Self-Normalized KPSS Tests with Power Enhancement. (With Xiaojun Song)
Much research has focused on testing the null hypothesis of stationarity againstthe unit root alternative. In this paper, we propose a novel class of self-normalized
KPSS tests without needing a consistent estimation of the long-run variance. Under persistent autocorrelation, the widely used heteroskedasticity and autocorrelation
consistent long-run variance estimator proposed by Newey and West (1987), Newey and West (1994), and Andrews (1991) may not be reliable and often lead to tests
with size distortions and power losses in finite samples. In addition, the practitioner has to choose the truncation lag that is ultimately arbitrary. To improve the
finite sample performance of the stationarity tests and make them robust to realistic amounts of dependence, we propose the use of self-normalizing methods, for
example, range-based self-normalization by Hong et al. (2024) and fixed-b asymptotics by Kiefer and Vogelsang (2005) and Amsler et al. (2009) to control the effect
of the long-run variance. The self-normalized tests are inconsistent in that, under
the unit root alternative, they do not diverge as the sample size approaches infinity. To recover the tests’ consistency, we devise a mechanism similar to the power
enhancement mechanism proposed by Fan et al. (2015). Under the null hypothesis, this mechanism is asymptotically negligible. However, under the alternative
hypothesis, the mechanism diverges as the sample size increases. We show that
this mechanism also enhances the power of the self-normalized stationarity tests.
A simulation study and an empirical application demonstrate the merits of the
approach advocated.
Self-normalized Tests for Skewness, Kurtosis, and Normality for Time Series Data. (With Xiaojun Song)
Testing for skewness, kurtosis, and normality for time series data is highly relevant for
modeling and testing purposes in econometrics. It also affects our understanding of many
economic and financial phenomena and the validity of the models developed to explain them.
In this paper, we propose self-normalized tests for skewness, kurtosis, and normality that
can eliminate the effect of the long-run variance. In particular, our tests allow us to avoid
using the long-run variance estimator, which is poorly approximated in finite samples. Consequently, our tests rule out the need to choose the lag-truncation parameter. We present
general conditions on the self-normalization function and give two simple examples using
the fixed-b asymptotics and the simple normalization proposed by Lobato (2001). Monte
Carlo simulations show that the self-normalized tests for skewness and normality have good finite-sample size and power properties, while the test for kurtosis presents substantial size
distortions unless the distribution has thin tails like the normal distribution. Finally, we apply the tests to 18 macroeconomic and financial series to study their symmetry, kurtosis, and normality.
You can find the latest version at SSRN
Threshold Stochastic Unit Root Models.
In this study, we introduce a new class of stochastic unit-root (STUR) processes, where a threshold variable drives the randomness of the autoregressive unit root, thereby allowing us to explain the existence of unit roots.
This new model, namely the threshold autoregressive stochastic unit root (TARSUR) process, is strictly stationary, but if we do not consider the threshold effect, it can mislead to conclude that the process has a unit root.
The TARSUR models are not only an alternative to fixed unit root models but present interpretation, estimation and testing advantages with respect to the existent STUR models.
This study analyzes the properties of the TARSUR models and proposes two simple tests to identify this type of processes. The first test will allow us to detect the presence of unit roots, which can be fixed or stochastic,
and the asymptotic distribution (AD) of this test presents a distribution discontinuity depending if the unit root is fixed or stochastic. The second test we propose is a simple t-statistic (or the supremum of a sequence of t-statistics)
for testing the null hypothesis of a fixed unit root versus a stochastic unit root hypothesis. It is shown that its asymptotic distribution (AD) depends if the threshold value is identified under the null hypothesis or not.
When the threshold parameter is known, the AD is a standard normal distribution, while in the case of an unknown threshold value, the AD is a functional of Brownian Bridge. A Monte Carlo simulation shows that the proposed tests behave very well in finite sample,
and the Dickey-Fuller test cannot easily distinguish between exact unit roots and threshold stochastic unit roots. The study concludes with applications to U.S. stock prices, U.S. house prices, U.S. interest rates, and USD/Pound exchange rates.
Multiple Long Run Equilibria Through Cointegration Eyes.
Cointegration has succeeded in capturing the unique long-run linear equilibrium. Specific non-linearities have been incorporated into cointegrated models but always assuming the existence of a single equilibrium.
In this study, we explore the possibility of different long-run equilibria depending on the state of the world (i.e., good and bad times, optimism and pessimism, frictional coordination) in a threshold framework.
Starting from the present-value model (PVM) with different discount factors and depending on the state of the economy, we show that this type of PVM implies threshold cointegrated with different long-run equilibria.
We present the estimation and inference theory. The study completes two applications where the variables are not linearly cointegrated but threshold cointegrated.
Behavioral/Experimental Economics
Unveiling Lies in Disguise: A Test of Lying Aversion Theories. (With Jin-yeong Sohn, and Xu Cheng)
We provide an experimental test to distinguish two prominent theories of lying aversion: perceived cheating aversion (Dufwenberg & Dufwenberg 2018) and reputation for
honesty (Gneezy et al. 2018, Khalmetski & Sliwka 2019). We use a novel belief-elicitation method, which allows us to estimate the subjects’ strategies, i.e. probability of reporting
y ∈ {1, · · · , 6} conditional on die roll x ∈ {1, · · · , 6}. We also compare lying behavior across various non-linear payoff schemes. Our results support no-downward-lies and
uniform-cheating properties proposed by Dufwenberg & Dufwenberg (2018). We find partial support for an implication of reputation for honesty; people use maximal lies more
(less) frequently under a convex (concave) payoff scheme.
You can find the latest version at SSRN
Private Card-shredding Game: A Trade-off between Truth and Lies. (With Jin-yeong Sohn)
Fischbacher & Föllmi-Heusi’s (2013) die-roll paradigm provides an experimental method of measuring cheating behavior by comparing report distributions to the uniform distribution. We point out that, given the sample sizes commonly used in the
literature, this method may lead to erroneous inferences about lying behavior. To address this issue, we provide an alternative design called private card-shredding game.
We conduct an experiment using this alternative design and find that there is a tradeoff between the experimenters’ access to truths and lies; if they are to observe lies, they
must give up on knowing the empirical distribution of private signals.
