On the predictability of time-varying VAR and DSGE models

Over the last few years, there has been a growing interest in DSGE modelling for predicting macroeconomic fluctuations and conducting quantitative policy analysis. Hybrid DSGE models have become popular for dealing with some of the DSGE misspecifications as they are able to solve the trade-off between theoretical coherence and empirical fit. However, these models are still linear and they do not consider time variation for parameters. The time-varying properties in VAR or DSGE models capture the inherent nonlinearities and the adaptive underlying structure of the economy in a robust manner. In this article, we present a state-space time-varying parameter VAR model. Moreover, we focus on the DSGE–VAR that combines a microfounded DSGE model with the flexibility of a VAR framework. All the aforementioned models as well simple DSGEs and Bayesian VARs are used in a comparative investigation of their out-of-sample predictive performance regarding the US economy. The results indicate that while in general the classical VAR and BVARs provide with good forecasting results, in many cases the TVP–VAR and the DSGE–VAR outperform the other models.     

Short-term inflation forecasting models for Turkey and a forecast combination analysis

In this paper, we produce short term forecasts for the inflation in Turkey, using a large number of econometric models. In particular, we employ univariate models, decomposition based approaches (both in frequency and time domain), a Phillips curve motivated time varying parameter model, a suite of VAR and Bayesian VAR models and dynamic factor models. Our findings suggest that the models which incorporate more economic information outperform the benchmark random walk, and the relative performance of forecasts are on average 30% better for the first two quarters ahead. We further combine our forecasts by means of several weighting schemes. Results reveal that, the forecast combination leads to a reduction in forecast error compared to most of the models, although some of the individual models perform alike in certain horizons.     

Asset prices, exchange rates and the current account

This paper analyses the role of asset prices in comparison to other factors, in particular exchange rates, as a driver of the US trade balance. It employs a Bayesian structural VAR model that requires imposing only a minimum of economically meaningful sign restrictions. We find that equity market shocks and housing price shocks have been major determinants of the US current account in the past, accounting for up to 30% of the movements of the US trade balance at a horizon of 20 quarters. By contrast, shocks to the real exchange rate have been less relevant, explaining about 9% and exerting a more temporary effect on the US trade balance. Our findings suggest that large exchange rate movements may not necessarily be the key element of an adjustment of today's large current account imbalances, and that in particular relative global asset price changes could be a potent source of adjustment.     

A DSGE-VAR model for forecasting key South African macroeconomic variables

The paper develops a Small Open Economy New Keynesian DSGE-VAR (SOENKDSGE-VAR) model of the South African economy, characterised by incomplete pass-through of exchange rate changes, external habit formation, partial indexation of domestic prices and wages to past inflation, and staggered price and wage setting. The model is estimated using Bayesian techniques on data from the period 1980Q1 to 2003Q2, and then used to forecast output, inflation and nominal short-term interest rate for one-to eight-quarters-ahead over an out-of sample horizon of 2003Q3 to 2010Q4. When the forecast performance of the SOENKDSGE-VAR model is compared with an independently estimated DSGE model, the classical VAR and six alternative BVAR models, we find that, barring the BVAR model based on the SSVS prior on both VAR coefficients and the error covariance, the SOENKDSGE-VAR model is found to perform competitively, if not, better than all the other VAR models.     

Forecasting house prices for the four census regions and the aggregate US economy in a data-rich environment

This article considers the ability of large-scale (involving 145 fundamental variables) time-series models, estimated by dynamic factor analysis and Bayesian shrinkage, to forecast real house price growth rates of the four US census regions and the aggregate US economy. Besides the standard Minnesota prior, we also use additional priors that constrain the sum of coefficients of the VAR models. We compare 1- to 24-months-ahead forecasts of the large-scale models over an out-of-sample horizon of 1995:01–2009:03, based on an in-sample of 1968:02–1994:12, relative to a random walk model, a small-scale VAR model comprising just the five real house price growth rates and a medium-scale VAR model containing 36 of the 145 fundamental variables besides the five real house price growth rates. In addition to the forecast comparison exercise across small-, medium- and large-scale models, we also look at the ability of the ‘optimal’ model (i.e. the model that produces the minimum average mean squared forecast error) for a specific region in predicting ex ante real house prices (in levels) over the period of 2009:04 till 2012:02. Factor-based models (classical or Bayesian) perform the best for the North East, Mid-West, West census regions and the aggregate US economy and equally well to a small-scale VAR for the South region. The ‘optimal’ factor models also tend to predict the downward trend in the data when we conduct an ex ante forecasting exercise. Our results highlight the importance of information content in large number of fundamentals in predicting house prices accurately.     

