CI018 Room Senate SPECIAL SESSION ON BOOTSTRAP INFERENCE Chair: Jean-Pierre Urbain
CI0665: A discrete model for bootstrap iteration Presenter: Russell Davidson, McGill University, Canada
In an attempt to free bootstrap theory from the shackles of asymptotic considerations, the aim is to study the possibility of justifying, or validating, the bootstrap, not by letting the sample size tend to infinity, but by considering the sequence of bootstrap p−values obtained by iterating the bootstrap. The main idea is that, if this sequence converges to a random variable that follows the uniformU(0,1)distribution, then the bootstrap is valid. The idea is studied by making the model under test discrete and finite, so that it is characterised by a finite three-dimensional array of probabilities. This device, when available, renders bootstrap iteration to any desired order feasible. It is used for studying a unit-root test for a process driven by a stationaryMA(1)process, where it is known that the unit-root test, even when bootstrapped, becomes quite unreliable when theMA(1)parameter is in the vicinity of−1. Iteration of the bootstrapp−value to convergence achieves reliable inference except for a parameter value very close to−1. We then endeavour to see these specific results in a wider context, and try to cast new light on where bootstrap theory may be going.
CI0802: Bootstrap inference for VAR models under rank uncertainty Presenter: Stephan Smeekes, Maastricht University, Netherlands Co-authors:Lenard Lieb
In impulse response analysis using VAR models it is common to construct confidence intervals using bootstrap techniques. However, in many practical applications, uncertainty regarding the true (unknown) cointegration rank is typically ignored, and impulse responses and their confidence intervals are constructed as if the cointegration rank were known, whether the rank has been estimated or simply assumed to be equal to a certain rank. Recently many methods have been proposed to robustify impulse responses to an unknown rank or order of integration. These studies however do not consider how the bootstrap is affected when constructing confidence intervals. We therefore investigate how bootstrap inference for VAR models, such as used for impulse response analysis, is affected by a misspecified cointegration rank. We derive theoretical results on the asymptotic validity of the bootstrap in this setting, and analyze finite sample effects through Monte Carlo simulation. This allows us to quantify how serious the problems are for empirical work if the uncertainty regarding the rank is ignored. We also consider modifications of the bootstrap that provide better guard against misspecification of the cointegration rank. Finally our results are illustrated with an empirical example.
CI1602: Dependent wild bootstrap for the empirical process and von Mises-statistics Presenter: Michael H Neumann, Friedrich Schiller University, Germany
Many important quantities in statistics can be written as a functional of the empirical process. Von Mises (V-) statistics appear as approximations of test statistics of Cramer-von Mises-type. It will be shown how modifications of the dependent wild bootstrap, which was originally introduced for smooth functionals of the sample mean, can be used for bootstrapping the empirical process and degenerate V-statistics of dependent random variables. Consistency of the bootstrap approximations is proved under minimal conditions.
CO438 Room G21A LARGE DIMENSIONAL PANEL MODELS Chair: Xun Lu
CO0188: Determining the number of groups in latent panel structures with an application to income and democracy Presenter: Xun Lu, Hong Kong University of Science and Technology, China
A latent group panel structure where the number of groups is unknown and has to be determined empirically is considered. We propose a testing procedure to determine the number of groups. Our test is a residual-based LM-type test. We show that after being appropriately standardized, our test is asymptotically normally distributed under the null hypothesis of a given number of groups and has power to detect deviations from the null.
Monte Carlo simulations show that our test performs remarkably well in finite samples. We apply our method to study the effect of income on democracy and find strong evidence of heterogeneity in the slope coefficients. Our testing procedure determines three latent groups among eighty two countries.
CO0658: Estimation of principal functional coefficient models for longitudinal data Presenter: Degui Li, University of York, United Kingdom
The estimation of the functional coefficient longitudinal data models is studied. In order to achieve dimension reduction for the nonparametric functional coefficients and improve the estimation efficiency, we introduce a novel semiparametric estimation procedure which combines a principal component analysis of the functional coefficients and a Cholesky decomposition of the within-subject covariance matrices. Under some regularity conditions, we derive the asymptotic distribution theory for the proposed semiparametric estimators and show that the efficiency of the estimation of the (principal) functional coefficients can be improved when the within-subject covariance structure is correctly specified. Furthermore, we apply two approaches to consistently estimate the autoregressive coefficients in the Cholesky decomposition, which help avoid a possible misspecification of the within-subject covariance structure and ensure the efficiency improvement for the estimation of the (principal) functional coefficients.
Some numerical studies including Monte Carlo experiments and an empirical application show that the developed semiparametric method works reasonably well in finite samples.
