Essays on housing and macroeconomic dynamics
- Autores: Alessandro Galesi
- Directores de la Tesis: Enrique Sentana Iváñez (dir. tes.), Claudio Michelacci (codir. tes.)
- Lectura: En la Universidad Internacional Menéndez Pelayo (UIMP) ( España ) en 2015
- Idioma: español
- Tribunal Calificador de la Tesis: Tommaso Monacelli (presid.), Hernan D. Seoane (secret.), Josep Pijoan-Mas (voc.), Luca Gambetti (voc.), Evi Pappa (voc.)
- Materias:
- Texto completo no disponible (Saber más ...)
- Resumen
Recent macroeconomic developments across advanced economies have renewed the interest in explicitly considering the interaction between housing and the overall economy for the purpose of business cycle analysis. In this respect, the recent research on business cycle analysis is revisiting its macroeconomic models by explicitly incorporating housing and housing nance so to deepen our understanding of how business cycles in modern economies work.
This dissertation contributes to this research in two directions: rst, by presenting novel theories in housing, macroeconomics, and their interaction using the last generation of Dynamic Stochastic General Equilibrium (DSGE) models which explicitly feature housing and housing nance; and second, by proposing new estimation techniques for dynamic factor models, which are widely used tools in business cycle analysis for capturing comovements across multiple macroeconomic series.
Chapter 1 addresses the question of what drives the recent developments of house prices in the United States. Using US industry-level data I document a downward trend in construction productivity relative to other industries which started around the end of the 1960's and still persists. First I study how this productivity slowdown can a ect house prices by considering an open economy with collateralized debt and two sectors, construction and non-construction, whose productivities grow at di erent rates. I show that when productivity in construction falls, house prices increase. The e ect is ampli ed when the rate at which the economy can borrow is low. As house prices increase, the collateral constraint gets relaxed. If the borrowing rate is suciently low, households nd optimal to accumulate foreign debt and these capital in ows lead to surges in both residential investment and land prices. In this way house prices rise even further.
Subsequently I calibrate the economy to match the US evidence, which is characterized by a pronounced slowdown in construction productivity and a low borrowing rate. While the productivity slowdown in construction alone can account for the long-run trend in house prices over the 1970's - 2000's, its combination with the low interest rate is crucial to generate the increases in prices of land and housing of the early 2000's. This interaction also accounts for key stylized facts of the US housing cycle: (i) the positive correlation between housing prices and residential investment; (ii) land prices are twice as volatile as house prices; (iii) part of the worsening of the current account. In a closed economy the productivity slowdown in construction cannot account for these facts, hence I conclude that its interaction with the low interest rate is what helps to explain US house prices.
Using a Panel VAR with sign restrictions I also show that this mechanism can be important to explain the dynamics of house prices across OECD countries.
Chapter 2, co-authored with Claudio Michelacci and Vincenzo Quadrini, tackles the debate on the justi ability of policies aimed at supporting economic activity during recessions.
1 We emphasize one possible reason for subsidizing aggregate activity that applies to nancially driven recessions in which the real estate value matters for economic activity.
The conventional wisdom is that house prices matter for the business cycle because housing can be used as a collateral and a higher value of collateral relaxes the collateral constraints thereby promoting consumption, investment, and overall aggregate activity.
However the aggregate level of income and employment a ects the value of all non-tradable assets such as houses and capital. And since private rms take prices as given, they do not internalize that higher economic activity pushes up prices of all non-tradables and relaxes the collateral constraints, which is a pecuniary externality as in Lorenzoni (2008).
Thus in recessions in which households are nancially constrained there exists a price externality that justi es subsidizing business activity. We compute the optimal subsidies and study their properties. Moreover, using a unique new data set collecting information on business subsidies granted by US state governments since the 2000's, we document that over the US Great Recession the incidence of business subsidies has dramatically increased. By employing US States data over the 2000's, an estimated Panel VAR with sign restrictions documents that business subsidies lead to an increase in the value of 1Each coauthor was involved in developing each part of this research. Nonetheless, my contribution specializes to the motivating evidence and theoretical framework (Sections 2.2 and 2.3), while the contributions of Claudio Michelacci and Vincenzo Quadrini specialize to the social planner problem and properties of optimal subsidies (Sections 2.4, 2.5, 2.6, and 2.7).houses and in the level of household debt, consistently with our mechanism.
Chapter 3, co-authored with Claudio Michelacci, documents that in recent US business cycle episodes the correlation between manufacturing and service employment has increased, and more so in recessions and US states where households are highly indebted.2 We argue that this increased sectoral synchronization is a major implication of the dramatic US households' debt expansion experienced since the 1990's, which has made households more prone to be nancially constrained especially during severe recessions. When households are nancially constrained, manufacturing output matters for the demand in services. In fact, while manufacturing produces tradable goods whose demand is determined internationally, most services are non-tradable with their demand set just by local economic conditions. Hence when manufacturing employment falls, households disposable income falls which, due to the binding nancial constraint, forces households to deleverage. Debt repayments crowd out manufacturing consumption, and due to complementarity in consumption of goods and services, consumption of services contracts.
As a result economic activity falls more, households become even more constrained, and this further contracts demand for services. This prevents job creation in services and comovement increases.
Chapter 4, co-authored with Gabriele Fiorentini and Enrique Sentana, presents a novel algorithm for estimating general dynamic factor models which allows for dynamic loadings and ARMA dynamics. Our algorithm reduces the computational burden so much that researchers can estimate such models with a large number of series without good initial values.3 Our estimator of the model's parameters is a spectral variant of Maximum Likelihood, which maximizes the Whittle (spectral) approximation to the log-likelihood function. We exploit the fact that the spectral density matrix of a dynamic factor model has the structure of the unconditional covariance matrix of a static factor model, but with di erent common and idiosyncratic variances for each frequency. Hence by working 2Each coauthor contributed by 50 percent in developing each part of this research.
3Each coauthor was involved to various degrees in developing each part of this research. Nonetheless, the key contributions of Gabriele Fiorentini and Enrique Sentana are in suggesting the research topic and the analytical results (Sections 4.2 and 4.3), while my contribution specializes in the computational part as well as the empirical application (Section 4.4).
in the frequency domain we sidestep the computational diculties associated with the intrinsic recursivity of the dynamic factor model. Although our proposed procedure allows researchers to estimate such models by maximum likelihood with many series even without good initial values, we recommend switching to a gradient method that uses the EM principle to swiftly compute frequency domain analytical scores near the optimum. In an empirical application we successfully employ our algorithm to construct an index that captures the common movements of 81 US sectoral employment growth rates over the period 1990-2014.
Chapter 5, co-authored with Gabriele Fiorentini and Enrique Sentana, generalizes the mentioned algorithm to bifactor models in which pervasive common factors are complemented by block factors.4 We explain how to eciently exploit the sparsity of the loading matrix to reduce the computational burden so much that researchers can estimate such models by maximum likelihood with a large number of series from multiple regions. As for the case of models featuring a single factor, our algorithm allows for dynamic loadings and ARMA dynamics. We also derive convenient expressions for the spectral scores and information matrix, which allows us to switch to the scoring algorithm near the optimum.
In an empirical application we explore the ability of a model with a global factor and three regional ones to capture in ation dynamics across 25 European countries over the period 1999-2014.
4Each coauthor was involved to various degrees in developing each part of this research. Nonetheless, the key contributions of Gabriele Fiorentini and Enrique Sentana are in suggesting the research topic and the analytical results (Sections 5.2 and 5.3), while my contribution specializes in the computational part as well as the empirical application (Section 5.4).