Evaluating the effect of monetary policy on unemployment with alternative inflation forecasts

This paper explores the role that inflation forecasts play in the uncertainty surrounding the estimated effects of alternative monetary rules on unemployment dynamics in the euro area and the US. We use the inflation forecasts of 8 competing models in a standard Bayesian VAR to analyse the size and the timing of these effects, as well as to quantify the uncertainty relative to the different inflation models under two rules. The results suggest that model uncertainty can be a serious issue and strengthen the case for a policy strategy that takes into account several sources of information. We find that combining inflation forecasts from many models not only yields more accurate forecasts than those of any specific model, but also reduces the uncertainty associated with the real effects of policy decisions. These results are in line with the model-combination approach that central banks already follow when conceiving their strategy.     

A national or local phenomenon? A comparison of the US and the Chinese housing markets

This article investigates the common movements of house prices across cities as well as the macroeconomic underpinnings of the comovements in the US and China. Our empirical results indicate more differences than similarities between the US and the Chinese housing markets. The results from a Bayesian dynamic latent factor model indicate that the fluctuations of house prices across cities in the US are more a national phenomenon, while the dynamics of house prices across cities in China are mainly driven by the city-specific component. We further use VAR models to compare the roles of the underlying determinants in these two housing markets. The results show that the roles of monetary policy shocks and aggregate fluctuations in driving the common movements of house prices across cities differ substantially between the US and China at both short and long horizons.     

Forecasting with a state space time-varying parameter VAR model: Evidence from the Euro area

Standard VAR and Bayesian VAR models are proven to be reliable tools for modeling and forecasting, yet they are still linear and they do not consider time-variation in parameters. VAR modeling is subject to the Lucas critique and fails to take into account the inherent nonlinearities of the economy, while it can only be utilized in the analysis of stationary series and in many cases stationarity assumptions are too restrictive. A novel time-varying multivariate state-space estimation method for vector autoregression models is introduced. For the time-varying parameter model (TVP-VAR), the parameters are estimated using a multivariate specification of the standard Kalman filter (Harvey, 1990) combined with a suitable extension of the univariate methodology framework of Kim and Nelson (1999). The TVP-VAR model as well as standard VARs and Bayesian VARs, are used in a comparative investigation of their predicting performance for the monthly IP, CPI and Euribor rate of the EU economy. The total period covers 1999:1–2011:2 with an out-of-sample testing period of 2007:2 to 2011:2, which included the US sub-prime and the EU debt crisis sub-periods. The results varied across the investigated time series and indicated that the TVP-VAR model consistently outperforms the other models in case of the EU monthly CPI, while different specifications of the VAR and BVAR models for the IP and Euribor series provide with better forecasting performance. Interestingly, the robustness analysis showed that the TVP-VAR model provided with enhanced predictability in particular during “crisis times”.     

Estimating Sectoral Disinflation Cost in India: Some Structural VAR Evidence

This paper attempts a sectoral estimation of sacrifice ratios for India. Two Structural VAR models are estimated using quarterly data for the period 1997–98 Q1 to 2016–17 Q1. The empirical findings suggest that the real cost of disinflation policy is not negligible in India. The estimate of sacrifice ratios in terms of real GDP range from 0.16 to 0.17 depending upon the model employed. The calculation of sectoral sacrifice ratios show that all the three sectors are affected negatively and the largest impact is found in the manufacturing sector followed by the other two sectors, i.e., agriculture sector and service sector. The sacrifice ratio in the manufacturing sector is found to be 1.10 and 0.72 indicating huge negative impact of tight monetary policy. Similarly, for the agriculture sector the sacrifice ratios are 0.40 and the service sector shows sacrifice ratios of 0.36 and 0.37 respectively for different models. Further, from 10-year rolling estimation, we find that sacrifice ratios are time varying. The sacrifice ratio is declining in the last few quarters at the aggregate and sectoral levels indicating that disinflation could be less costly in recent times. However, the high disinflation cost experienced prior to the year 2015 can’t be neglected. So, it is important to take caution while interpreting the results of disinflation cost in India.