CC1093: Semiparametric trending regression for unbalanced panel data with application to realized volatility Presenter: Alev Atak, City University London, United Kingdom
A methodology is outlined for developing a semiparametric panel data model to describe the realized volatility and the trend in monthly dataset of US equity returns by using the Center for Research in Security Prices (CRSP) while relinquishing the assumption of global stationarity. We allow the trend to evolve in a nonparametric way, with an unknown smooth function. While we first provide idiosyncratic trends for each individual i, we aim to test for the common trends assumption based on a measure of nonparametric goodness-of-fit test before imposing it. We propose a semiparametric profile likelihood approach to estimate the model. We assume an asymptotic framework in whichTis large; but not necessarilyN.
CC1680: Generalized least squares estimation of panel with common shocks Presenter: Marco Avarucci, University of Glasgow, United Kingdom
Co-authors:Paolo Zaffaroni
The estimation of linear regression such asYi=Xiβi0+uiis considered, whereYi= (yi1, ...yiT)0is aT x1 vector of dependent variables,Xiis a T×Kmatrix of regressor andβi0are individual-specific parameters. The innovationui= (ui1, ...,uiT)0has a factor structureui=Fbi+ei, for a TxM matrix of latent factorsF= (f1, ...,fT)0with loadingsbiandei= (ei1, ...,eiT)0is a vector of idiosyncratic innovations. A factor structure in both the innovationuiand the regressorsXican make the ordinary least squares estimator inconsistent for the true regression coefficients. To overcome this problem, we propose a GLS-type estimator. The procedure can be summarized as follows: (i) Obtain theT×1 vector of residuals ˆui
by OLS. (ii) Construct theT×Tvariance covariance matrixW=N1∑Ni=1uˆiuˆ0i. (iii) Compute the GLS estimator using the matrixW. We show that,
ifT2/Napproaches zero forT,Ndiverging to infinity, the GLS estimator is consistent and asymptotically normal. This result is due to an important insight, namely the existence of a form of asymptotic orthogonality between the latent factorF and inverse ofW. This result holds despite the inconsistency of the OLS and does not require a preliminary estimate of the factor or a priory knowledge of their number.
CC1784: Canonical correlation analysis of panel VEC models Presenter: Piotr Keblowski, University of Lodz, Poland
The focus is on performance of the fully parametric system estimators of long-run relationships in the panel vector error correction framework that are based on eigenvalue problem. Therefore, we compare performance of Box-Tiao levels canonical correlation analysis with the classical approach proposed by Johansen in the panel data setting. The panel VEC model is considered with different sets of restrictions on the system’s structure. Therefore, it is consecutively allowed that (i) cross-sectional dependence in the error terms occurs, (ii) there is interaction of the short-run dynamics between cross-sections, (iii) there is interaction of the error-correction terms between cross-sections. The results for the individual time-series analyses, where cross-sections are assumed to be independent, are related to the results of the panel analysis. It is showed that there is a trade-off between the dimensionality effect which is well known from the standard time-series analysis and efficiency gains, which are due to cross-sectional dependencies. However, if there is a common cointegration rank and significant cross-sectional relationships, then the results of the MLE of long-run parameters in panels usually outperform the results for the standard time-series analyses, where cross-sections are assumed to be independent. Moreover the performance of the MLE of long-run parameters in the panel VEC is enhanced if the cross-sections share the same long-run structure.
CO456 Room Woburn INFLATION ANALYSIS AND FORECASTING Chair: Till Strohsal
CO0226: Analysis of aggregated inflation expectations based on the ECB SPF survey Presenter: Maritta Paloviita, Bank of Finland, Finland
The aim is to examine aggregated inflation expectations based on the ECB Survey of Professional Forecasters (ECB SPF). The focus of the analysis is on possible impacts of changing panel composition on aggregated short and long term point forecasts. We also investigate corresponding forecast uncertainties, which are based on subjective probability distributions. We compare changes in aggregated forecasts in the original unbalanced panel data with aggregated forecast changes based on a set of sub-panels of fixed composition. We also construct lower and upper bounds around aggre-gated forecast revisions. Our results indicate that the unbalanced panel data do not cause systematic distortions to aggreaggre-gated survey information, but there are some minor differences between alternative survey aggregates, which are not necessarily non-negligible from the monetary policy point of view. We provide evidence that both micro and macro level analysis of the ECB SPF survey information is needed, especially in times of wide disagreement across forecasters and high levels of inflation uncertainty.
CO0216: How oil price forecast errors impact inflation forecast errors
Presenter: Frederique Bec, THEMA University of Cergy-Pontoise and CREST, France Co-authors:Annabelle De Gaye
The aim is to propose an empirical investigation of the impact of oil price forecast errors on inflation forecast errors for two different sets of recent forecasts data: the median of SPF inflation forecasts for the U.S. and the Central Bank inflation forecasts for France. Mainly two salient points emerge from our results. First, there is a significant and dominant contribution of oil price forecast errors to the explanation of inflation forecast errors, whatever the country or the period considered. Second, the pass-through of oil price forecast errors to inflation forecast errors is multiplied by around 2 when the oil price volatility is large.
