Fitting dynamic factor models to non-stationary time series

Factor modelling of a large time series panel has widely proven useful to reduce its cross-sectional dimensionality. This is done by explaining common co-movements in the panel through the existence of a small number of common components, up to some idiosyncratic behaviour of each individual series. To capture serial correlation in the common components, a dynamic structure is used as in traditional (uni- or multivariate) time series analysis of second order structure, i.e. allowing for infinite-length filtering of the factors via dynamic loadings. In this paper, motivated from economic data observed over long time periods which show smooth transitions over time in their covariance structure, we allow the dynamic structure of the factor model to be non-stationary over time by proposing a deterministic time variation of its loadings. In this respect we generalize the existing recent work on static factor models with time-varying loadings as well as the classical, i.e. stationary, dynamic approximate factor model. Motivated from the stationary case, we estimate the common components of our dynamic factor model by the eigenvectors of a consistent estimator of the now time-varying spectral density matrix of the underlying data-generating process. This can be seen as a time-varying principal components approach in the frequency domain. We derive consistency of this estimator in a “double-asymptotic” framework of both cross-section and time dimension tending to infinity. The performance of the estimators is illustrated by a simulation study and an application to a macroeconomic data set.     

Forecasting macroeconomic variables using collapsed dynamic factor analysis

We explore a new approach to the forecasting of macroeconomic variables based on a dynamic factor state space analysis. Key economic variables are modeled jointly with principal components from a large time series panel of macroeconomic indicators using a multivariate unobserved components time series model. When the key economic variables are observed at a low frequency and the panel of macroeconomic variables is at a high frequency, we can use our approach for both nowcasting and forecasting purposes. Given a dynamic factor model as the data generation process, we provide Monte Carlo evidence of the finite-sample justification of our parsimonious and feasible approach. We also provide empirical evidence for a US macroeconomic dataset. The unbalanced panel contains quarterly and monthly variables. The forecasting accuracy is measured against a set of benchmark models. We conclude that our dynamic factor state space analysis can lead to higher levels of forecasting precision when the panel size and time series dimensions are moderate.     

Linear prediction of temporal aggregates under model misspecification

This paper aims to demonstrate a possible aggregation gain in predicting future aggregates under a practical assumption of model misspecification. Empirical analysis of a number of economic time series suggests that the use of the disaggregate model is not always preferred over the aggregate model in predicting future aggregates, in terms of an out-of-sample prediction root-mean-square error criterion. One possible justification of this interesting phenomena is model misspecification. In particular, if the model fitted to the disaggregate series is misspecified (i.e., not the true data generating mechanism), then the forecast made by a misspecified model is not always the most efficient. This opens up an opportunity for the aggregate model to perform better. It will be of interest to find out when the aggregate model helps. In this paper, we study a framework where the underlying disaggregate series has a periodic structure. We derive and compare the efficiency loss in linear prediction of future aggregates using the adapted disaggregate model and aggregate model. Some scenarios for aggregation gain to occur are identified. Numerical results show that the aggregate model helps over a fairly large region in the parameter space of the periodic model that we studied.     

Combining filter design with model-based filtering (with an application to business-cycle estimation)

Filters used to estimate unobserved components in time series are often designed on a priori grounds, so as to capture the frequencies associated with the component. A limitation of these filters is that they may yield spurious results. The danger can be avoided if the so-called ARIMA-model-based (AMB) procedure is used to derive the filter. However, parsimony of ARIMA models typically implies little resolution in terms of the detection of hidden components. It would be desirable to combine a higher resolution with consistency of the structure of the observed series.We show first that for a large class of a priori designed filters, an AMB interpretation is always possible. Using this result, proper convolution of AMB filters can produce richer decompositions of the series that incorporate a priori desired features of the components and fully respect the ARIMA model for the observed series (hence no additional parameter needs to be estimated).The procedure is discussed in detail in the context of business-cycle estimation by means of the Hodrick-Prescott filter applied to a seasonally adjusted series or a trend–cycle component.     

What do professional forecasters actually predict?

This paper studies what professional forecasters predict. We use spectral analysis and state space modeling to decompose economic time series into trend, business cycle, and irregular components. We examine which components are captured by professional forecasters by regressing their forecasts on the estimated components extracted from both the spectral analysis and the state space model. For both decomposition methods, we find that, in the short run, the Survey of Professional Forecasters can predict almost all of the variation in the time series due to the trend and the business cycle, but that the forecasts contain little or no significant information about the variation in the irregular component.     

THE ENTERPRISE ACT:ASPECTS OF THE NEW REGIME1

The forthcoming Enterprise Act makes the Competition Commission (CC) determinative in relation to merger and market inquiries. It also introduces new competition‐based tests, the rationale for which is examined. Several procedural aspects of the new regime are explored, in particular the need for economic guidance to be published on the application of the new tests. A number of key economic considerations are then examined, including market definition, oligopoly pricing, entry and the scope for different perspectives as between economic analysis and business practice.     