     

A simple test of exogeneity for recursively structured VAR models

The restriction of exogeneity of certain variables in structural VAR models is rarely tested for consistency with the actual data. The reason is obvious: such a test requires estimates of the structural parameters. This paper proposes a solution for models that assume long-run or contemporaneous recursive structures in identification. We show that in such cases, the exogeneity restriction can be assessed statistically using the well-known Granger non-causality test which is conveniently performed in the reduced-form VAR model. Two empirical examples are offered to demonstrate the usefulness of this result.     

Disinflationary boom in a price–wage spiral model

This paper analyses the impact of the disinflation policy timing on the sign and the magnitude of the sacrifice ratio in a modified price and wage staggered model of Blanchard (1986). When wages are updated every four quarters and prices every two quarters, we show that a “cold-turkey” disinflation is associated to an output boom when the policy is implemented during the last period of life of the wage contract and a recession in the other quarters.     

The relative importance of demand and supply shocks some international evidence

In this paper the contribution of demand and supply shocks of the fluctuation of output and employment is analyzed using quarterly data for four countries (USA, UK, FRG and Switzerland) covering the period 1966 to 1988 in a long run restricted VAR framework for employment, output, the interest rate and inflation. For the USA output and employment fluctuations seem to be dominated by supply shocks in theshort and long run, whereas interest and price level variations are dominated by demand shocks. The result obtained with German and Swiss data are, however, different with two respects: The contribution of demand shocks to output and employment fluctuations are substantial up to a time horizon of eight quarters and price level variations, are at least in the short run, dominated by supply disturbances. Finally, the analysis of the UK data shows the US pattern of results for employment and the price level, but demand shocks are important for output up to four quarters and interest rate movements are dominated by supply shocks.     

Forecasting the US real house price index: Structural and non-structural models with and without fundamentals

We employ a 10-variable dynamic structural general equilibrium model to forecast the US real house price index as well as its downturn in 2006:Q2. We also examine various Bayesian and classical time-series models in our forecasting exercise to compare to the dynamic stochastic general equilibrium model, estimated using Bayesian methods. In addition to standard vector-autoregressive and Bayesian vector autoregressive models, we also include the information content of either 10 or 120 quarterly series in some models to capture the influence of fundamentals. We consider two approaches for including information from large data sets — extracting common factors (principle components) in factor-augmented vector autoregressive or Bayesian factor-augmented vector autoregressive models as well as Bayesian shrinkage in a large-scale Bayesian vector autoregressive model. We compare the out-of-sample forecast performance of the alternative models, using the average root mean squared error for the forecasts. We find that the small-scale Bayesian-shrinkage model (10 variables) outperforms the other models, including the large-scale Bayesian-shrinkage model (120 variables). In addition, when we use simple average forecast combinations, the combination forecast using the 10 best atheoretical models produces the minimum RMSEs compared to each of the individual models, followed closely by the combination forecast using the 10 atheoretical models and the DSGE model. Finally, we use each model to forecast the downturn point in 2006:Q2, using the estimated model through 2005:Q2. Only the dynamic stochastic general equilibrium model actually forecasts a downturn with any accuracy, suggesting that forward-looking microfounded dynamic stochastic general equilibrium models of the housing market may prove crucial in forecasting turning points.     

An inference about the length of the time-to-build period

This study attempts to infer the length of aggregate time-to-build period by estimating DSGE models with different investment lags and comparing their fits to the data. The models considered in this study use two, four, six, and eight quarters of investment lags. The Bayesian estimation result indicates that the model with six quarters of investment lags fits the data significantly better than the others.     

Hysteresis vs. natural rate of US unemployment

This paper investigates the stochastic nature of the unemployment rate allowing for cross-section dependence from a panel of US state-level data. We first employ the PANIC method to identify the common and idiosyncratic components. Powerful recursive mean adjustment (RMA) methods are used to test for unit roots. We find significant evidence of a nonstationary common component when the data from the most recent recession are included. Even when stationarity is empirically supported, the bias-corrected half-life of the common component appears very long, casting doubt on the usefulness of the natural rate hypothesis.     