CO0320: The time-varying degree of inflation expectations anchoring Presenter: Till Strohsal, Freie Universitaet Berlin, Germany
Co-authors:Rafi Melnick, Dieter Nautz
Well-anchored inflation expectations have become a key indicator for the credibility of a central bank’s inflation target. Since the outbreak of the recent financial crisis, the existence and the degree of de-anchoring of U.S. inflation expectations have been under debate. An encompassing time-varying parameter model is introduced to analyze the changing degree of U.S. inflation expectations anchoring. Our model nests the two most common existing approaches as special cases. We confirm that inflation expectations have been partially de-anchored during the financial crisis.
Yet, our results suggest that inflation expectations have been successfully re-anchored ever since.
CO1095: A credit-based indicator for the risk of low inflation Presenter: Roberta Colavecchio, Hamburg University, Germany
We employ a credit-based early warning model in order to analyse the risk of a low inflation regime in the four major Euro area countries. The model specification allows for three different inflation regimes: Low, Medium and High inflation, with time-varying transition probabilities depending on a national credit aggregate. Using Bayesian techniques, we estimate the model with quarterly data from the early 1970s up to the end of 2014. Our analysis uncovers several country-specific features and suggests that, from 2011 on, the risks of a Low inflation regime have been increasing in Italy, Spain and, to a lesser extent in Germany while in France they have started to alleviate in the course of the last seven quarters of the sample.
Moreover, credit growth appears to play a role in the assessment of the risk of entering a low inflation state: the inclusion of a credit indicator variable signals an increase in such risk, especially for Italy and Spain.
CO422 Room Bedford ECONOMETRICS OF ART MARKETS Chair: Douglas Hodgson
CO0313: Efficiency of Italian opera houses: A stochastic frontier production function approach Presenter: Sabrina Auci, University of Palermo, Italy
Co-authors:Antonio Cognata
The empirical literature on the production of performing arts has mainly focused on cost functions. Studies have explored the cost structure of symphony orchestras, theatres and museums, mostly with the aim of finding evidence of scale economies. Until now no work has studied opera houses production or cost functions. Only recently there have been attempts in the use of a more suitable methodology and in finding evidence of the efficiency of performing arts institutions. The aim is to investigate efficiency of Italian opera houses using a stochastic frontier approach (SFA). The empirical analysis based on the concept of output maximization is performed on firm level unique database of 14 major Italian opera houses in the period 2001-2012. Dividing the error component into two aspects - the systematic and the noise components - the SFA allows to consider separately inputs of the production function, such as physical, labour and human capital from factors of the inefficiency model influencing the behaviour of opera houses. These latter factors represent the opera houses heterogeneity and show the influence on technical efficiency scores.
Finally, we rank opera companies on the basis of the estimated technical inefficiency.
CO0326: The relationship between artistic movements and artist careers: Evidence from individual-level hedonic regression Presenter: Douglas Hodgson, UQAM, Canada
Co-authors:John Galbraith, Christiane Hellmanzik
The literature on age-valuation profiles of artists has paid limited attention to the effects of membership in artistic movements. There are many reasons why membership in a movement can be important for the career dynamic of an artist. The relation between careers and movement membership has been previously studied by considering data on numbers of reproductions in art history books. The hedonic analysis of auction data in this area is limited, with results of regressions of pooled groups of artists being reported. Ideally, one would like to estimate individual-artist profiles relating valuation to date of production, and compare these with pooled profiles estimated for groups or movements to which the artists belong, to assess the relation between individual- and group-level price dynamics. Until recently, such an endeavour was rendered difficult by the small number of observations, compared to a large number of hedonic covariates, often available at the individual-artist level. But the successful application to this problem of recent dimensionality-reduction and model-averaging methods in the context of estimating individual age-valuation profiles suggests the utility of applying the same approach to estimating individual profiles in the context of movement membership. We thus apply these methods to a large data set on auction prices for major modern painters.
CO0583: A first approach to pricing on the painting secondary market in the Argentine Republic Presenter: Carolina Boufflet, University of Salvador and National Ministry of Industry, Argentina Co-authors:Marcos Leonardo Chamorro
The ongoing discussion of methods for predicting prices on the painting secondary market in the Argentine Republic is advanced. The research is bounded to 3542 sales transactions (not repeated) conducted by auction houses located in the city of Buenos Aires in the period from January to December 2014. Given the specificity for works of arts demand, one approach to the demand from the Theory of characteristics was proposed. The estimation of a hedonic price index was performed by translog regressions. According to preliminary results, Argentines market of art shows the existence of “Masterpiece effect”, the “Main auctions effect” and a “batch effect” in relation to the sales strategy conducted by auction houses. The model threw anR2 of 65% and this pose the challenge of running a model with 2SLS instrumental variable. Artist index was taken as instrumental variable and elements of his biography and career as dependent variables that were not explicitly included in the previous model. Also a model of latent or hidden variables was developed to find out that the artwork has Its own life beyond the artist. The goodness of fit was tested to analyze the strengths and weaknesses (or shortcomings) of each model.
