Skip to main content

As of December 19th, 2016 Xinxi Song is no longer a current member of the department

Xinxi Song

PhD candidate


I am on the economics job market in 2014-15, and will be available for interview at the ASSA meeting in Boston and the RES meeting in London.

Research Interests

Microeconomic theory; Applied microeconometrics

Supervisors: Prof. Herakeles Polemarchakis and Dr. Andres Carvajal

Working Papers

"Recovering risk and ambiguity aversion: theory and evidence" (Job Market Paper)

Abstract: Although ambiguity preference has been widely used in finance and macroeconomics, there is only few experimental evidence on individual risk and ambiguity aversion. For the first time, this paper systematically investigates the nature of household ambiguity preference using household survey data. We derive explicit solution in a two-period smooth ambiguity model (Klibanoff, Marinacci and Mukerji, 2005), and show that time preference, risk aversion and ambiguity aversion can be uniquely identified from a special panel dataset. Using Bank of Italy Survey on Household Income and Wealth (SHIW) 2008 and 2010, which contain detailed information on household consumption, saving, portfolio, and stock return expectation, we show, firstly, individual expectation is very pessimistic, and is subject to much ambiguity; secondly, constant relative risk aversion and relative ambiguity aversion can be a good approximation; thirdly, the recovered preference parameters display quite heterogeneity, the average relative risk aversion is much smaller than 1, and the average relative ambiguity aversion is around 3 or larger; fourthly, the overidentification restriction implied by subjective expected utility model rejects the null hypothesis that households are subjective expected utility maximizer, in favor of ambiguity model; finally, household risk aversion and ambiguity aversion are not correlated, can not be explained by observable household characteristics, and quantitatively have significant effect on consumption and portfolio holding.

"The identification of uncertainty and risk" (with Herakles Polemarchakis and Larry Selden)

Abstract: Uncertainty is not risk, and individuals behave differently when they know the objective probability (under risk) than when they are ignorant (under uncertainty). This paper establishes the identifiability of individual risk and ambiguity preference from his consumption and/or portfolio choice. Assuming individuals are endowed with smooth ambiguity preference (Klibanoff, Marinacci and Mukerji, 2005), we show that if there exist a riskfree and an ambiguityfree asset, and the matrix of expected return has full row rank, then individual risk aversion index and uncertainty aversion index can be uniquely recovered from his portfolio choice. If there does not exist ambiguityfree asset, but we can observe individual consumption in addition to portfolio choice, recovery can be restored. However, without riskless asset, recovering individual risk and ambiguity preference requires the underlying preferences being analytic at 0 point.

"Testing empirical content of Pareto optimality and competitive equilibrium with public goods" (with Andres Carvajal)

In this paper, we test the empirical implications of Pareto optimal provision of public goods and that of competitive equilibrium with public goods. The literature is full of theoretical prediction that competitive market will fail in presence of public goods, and the resulting allocation will be Pareto suboptimal; however, to emprically test such prediction is difficult and evidence is sparse. Based on Carvajal (2010) and Snyder (1999), we characterize by Mixed-integer programming proposed by Cherchye et al (2009) necessary and sufficient conditions for observable data consisting of market prices and individual endowments to be consistent with Pareto optimal provision of public goods and competitive equilibrium with public goods. Since these necessary and sufficient conditions consist of nonlinear inequalities, Mixed-integer programming makes checking these conditions easier. Our test is nonparametric, and is immune to misspecification of underlying preference and technology function. From the result of implementation, we can see whether market mechanism fails in presence of public goods and whether there exists other mechanism to achieve Pareto efficiency.


EC 226 Econometrics

Contact details


Email: X dot Song at warwick dot ac dot uk


Office Hours:

Friday 11:30-12:30

Other information

(PDF Document)CV