A modification of the CUSUM test in the linear regression model with lagged dependent variables

We consider testing for structural change in a dynamic linear regression model, and show that the well known CUSUM test, which has been initially devised only for the standard static model, can easily be modified such as to remain asymptotically valid also in this nonstandard situation.     

A new test for structural change

In this paper we are interested in detecting structural change at an unknown point. We argue that the CUSUM test may not ideal for this, and propose an alternative test. Critical values for this test are determined based on the multiple studentt test procedure. A Monte Carlo study suggests that the new test is quite powerful when the structural change occurs in the latter part of the sample. The new test also provides information about the location of structural change.     

Power monotonicity in detecting volatility levels change

We show that the CUSUM and LM tests for structural change in the volatility process enjoy monotonic power. The framework is general including many recently proposed non-stationary GARCH-type models. The result is in contrast to the well-known issue of non-monotonic power for the CUSUM-based tests for changing mean. Simulations and an empirical example provide further support.     

Power properties of the CUSUM and CUSUMSQ tests for parameter instability

This article investigates the power of CUSUM and CUSUMSQ tests for parameter stability and demonstrates that this depends on the nature of the structural change taking place. If the break is in the intercept of the regression equation then the CUSUM test has higher power. However, if the structural change involves a slope coefficient or the variance of the error term, then the CUSUMSQ test has higher power. This may help to explain why the two tests often produce contradictory findings.     

Financial development and economic growth in Australia: An empirical analysis

The paper empirically examines the dynamic relationship between financial development and economic growth in Australia in terms of bank-based and market-based financial structure. A time-series approach using the VAR Model is used to provide evidence for the dynamic relationship. The paper provides empirical evidence on the causal impact of the financial market on the economic growth of the Australian economy. The results suggest that financial intermediaries and financial markets have different impacts on economic growth given their diverse roles in the domestic economy. In particular there is evidence of causality from economic growth to the development of the financial intermediaries. On the other hand, development in the financial markets causes economic growth but there is no evidence of any causality from economic growth to financial markets. The sensitivity test using different interest rates does not change the results.I Jel classification: O16, G18, G28I We would like to thank Tilak Abeysinghe and Rajagurn Gunasekaran for their helpful comments on the first draft. Also, we would like to thank the Editor, Prof. Baldev Raj, and two anonymous referees for their helpful comments.First version received: October 2001/Final version received: October 2002     

A CUSUM test for a long memory heterogeneous autoregressive model

A CUSUM test is proposed for testing structural breaks in a long-memory heterogeneous autoregressive model. The limiting distribution of the CUSUM test is shown to be a simple function of a standard Brownian bridge, contrasting with the nuisance parameter dependent asymptotics of other CUSUM tests based on fractional integration models. A Monte-Carlo experiment investigates finite sample size and power of the test. The proposed test is applied to a set of daily realized volatilities of the log-return of the Korean Won US Dollar exchange rate to reveal some evidence of a break in addition to a long-memory.     

Peaks or tails – What distinguishes financial data?

We argue against the view that it is mostly the peaks of the empirical densities of stock returns (and of other risky returns as well) that set such data aside from “normal” variables. We show that peaks depend on sample size and on the way returns are standardized, and that for given data sets of stock returns, both higher peaks and lower peaks than in a standard normal case can be obtained. First version received: March 1998/Final version received: April 2000     

The sensitivity of the RESET tests to disturbance autocorrelation in regression analysis

This paper investigates the sensitivity of the RESET tests, proposed by Ramsey (1969) and modified by Thursby and Schmidt (1977), to disturbance autocorrelation in regression analysis. Porter and Kashyap (1984) show that RESET is not robust to autocorrelated disturbance when there is a highly autocorrelated regressor in the model. We show that RESET is sensitive to disturbance autocorrelation even when the regressors are not autocorrelated. We explain the findings of Thursby (1979) and Porter and Kashyap (1984) as well as our result by showing that a spurious correlation between the regressor and the disturbance is responsible for the serious size distortion of the RESET tests. First version received: June 1999/Final version received: November 2000     

Structural Changes, Common Stochastic Trends, and Unit Roots in Panel Data

This paper studies the problem of unit root testing in the presence of multiple structural changes and common dynamic factors. Structural breaks represent infrequent regime shifts, while dynamic factors capture common shocks underlying the comovement of economic time series. We examine the modified Sargan–Bhargava (MSB) test in the panel data setting and propose ways to handle multiple structural changes and dynamic factors. Properties of the MSB test under these non-standard conditions are derived. For example, the test statistics are shown to be invariant, in the limit, to mean breaks. This invariance does not carry over to breaks in linear trends, where the test statistics will converge to functionals of weighted Brownian bridges. A simplified test statistic is then proposed, which is invariant to both mean and trend breaks. We further study pooled test statistic based on standardization and combination of p -values. Response surfaces for p -values of all test statistics are computed to facilitate the empirical implementation of the proposed methodology. The pooled tests are shown to have good finite sample performance.     

