Placement Director: Vijay Krishna
    (814) 863-8543
    vkrishna@psu.edu

Contact Information:
Li Wang
  Office: (814) 863-2657
  Cell: (814) 321-3696
E-mail: luw119@psu.edu
Website:http://www.personal.psu.edu/luw119/

CITIZENSHIP:

• China, F-1 Visa

EDUCATION:

• Ph.D., Economics, Penn State Univ., May 2008 (expected)
• M.A., Statistics, Penn State Univ., Dec. 2007 (expected)
• M.A., Finance, Beijing Univ. , June 2002
• B.A., Economics and English Literature, Beijing Univ., June 1999

PH.D. THESIS:

• “Integrated Conditional Moment Tests for Parametric Conditional Distributions”
Thesis Advisor: Professor Herman Bierens

 

FIELDS:

• Primary: Econometrics
• Secondary : Finance, Industrial Organization

PAPERS:

• “Integrated Conditional Moment Test for Parametric Conditional Distributions”, November 2007 (job market paper)
• “Consistent Integrated Conditional MomentTests for Parametric Conditional Distributions  with Infinitely Many Conditioning Variables”,  December 2007
• “Estimating the Dynamic Stochastic General Equilibrium  Models via Simulated Integrated Conditional Moment Method ” (work in progress)

GRANTS &
FELLOWSHIPS:

• Dissertation Support Grant, College of Liberal Arts, Penn State University, 2007
• Graduate Scholar Award, Penn State Univ., 2002-2005
• University Graduate Fellowship, Penn State Univ., 2002-2003
• ANTNA Scholarship, Beijing University , 2000
• Social Work Award, Beijing University , 2000

TEACHING EXPERIENCE:

• Instructor: taught 3 semesters of Introductory Econometrics
• Teaching Assistant: Graduate and Undergraduate econometrics

PRESENTATIONS & OTHER PROFESSIONAL ACTIVITIES:

• Internal Workshop, Economics Department, Penn State , 2005

REFERENCES:

• Professor Herman Bierens, Penn State University.
• Professor Paul Graf, Penn State University
• Professor Thomas Hettmansperger, Penn State Univ. (Statistics)

THESIS ABSTRACT

The correctness of the conditional distribution specification is crucial for maximum likelihood inference. This paper proposes a consistent test for the validity of parametric specifications of conditional distributions. Using the fact that distributions of bounded random variables and vectors are completely determined by their characteristic functions on an arbitrary compact set, the test is formed on the basis of the squared difference between the empirical characteristic function of the actual data and the characteristic function implied by the model , integrated over a compact set. The asymptotic null distribution of the test is a weighted sum of independent chi-square distributions, and its critical values are derived via parametric bootstrap. The test is consistent against all alternatives.

Since the conditional characteristic function of the model distribution has to be derived by integration, I will propose a simulated integrated conditional moment (SICM) test, where each theoretical conditional characteristic function is replaced by a simulated counterpart, based on a single random drawing from the corresponding conditional distribution. All the theoretical properties of the previous ICM test carry over with a much faster computation speed.

In contrast with other existing consistent tests, the SICM test is easier to compute, has a faster rate of convergence than non-parametric approaches, and does not have continuity restrictions on the conditional model distribution. Simulation results show that this test works very well.

Essay 2. “Consistent Integrated Conditional Moment Tests for Parametric Conditional Distributions with Infinitely Many Conditioning Variables”

Time series models aim to represent conditional distributions, or conditional expectations or moments, relative to the entire past of the time series involved, even when the actual model employs only a finite number of lagged conditioning variables. Therefore, in the time series case, tests that use just a few conditioning lagged variables can be inconsistent. There are only a few papers that deal with this tricky issue of conditioning on infinite past, but limited to conditional expectation models or continuous conditional distribution models.

This paper utilizes the Bierens' ARMA memory index approach to get around this curse of dimensionality. This ARMA memory index is a weighted average of truncated past variables which under some mild conditions reduces the dimension from infinity to one without losing any information. The ICM test in the previous paper is extended to a test for the parametric conditional distribution relative to the entire past of the time series involved, where instead of the infinite dimensional past this one-dimensional ARMA memory index is used. It is shown that the test is consistent, has nontrivial power against n^(-1/2)-local alternatives, and has a similar null distribution as in the previous case.

Simulation results show that this test works very well. This test is applied to GARCH models of stock market returns.

Essay 3. “Estimating and Testing Dynamic Stochastic General Equilibrium Models via Simulated Integrated Conditional Moment Method” (work in progress)

The DSGE model is a powerful workhorse for modern macroeconomics, but how to take real data into the DSGE model besides calibration remains a challenging issue, due to the model's singular joint distribution. This paper estimates a DSGE model, namely the King-Plosser-Rebelo (KPR) model, by the SICM method proposed in the previous papers. One of the policy functions involved in the model is parametrized to recursively solve the theoretical DSGE model, and these policy parameters are estimated together with the parameters of the DSGE models via SICM, where the conditional empirical characteristic function of the real data is compared to the theoretical characteristic function implied by the DSGE model. Statistical inference can be obtained through parametric bootstrap.