Expectations and Fiscal Policy: How Foresight Affects Policy Transmission

This thesis studies how expectations a ect scal policy transmission. It revisits the\nempirical evidence on the domestic and cross-border e ects of fiscal policy at the light\nof scal foresight.\nFiscal measures are the result of a complex decision process, entailing long lags be-\ntween the moment when the decision is taken and when it is e ectively implemented. It\nfollows that scal policy can be easily anticipated, with agents modifying their actions\nwhen they receive signals about changes in policy, well before the actual implementation.\nThis anticipation poses non-trivial challenges for the empirical analysis. It implies that\neconomic variables may move well before the policy shocks estimated in macroeconomic\nmodels take place. Overlooking the e ects of anticipation can be a major problem in VAR\nmodels. The parsimony required to conserve degrees of freedom makes the information\ncarried by the few endogenous variables contained in the model much smaller than the\ninformation possessed by private agents. This leads to a problem of non-fundamentalness\nthat might bias estimation results.\nThe predictability of scal actions is likely to in uence their cross-border e ects\nas well. National scal policies can a ect other countries through a variety of potential\nchannels. First, an increase in public spending could fall on foreign products, stimulating\ndirectly the foreign economy. Second, an expansionary policy could stimulate domestic\neconomic activity, leading to more imports from other countries and thus helping to\nstimulate foreign activity as well. Third, if the scal expansion results in higher public\ndebt, it could push long-term interest rates up in nancial partners, crowding out private\ninvestment. Government's nances could even become unsustainable with the risk of a\nsystemic crisis.\nThe innovative contribution of this thesis is to study scal spillovers and scal an-\nticipation together. The aim is to estimate how national scal measures a ect foreign\neconomies through an approach that allows to solve the problem of non-fundamentalness\nin VAR analysis, while at the same time assessing the role of expectations for the inter-\nnational transmission of scal policy. The questions of interest that this analysis aims at\nanswering are the following: are policy shocks estimated in standard scal VAR models\nreally unanticipated? How does anticipation in uence domestic and cross-border e ects?\nWhat are the implications for real economic activity of a change in forecasts about policy\nactions? Should governments consider the implications of their claims and actions for\nthe formation of expectations? Should national scal policies be coordinated at a supra-\nnational level? These questions have important policy implications that encouraged me\nto delve into these issues further and study them in depth.\nThe thesis comprises three chapters. The rst chapter provides a survey of the VAR\nmodels used in the empirical literature. The second and third chapters assess from a\n1\ndi erent perspective and with a di erent focus how the anticipation of future policy\nregimes changes the transmission of scal shocks.\nThe rst chapter reviews how the econometric model adopted in this thesis -the vector\nautoregression (VAR) - has been used in the empirical literature on scal policy. First, it\nillustrates the basic characteristics of the VAR approach, describing the reasons behind\nits widespread use in applied macroeconomic research. Second, it analyzes the challenges\nthat the VAR approach poses. Standard VAR models need to be identi ed in order to\nperform structural analysis, giving economic meaning to the shocks estimated in these\nsystems. This chapter reviews the di erent approaches proposed in the literature to solve\nthe problem, classifying them in four main categories. Standard VAR models are also low-\ndimensional models. The parsimony is imposed in order to conserve degrees of freedom\nand implies that VAR models can carry only a limited amount of information. The\nmisalignment of the information space between private agents and econometricians may\nlead to a problem of non-fundamentalness, biasing the estimation of policy innovations.\nDiscussing the solutions to this problem, I describe some innovative speci cations of\nthe basic approach which address this issue by taking advantage of the improvements\nin computing power and data availability (GVAR, FAVAR, Panel VAR). An ulterior\nlimitation of standard VAR models is linearity. In this regard, this chapter outlines\na further enhancement which allows for time-varying relationships among endogenous\nvariable (TVC-VAR). Finally, the rst chapter focuses on reviewing how VAR models\nhave been used to study the international transmission of scal shocks, addressing the\nissue of cross-border e ects.