Lead-lag effects in the mean and variance of returns of size-sorted UK equity portfolios

We test for lead-lag effects in the mean and variance among size-sorted portfolios for the UK stock market. We construct three sets of portfolios, namely a set of size-sorted equally-weighted portfolios of different capitalization size, a set of size-sorted value-weighted portfolios of different capitalization size, and a third set of portfolios of the same capitalization size. The recently proposed Cross Correlation Function test is employed. For both sets of portfolios with different capitalization size, we find evidence of a lead effect in both the mean and the variance from large-firm portfolios to small-firm portfolios. This result does not depend on the weighting scheme used to construct portfolios, and indicates that contrarian trading strategies on large-firm portfolios are profitable. For portfolios of equal capitalization size, there is hardly any evidence of a lead-lag effect in either the mean or the variance. This suggests that the lead-lag effect is due to the difference in the capitalization size among portfolios.I wish to thank two anonymous referees for helpful comments on a previous draft of this paper. The usual disclaimer applies.ing scheme.First version received: February 2002/Final version received: May 2003     

Nonlinearity in dynamic adjustment: Semiparametric estimation of panel labor supply

We estimate a semiparametric dynamic panel data model by the local linear kernel method and we interpret the slope of the nonparametric component function as a varying slope coefficient. Thus, the slope coefficient is a smooth, but otherwise unknown, function of some of the regressors. A Monte Carlo experiment is reported to examine the finite sample performance of the local linear estimator. We apply the estimation method to a labor supply equation for men from the triannual Survey of Income and Program Participation (SIPP). Specification tests based on the estimated labor supply elasticities, partial adjustment coefficients, and residuals demonstrate the improvements from a semiparametric partially linear model. Our empirical results point to a need by economists to revisit the issue of the speed of labor market adjustment to policy induced shifts in labor demand and to take more formal econometric account of heterogeneity in wage effects when studying the distributional consequences of tax reforms for labor supply earnings. First version received: July 2000/Final version received: January 2001     

Dynamic effects of public investment: Vector autoregressive evidence from six industrialized countries

We analyze the dynamic relationship between public investment and output. Whereas existing empirical studies on the effects of public capital typically rely on single-equation models of the private sector, we investigate the role of public investment in an economy by examining impulse responses derived from vector autoregressions. Using data from six industrial countries, we specifically examine the following questions: does higher public investment lead to GDP increases; is there reverse causation from output to public investment; and what are the effects of expenditure-neutral budget shifts from public consumption to public investment. First version received: April 1999/Final version accepted: August 2000     

The dynamic specification of the modified pecking order theory: Its relevance to Italy

This paper proposes an empirical model for the modified pecking order theory (MPO) in which both trade-off (TO) and pecking order (PO) models are nested. The MPO model is specified as an error-correction mechanism and applied to a vast panel data-set. Unlike previously estimated financial models, it avoids a number of problems: the mis-specification of dynamics, the approximation of the target leverage using the historical mean, the constrained estimation of the free cash flow components in a unique parameter. The MPO model is particularly good at explaining “hybrid” systems (neither market-based nor bank-based) such as the Italian one, in which companies are a mixture of two types: TO-type firms with a long-term optimal debt ratio towards which they converge; PO-type firms for whom the short-term availability of internal funds for investment may interfere with the process of adjustment towards the target leverage. Finally, the MPO model enables us to separately test the individual relevance of each of the restricted (“pure”) TO and PO models: results confirm their mis-specification and clearly point towards the excellent empirical performance of the MPO model. First version received: May 2000/Final version received: September 2000     

Testing for negativity in a demand system: A Bayesian approach

The paper introduces Bayesian inference into a demand model. This allows us to test for the negativity condition of the substitution matrix which is difficult to handle directly in the traditional approach. To illustrate the Bayesian inference procedures, we estimate the Rotterdam model and test the demand properties using Japanese data. The empirical results show the importance of specifically considering negativity in demand analysis. First version received: September 1997/final version received: February 1998     

Do stock market returns predict changes to output? Evidence from a nonlinear panel data model

