscal instruments themselves. I find that the feedback from debt to government spending is weaker when the model is estimated including the narrative tax measures. This feedback is also weaker than estimated elsewhere in the literature. The feedback from debt to the model’s tax rates is also lower when the narrative measures are included. These results suggest that the conventional methods, which assume the tax changes are latent variables, overestimate the degree of feedback from debt to the fiscal policy instruments.
Second, properly accounting for the endogenous component of fiscal policy is im- portant for understanding the transmission of structural fiscal shocks. The effects of a structural shock to fiscal policy depend on the current and future feedback to the fiscal instruments. For example, will the tax cut be financed with debt and, over time, how will that debt be financed? I therefore examine the effect of fiscal shocks in my estimated model using the narrative tax changes. I find that the tax multipliers are higher than estimated elsewhere — for example in Leeper et al. (2010) or Zubairy (2010) — but that the government spending multiplier is lower than found by Zubairy (2010).5 The capital tax multiplier is also estimated to be significantly higher than would be obtained without incorporating the Romer–Romer narrative data in the estimation. In addition, I show that the estimated model implies that exogenous tax changes have a peak effect between estimates found by Romer and Romer (2010) and in the SVAR literature.
Third, I consider several variations of the model and examine which specification the data prefer. I exclude depreciation allowances, consider a flexible price model and, importantly, I consider a model without any endogenous feedback to the fiscal instru- ments. I call this the exogenous fiscal policy specification. I show that the data prefer the baseline model over these three specifications. If fiscal policy had behaved more like an exogenous process, monetary policy would have needed to violate the Taylor principle (so-called ‘passive’ monetary policy) to ensure stable debt dynamics.6 That the data prefer a model where fiscal policy endogenously responds to debt, output and spending is interesting and implies that movements in the price level (at least over the full sample) have not been required to satisfy the government’s budget constraint.
The remainder of this chapter is structured as follows. Section 3.2 presents the DSGE model and discusses how the narrative measures of tax changes can be integrated. Sec- tion 3.3 describes the estimation approach and presents the baseline parameter estimates. Section 3.4 conducts impulse response analysis for various structural fiscal shocks, ex-
5Zubairy (2010) includes the deep habits mechanism of Ravn et al. (2006). 6