What's Unique About the Federal Funds Rate? Evidence from a Spectral Perspective*

This paper compares the behaviour of the effective federal funds rate to 10 US interest rates with maturities ranging from overnight to 10 years. Using spectral estimation methods, we identified idiosyncratic shocks to the funds rate and provided evidence on their impact on other rates at various frequencies. Our results suggest that, while all of the interest rates examined have common shocks at low frequencies, the federal funds rate contains some unique information at high frequency, although this information appears to be relevant only at the short end of the term structure. In turn, these results are open to various alternative interpretations.     

A Unit Root Test Using a Fourier Series to Approximate Smooth Breaks*

We develop a unit‐root test based on a simple variant of Gallant's (1981) flexible Fourier form. The test relies on the fact that a series with several smooth structural breaks can often be approximated using the low frequency components of a Fourier expansion. Hence, it is possible to test for a unit root without having to model the precise form of the break. Our unit‐root test employing Fourier approximation has good size and power for the types of breaks often used in economic analysis. The appropriate use of the test is illustrated using several interest rate spreads.     

Estimation of a dynamic multi-level factor model with possible long-range dependence

A dynamic multi-level factor model with possible stochastic time trends is proposed. In the model, long-range dependence and short memory dynamics are allowed in global and local common factors as well as model innovations. Estimation of global and local common factors is performed on the prewhitened series, for which the prewhitening parameter is estimated semiparametrically from the cross-sectional and local average of the observable series. Employing canonical correlation analysis and a sequential least-squares algorithm on the prewhitened series, the resulting multi-level factor estimates have centered asymptotic normal distributions under certain rate conditions depending on the bandwidth and cross-section size. Asymptotic results for common components are also established. The selection of the number of global and local factors is discussed. The methodology is shown to lead to good small-sample performance via Monte Carlo simulations. The method is then applied to the Nord Pool electricity market for the analysis of price comovements among different regions within the power grid. The global factor is identified to be the system price, and fractional cointegration relationships are found between local prices and the system price, motivating a long-run equilibrium relationship. Two forecasting exercises are then discussed.     

COMMON SEASONAL FEATURES: GLOBAL UNEMPLOYMENT

Seasonal patterns in economic time series are generally examined from a univariate point of view. Using extensions of the unit root literature, important classes of seasonal processes are deterministic, stationary stochastic or mean reverting, and unit root stochastic. Time series tests have been developed for each of these. This paper examines seasonality in a multivariate context. Systems of economic variables can have trends, cycles and unit roots as well as the various types of seasonality. Restrictions such as cointegration and common cycles are here applied also to examine multivariate seasonal behaviour of economic variables. If each of a collection of series has a certain type of seasonality but a linear combination of these series can be found without seasonality, then the seasonal is said to be ‘common’. New tests are developed to determine if seasonal characteristics are common to a set of time series. These tests can be employed in the presence of various other time series structures. The analysis is applied to OECD data on unemployment for the period 1975.1 to 1993.4, and it is found that four diverse countries (Australia, Canada, Japan and USA) not only have common trends in their unemployment, but also have common deterministic seasonal features and a common cycle/stochastic seasonal feature. Such a collection of characteristics were not found in other groups of OECD countries.     

MULTIVARIATE METHODS FOR MONITORING STRUCTURAL CHANGE

Detection of structural change is a critical empirical activity, but continuous ‘monitoring’ for changes in real time raises well‐known econometric issues that have been explored in a single series context. If multiple series co‐break then it is possible that simultaneous examination of a set of series helps identify changes with higher probability or more rapidly than when series are examined on a case‐by‐case basis. Some asymptotic theory is developed for maximum and average CUSUM detection tests. Monte Carlo experiments suggest that these both provide an improvement in detection relative to a univariate detector over a wide range of experimental parameters, given a sufficiently large number of co‐breaking series. This is robust to a cross‐sectional correlation in the errors (a factor structure) and heterogeneity in the break dates. We apply the test to a panel of UK price indices. Copyright © 2011 John Wiley & Sons, Ltd.     

Are more data always better for factor analysis?

Factors estimated from large macroeconomic panels are being used in an increasing number of applications. However, little is known about how the size and the composition of the data affect the factor estimates. In this paper, we question whether it is possible to use more series to extract the factors, and yet the resulting factors are less useful for forecasting, and the answer is yes. Such a problem tends to arise when the idiosyncratic errors are cross-correlated. It can also arise if forecasting power is provided by a factor that is dominant in a small dataset but is a dominated factor in a larger dataset. In a real time forecasting exercise, we find that factors extracted from as few as 40 pre-screened series often yield satisfactory or even better results than using all 147 series. Weighting the data by their properties when constructing the factors also lead to improved forecasts. Our simulation analysis is unique in that special attention is paid to cross-correlated idiosyncratic errors, and we also allow the factors to have stronger loadings on some groups of series than others. It thus allows us to better understand the properties of the principal components estimator in empirical applications.     