Recent empirical work suggests a predictive relationship between stock returns and output growth. We employ quarterly data from a panel of 27 countries to test whether stock returns as useful in predicting growth. Unlike previous research, our approach allows for the possible non-linear effect of recessions on the growth-return relationship. There is strong evidence to suggest that a linear model would be misspecified and provide potentially misleading inference. Using a switching regression approach, we find evidence that returns are most useful in predicting growth when the economy is in recession.First version received: November 2002/Final version received: April 2003This paper benefited greatly from discussions with Kalvinder Shields, Mark Harris, Pete Summers, and Vance Martin. Two anonymous referees provided useful comments on an earlier version of the paper for which we are grateful. The usual disclaimer applies to any errors and omissions. Funding from The University of Melbourne greatly assisted in the completion of this paper.     

Fourier ADL cointegration test to approximate smooth breaks with new evidence from Crude Oil Market

Economic data is typically subject to a number of different forms of structural breaks. Ignoring structural breaks in a model can lead to misspecification issues and false conclusions. This paper proposes a new Autoregressive Distributive Lag (ADL) cointegration test in the presence of nonlinear breaks approximated by a Fourier function. The test offers a simple way to capture smooth structural change in time series data. Exact break dates are not required, and the suggested methodology can accommodate unknown number and form of gradual structural change. The testing procedure circumvents the potential power loss which can result from adding more dummy variables in the testing equation. Simulation results show that our procedure has good size and power properties. We demonstrate our test on the empirical example of real oil prices, oil production, and real economic activity, which are subject to structural breaks. The new test suggests that variables are cointegrated, while a conventional ADL test ignores structural breaks and concludes the opposite. This result casts some doubt on conventional oil price models.     

Long-run matching relationship in the Japanese labor market

This paper investigates the long-run relationship among new hiring, unemployment (job seekers), and unfilled vacancies in Japan, using an annual panel data on 47 prefectures for 1972-1999. We find that these three variables are I(1) processes, and are cointegrated in our panel data. Further, we estimate the panel cointegration equation derived from a Cobb-Douglas matching function by the heterogeneous fully modified OLS and heterogeneous dynamic OLS. The estimation results reveal that conventional within estimates could have non-negligible biases.First version received: November 2002/Final version received: October 2003All correspondence to Shigeki Kano. The authors are grateful to the associate editor and two anonymous referees of this journal for helpful comments. Doctor Stephen J. Turnbull (University of Tsukuba) is also acknowledged for correcting English errors in this paper. Remaining errors are due to the authors. The data set and GAUSS programming code used in this paperare available upon request.     

Okun's Law Does the Austrian unemployment–GDP relationship exhibit structural breaks?

Okun's Law postulates an inverse relationship between movements of the unemployment rate and the real gross domestic product (GDP). Initial empirical estimates for US data indicate that a two to three percent GDP growth rate above the natural or average GDP growth rate causes unemployment to decrease by one percentage point and vice versa. In this investigation we check whether this postulated relationship exhibits structural breaks by means of Markov-Chain Monte Carlo methods. We estimate a regression model, where the parameters are allowed to switch between different states and the switching process is Markov. As a by-product we derive an estimate of the current state within the periods considered. Using quarterly Austrian data on unemployment and real GDP from 1977 to 1995 we infer only one state, i.e. there are no structural breaks. The estimated parameters demand for an excess GDP growth rate of 4.16% to decrease unemployment by 1 percentage point. Since only one state is inferred, we conclude that the Austrian economy exhibits a stable relationship between unemployment and GDP growth. First version received: January 2000/Final version received: October 2000     

The relative weight of manufacturing and services in Europe: An innovation perspective

This paper uses an innovation-based taxonomy of both manufacturing and service industries to assess the role of the process of structural change of the last 25 years on the rate of productivity growth in Europe, US and Japan. The empirical analysis exploits the shift-share methodology for a decomposition of aggregate labor productivity growth. A modified version is applied, that allows to interpret whether employment has shifted to higher or lower productivity sectors. The results are discussed in light of the role that the different industries play according to the innovation-based taxonomy.     

Cointegration and common trends on the West German labour market

In this paper we analyze the West German labour market by means of a cointegrated structural VAR model. We find sensible stable long-run relationships that are interpreted as a labour demand, a wage setting and a goods market equilibrium. In order to study the dynamic behaviour of the model we identify two common trends that push unemployment. We find that goods market shocks have only transitory impacts on unemployment. In the long run, it is almost equally determined by technology and labour supply factors. However, transitory shocks have major importance in the shorter run since adjustment processes are rather sluggish. First version received: Sept. 1998/Final version accepted: Feb. 2000