PhD Lunch Seminars Amsterdam

Magdalena Rola-Janicka (University of Amsterdam) and David Garces Urzainqui (VU Amsterdam)
Tuesday, 27 June 2017

12:00-13:00 Magdalena Rola-Janicka (University of Amsterdam)
Supply of Bank Funding and Risk-Taking

We develop a theoretical framework to study how an expansion of bank funding supply can lead to a build-up of fragility in the banking sector. There is an informed and an uninformed bank, which choose an optimal mix of investment into risky asset, storage and safe technology with decreasing marginal returns. When funding supply is high the informed bank will invest in the risky asset and make an inefficiently low investment into safe technology. The uninformed bank infers the productivity from the total investment of the informed bank. Uncertainty about productivity emerges as the uninformed bank does not observe the supply of funding to the other bank. When supply is high and productivity is low the uninformed bank overestimates the productivity; if overestimation is sufficiently high, it will default at the final date. (joint with Enrico Perotti)

13:00-14:00 David Garces Urzainqui (VU Amsterdam)
Poverty Transitions Without Panel Data? An Appraisal of Synthetic Panel Methods

The scarcity of nationally representative panel data has historically held back the study of welfare dynamics in developing countries. Recently, different synthetic panel methods have been advanced that claim to provide point estimates of mobility on the basis of only two rounds of a cross-section survey. This paper compares the performance of the approaches suggested in Dang and Lanjouw (2013) and Bourguignon and Moreno (2016) in the measurement of transitions in and out of poverty against the benchmark of actual panel data from Thailand for years 2006 and 2007. Given a good estimate of income residual autocorrelation  ρ, a synthetic panel approach that allows for flexible treatment of residuals and opts for parsimonious income models delivers estimates very close to the panel. Results from  Monte-Carlo simulations and the validation exercise reveal that only pseudo-panel techniques based on cohort means, as opposed to cohort variances, may produce good estimates of this key parameter ρ. This is indeed the case for the current setting, but the methods rely on strong assumptions on the comparability of income residual persistence at the cohort and the individual level that may well not be fulfilled in other contexts.