DSGE Models in the Frequency Domains

We use frequency domain techniques to estimate a medium‐scale dynamic stochastic general equilibrium (DSGE) model on different frequency bands. We show that goodness of fit, forecasting performance and parameter estimates vary substantially with the frequency bands over which the model is estimated. Estimates obtained using subsets of frequencies are characterized by significantly different parameters, an indication that the model cannot match all frequencies with one set of parameters. In particular, we find that: (i) the low‐frequency properties of the data strongly affect parameter estimates obtained in the time domain; (ii) the importance of economic frictions in the model changes when different subsets of frequencies are used in estimation. This is particularly true for the investment adjustment cost and habit persistence: when low frequencies are present in the estimation, the investment adjustment cost and habit persistence are estimated to be higher than when low frequencies are absent. Copyright © 2014 John Wiley & Sons, Ltd.     

Estimation of Common Factors under Cross‐Sectional and Temporal Aggregation Constraints

The paper estimates a large‐scale mixed‐frequency dynamic factor model for the euro area, using monthly series along with gross domestic product (GDP) and its main components, obtained from the quarterly national accounts (NA). The latter define broad measures of real economic activity (such as GDP and its decomposition by expenditure type and by branch of activity) that we are willing to include in the factor model, in order to improve its coverage of the economy and thus the representativeness of the factors. The main problem with their inclusion is not one of model consistency, but rather of data availability and timeliness, as the NA series are quarterly and are available with a large publication lag. Our model is a traditional dynamic factor model formulated at the monthly frequency in terms of the stationary representation of the variables, which however becomes nonlinear when the observational constraints are taken into account. These are of two kinds: nonlinear temporal aggregation constraints, due to the fact that the model is formulated in terms of the unobserved monthly logarithmic changes, but we observe only the sum of the monthly levels within a quarter, and nonlinear cross‐sectional constraints, since GDP and its main components are linked by the NA identities, but the series are expressed in chained volumes. The paper provides an exact treatment of the observational constraints and proposes iterative algorithms for estimating the parameters of the factor model and for signal extraction, thereby producing nowcasts of monthly GDP and its main components, as well as measures of their reliability.     

Toward regional economic planning: The New England case

As a result of the current change in economic thinking toward planning, this article, using New England as a case, after some preliminary data analysis on the continuity of the two sets of time series data, and the rejection of a hypothesis on the similarity between the regional and national economic structures, proposes first to estimate and then to project the regional economic structure and its possible shift in terms of industrial shares of their 10 component industries. Possible contributions of this kind of study toward regional economic planning then conclude the article.     

A unified approach to validating univariate and multivariate conditional distribution models in time series

Modeling conditional distributions in time series has attracted increasing attention in economics and finance. We develop a new class of generalized Cramer–von Mises (GCM) specification tests for time series conditional distribution models using a novel approach, which embeds the empirical distribution function in a spectral framework. Our tests check a large number of lags and are therefore expected to be powerful against neglected dynamics at higher order lags, which is particularly useful for non-Markovian processes. Despite using a large number of lags, our tests do not suffer much from loss of a large number of degrees of freedom, because our approach naturally downweights higher order lags, which is consistent with the stylized fact that economic or financial markets are more affected by recent past events than by remote past events. Unlike the existing methods in the literature, the proposed GCM tests cover both univariate and multivariate conditional distribution models in a unified framework. They exploit the information in the joint conditional distribution of underlying economic processes. Moreover, a class of easy-to-interpret diagnostic procedures are supplemented to gauge possible sources of model misspecifications. Distinct from conventional CM and Kolmogorov–Smirnov (KS) tests, which are also based on the empirical distribution function, our GCM test statistics follow a convenient asymptotic $N(0,1)$ distribution and enjoy the appealing “nuisance parameter free” property that parameter estimation uncertainty has no impact on the asymptotic distribution of the test statistics. Simulation studies show that the tests provide reliable inference for sample sizes often encountered in economics and finance.     

Some models for simulating canadian manpower flows and related systems

The paper reports on a series of models designed for computer simulation of Canadian manpower flows at different levels of unemployment, either alone or in conjunction with related population flows. The models are disaggregated macromodels of the state-transition variety. One relates to annual flows, the others to monthly flows.The models have been used for various purposes, including simulation analyses of the effects of changes in economic conditions and government policy variables on two government programs, namely unemployment insurance and manpower training. Other applications are possible and, in general, the basic approach underlying the models is considered to have much potential for manpower policy analysis.     

Forecasting aggregates and disaggregates with common features

This paper focuses on the provision of consistent forecasts for an aggregate economic indicator, such as a consumer price index and its components. The procedure developed is a disaggregated approach based on single-equation models for the components, which take into account the stable features that some components share, such as a common trend and common serial correlation. Our procedure starts by classifying a large number of components based on restrictions from common features. The result of this classification is a disaggregation map, which may also be useful in applying dynamic factors, defining intermediate aggregates or formulating models with unobserved components. We use the procedure to forecast inflation in the Euro area, the UK and the US. Our forecasts are significantly more accurate than either a direct forecast of the aggregate or various other indirect forecasts.     

Sparse estimation of dynamic principal components for forecasting high-dimensional time series

We present the sparse estimation of one-sided dynamic principal components (ODPCs) to forecast high-dimensional time series. The forecast can be made directly with the ODPCs or by using them as estimates of the factors in a generalized dynamic factor model. It is shown that a large reduction in the number of parameters estimated for the ODPCs can be achieved without affecting their forecasting performance.