CO0669: From quality to utility, an empirical study of artist reputations impact on contemporary art market price Presenter: Simeng Chang, Erasmus University Rotterdam, Netherlands
The aim is to analyze how an artists reputation determines the market price of his or her artworks. Reputation is a proxy of quality as well as art price. In other words, reputation signals the artistic value of the artwork and generates the economic value. In the art field, an artistic reputation is built through the process where the artistic value is recognized and enhanced by getting constantly attention from the art experts. In the art market, consumers aesthetic utility is confirmed by assessing the quality from artistic reputation. The aesthetic utility is further enhanced by the added value derived from market reputation. Therefore, the price of an artwork with higher reputation is higher since the quality uncertainty is reduced and the consumer utility is raised. Using the reputation quantified tool, Artfacts artist ranking, and auction data, we are intended to empirically identify the reputation effect on the art market price.
CC0570: Historic art exhibitions and modern day auction results Presenter: Christiane Hellmanzik, University of Hamburg, Germany
Historic art exhibitions are used in order to investigate the impact of the contemporary success of artistic careers on today’s auction prices of modern paintings. That is, if an artist’s work was displayed in a historic art exhibition in a given year, paintings dated from this year fetch higher prices at auction today. This can be attributed to two effects: artists who participated in such shows were already acknowledged as superstars contemporaneously and participants in art exhibitions benefited from a longer-lasting career boost as reflected by positive mark-ups on paintings made in the years following a show. For both channels participation in historic art exhibitions is a strong quality signal for today’s art buyers.
The study is based on a global sample of 273 ‘superstars’ of modern art born between 1800 and 1945, 34,141 auction results of paintings and participation in important historic art exhibitions.
CO394 Room Torrington COMMODITY MARKETS: PRICING AND TRADING Chair: Ana-Maria Fuertes CO0318: The determinants of convenience yields
Presenter: Marcel Prokopczuk, Leibniz University Hannover, Germany Co-authors:Yingying Wu
The determinants of convenience yields across a broad range of commodities are investigated. We find that the convenience yields of commodities are exposed to both commodity-specific and systematic factors, but to a different extent. The difference in explanatory power of these factors for each commodity sheds light on the heterogeneity of commodity markets. One main difference between commodity sectors lies in their different sensitivities towards the state of the economy.
CO1001: The skewness of commodity futures returns Presenter: Joelle Miffre, EDHEC Business School, France Co-authors:Ana-Maria Fuertes, Adrian Fernandez-Perez, Bart Frijns
The relationship between skewness of the distribution of past returns and expected returns in commodity futures markets is explored. Both time-series tests and cross-sectional tests indicate that more positively skewed commodities accrue significantly lower mean excess returns. Sorting a cross-section of commodities by their past skewness, we demonstrate that a fully collateralized portfolio that buys commodities with the most negative skewness and shorts commodities with the most positive skewness earns an excess return of 8.01% a year. A commodity pricing model that utilizes as risk factors the excess returns of a long-only equally-weighted portfolio of all commodities, alongside term structure, momentum, and hedging pressure portfolios yields a significant alpha of 6.58% obtained which indicates that the profitability of skewness portfolios is not a mere manifestation of backwardation and contango risk. Skewness risk may uniquely relate to the preferences of investors for lottery-like commodity futures. The findings are robust to transaction costs, liquidity considerations and sample periods.
CO0975: The earnings-price ratio and predictability of earnings in the dry bulk shipping industry Presenter: Nikos Nomikos, Cass Business School, United Kingdom
Co-authors:Ioannis Moutzouris
We examine second-hand vessel prices, net earnings, and returns on capital in the dry bulk shipping industry. We demonstrate that the bulk of variation in net earnings-price ratios reflects varying expected net earnings growth. Furthermore, we contribute to the literature by examining a forward-looking definition of the net earnings-price ratio, and extending the variance decomposition framework to assets with limited economic lives. Our results strongly indicate that shipping net earnings-price ratios negatively forecast future net earnings growth. In addition, there is no statistical evidence that the net earnings-price ratio is negatively related to future returns. These results are in contrast to the recent empirical asset pricing literature in the U.S. equity and housing markets. Importantly, however, our findings agree with recent results obtained from global equity