Forecasting recessions using financial variables: the French case

In this article, we focus on the ability of two financial variables—the yield curve spread and the euro–US dollar exchange rate—to predict French recessions over the period 1979–2010. First, we propose a turning point chronology for the French business cycle based on a classical conception of economic cycles and a non-parametric dating algorithm applied to the real GDP series. Second, static and dynamic probit models are developed and estimated to produce the recession probabilities. In-sample results show that the dynamic specification performs better than the static one and, above all, that the exchange rate has a stronger predictive power than the yield curve. Out-of-sample results finally confirm the predominant role assigned to the exchange rate in predicting the latest recession occurred in 2008.     

Macroeconomic effects of fiscal policy in Pakistan: a disaggregate analysis

This study analyzes the macroeconomic effects of fiscal policy on macroeconomic variables in Pakistan. The study utilizes VAR framework and uses quarterly data of Pakistan from 1976Q1 to 2017Q4. The results showed that after an increase in government expenditures, private consumption and prices take three quarters to increase, while private investment follows the declining trend. Private consumption and interest rate are negatively related with taxes, while private investment and prices are positively related with taxes. Real GDP, private consumption and interest rate respond positively with an increase in developmental expenditure, public consumption and public investment. Private investment is negatively related with development expenditure, public consumption and public investment. An increase in direct tax as well as indirect tax leads to an increase in real GDP, private investment and interest rate, while private consumption takes three to five quarters to responds. Private investment and prices are positively related with non-tax revenue, while real GDP, private consumption and interest rate are negatively related. These results support the Keynesian view that government expenditure and taxes are useful tools to stimulate the economic activity, while crowding-out hypothesis holds in Pakistan as well. An active and efficient role of government is required for macroeconomic stability.     

Potential Contagion Effects on OECD Countries: A FAVAR Model under Bayesian Framework

This paper studies whether and how US shocks impact the OECD countries in the case of a simulated crisis. Using Bayesian estimation methods we extract constrained factors (global, country and variable type specific) from a sample of 153 economic and financial OECD variables from 1980–2008. These factors are the transmission channels through which national shocks spread to other countries, as in a pandemic. The Bayesian interpretable factors are used to estimate FAVAR models. Our main findings suggest that differences exist in the contagion effects. This implies that no generalizations can be made for OECD countries even of equal economic size and in the same geographic region. In addition, our results show that a large portion of the variance of domestic economic variables is explained by global factors; and that the interest rate shock appears to play an important role in the spillover mechanism from the United States to the rest of the world. More precisely, Australia, the United Kingdom and Scandinavian countries appear to be most sensitive to the US shocks.     

A Litterman BVAR approach for production forecasting of technology industries

Forecasting the production of technology industries is important to entrepreneurs and governments, but usually suffers from market fluctuation and explosion. This paper aims to propose a Litterman Bayesian vector autoregression (LBVAR) model for production prediction based on the interaction of industrial clusters. Related industries within industrial clusters are included into the LBVAR model to provide more accurate predictions. The LBVAR model possesses the superiority of Bayesian statistics in small sample forecasting and holds the dynamic property of the vector autoregression (VAR) model. Two technology industries in Taiwan, the photonics industry and semiconductor industry are used to examine the LBVAR model using a rolling forecasting procedure. As a result, the LBVAR model was found to be capable of providing outstanding predictions for these two technology industries in comparison to the autoregression (AR) model and VAR model.     

Economic policy uncertainty,financial markets and probability of US recessions

We use probit recession forecasting models to assess the ability of economic policy uncertainty indexes developed by Baker et al. (2013) to predict future US recessions. The model specifications include policy indexes on their own, and in combination with financial variables, such as interest rate spreads, stock returns and stock market volatility. Both in-sample and out-of-sample analysis suggests that the policy uncertainty indexes are statistically and economically significant in forecasting recessions at the horizons beyond five quarters. The index based on newspaper reports emerges as the best predictor, outperforming the term spread at the longer forecast horizons.