\nThe second chapter estimates the impact of expectations on scal spillovers. More\nprecisely, the analysis studies how the anticipation of the future scal stance a ects\nthe international transmission of scal measures in the US. To address this challenge use a set of two-country Bayesian VAR models. In each model, the US is the domes-\ntic economy, considering its leading role in the global economy. As foreign countries,\nI include Canada, France, Germany and UK because they represent the lion share of\nUS foreign trade. The innovative feature of this contribution is to apply an identi -\ncation strategy which di erentiates an unanticipated or surprise shock from a foresight\nor news shock. The former represents a discretionary increase in government spending\nthat was not foreseen by agents. The latter represents news received by agents which\na ect their expectations about prospective policy actions. The Philadelphia FED Survey\nof Professional Forecasters provides the data used to construct the indicators of scal\nforecast. This approach has a twofold advantage: it allows to address the problem of\nnon-fundamentalness, as well as to assess the international repercussions of both surprise\nand foresight shocks. Results show the importance of expectations for the international\ntransmission of US scal policies. A surprise scal stimulus has negligible cross-border\ne ects when it is associated with expectations of spending reversals. Foresight shocks,\non the contrary, are associated with expectations of increasing government spending and\nyield positive spillovers, despite no expansionary action is taken. The sign and magnitude of these external e ects are country sensitive, suggesting an active role of country-speci c\nfactors in a ecting the international transmission of US scal policies. These ndings\nprovide novel evidence in support of the hypothesis that foresight alters scal policy\ne ects on a national and international level.\nThe third chapter provides further evidence in support of the hypothesis that scal\npolicy is largely anticipated and its e ects depend on expectations. The analysis draws\non two-country VAR models between major European economies and applies the same\nidenti cation approach adopted in the second chapter, using this time the o cial fore-\ncasts of the European Commission to construct the indicators of scal forecast. Data\nrefer to Italy, France and Germany over the period 1971-2011. This third contribu-\ntion rst documents the forecasting accuracy of the European Commission forecasts and\nshows that they help address the problem of non-fundamentalness in scal VAR models.\nThen, it identi es surprise and foresight shocks through a recursive ordering in which\nthe realized policy does not react within the year to innovations in any other variable\nin the system. The expected policy, on the contrary, is allowed to react to innovations\nin the realized policy, re ecting the revision of expectations upon arrival of news. The\n ndings show that an unanticipated scal stimulus leads to expectations of strong de cit\nreversals over the subsequent two to three years. This in turn depresses domestic and\nforeign activity over the same horizon. Foresight shocks, on the contrary, have positive\ne ects on domestic activity. Di erences in the responses to surprise and foresight shocks\nre ect the role of expectations. The evidence is consistent with a regime where de cit\nreversals are mainly based on taxation alone and suggests that the incentive to reform\n scal regimes in an uncoordinated way may be small, while incentives for opportunistic\nbehavior may be strong.\nFinal Remarks\nThe thesis has examined how the anticipation of future policy regimes in uence the\ne ects of discretionary scal actions on a national and international level. The analysis\ndraws on the VAR approach with an identi cation scheme that allows to estimate the\ne ects of truly unanticipated scal measures as well as changes in expectations about\nfuture policies. The results obtained provide support to the hypothesis that foresight\na ects signi cantly the response of domestic and foreign economies to scal policy. More\nin details:\nChapter 2 tells us that a scal stimulus in the US produces a temporary boost in\ndomestic output and has no e ects on foreign economies as long as it is associated\nwith expectations of spending reversals. Conversely, the anticipation of an expansionary\npolicy has positive e ects on both domestic and foreign output, despite no scal measure\nis taken.\nChapter 3 tells us that expectations about future policy actions in uence the trans-\nmission of scal policy in Europe. Expansionary policies associated with previsions of\nstrong expenditure retrenchments over the subsequent two to three years depress domes-\ntic and foreign activity over the same period. Forecasts of expansionary de cit- nanced\nmeasures, on the contrary, have positive e ects on domestic output.\nThe evidence provided in this thesis has non-negligible policy implications. First,\nit shows that the e ects of scal policy depend

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.