Spyros Skouras

Electoral Misgovernance Cycles: Evidence from Wildfires and Tax Evasion in Greece and Elsewhere

We present detailed empirical evidence that around Greek elections, misgovernance results in significant increases in wildfires and tax evasion and with important economic implications: the cumulative cost of these effects in recent years has been over 8% of GDP and has therefore been a contributing factor to Greece’s debt crisis and any effect this has had on the global economy. We interpret this evidence as a type of misgovernance which arises from electoral cycles in two types of incumbent incentives: (i) to allocate effort or attention between governing vs. campaigning; and/or (ii) to adopt even very inefficient redistributive policies if they benefit special interests with a lead over when the costs are observed. While these incentives may manifest differently among countries, our analysis suggests that electoral cycles everywhere may be much more multifaceted and harmful than previous literature suggests.

Financial returns and efficiency as seen by an artificial technical analyst

Previous research has shown that simple trading rules can be useful tools for evaluating financial models. Here we introduce trading rules which are selected by an artificially intelligent agent who learns from experience - an Artificial Technical Analyst. We show that the rules used by this agent can lead to the recognition of subtle regularities in return processes whilst suffering from lesser data-mining problems than other rules commonly used as model evaluation devices. The relationship between the efficiency of financial markets and the efficacy of technical analysis is investigated and it is shown that the Artificial Technical Analyst can be used to provide a quantifiable measure of market efficiency. The measure is applied to the DJIA daily index from 1962 to 1986 and it is shown that a quantification of efficiency based on the profits of an Artificial Technical Analyst can lead to interesting results concerning the behaviour of other investors.

An ecological perspective on the future of computer trading

Computer trading in financial markets is a natural and inevitable consequence of technological progress, and is almost as old as computers themselves. Computers facilitate basic market activities s...

Gibrat's Law for (All) Cities: A Rejoinder

We establish that the debate between Eeckhout (2004; 2009) and Levy (2009) has still not resolved the key issue of whether the distribution of large US urban places in 2000 is consistent with a lognormal for the intire size range. We resolve this by introducing a new distribution function which switches between a lognormal and a power distribution and estimating it with the data used by Eeckhout and Levy (2009). We find that there is a sudden transition from lognormality to power behavior as city populations icrease above sudden transition from lognormality to power behavior as city populations increase above 100,000. Gibrats law holds for most cities but a power law holds for most of the population.

Markets Change Every Day: Evidence from the Memory of Trade Direction

We present empirical evidence that there are periodic, specifically daily, structural breaks in the trade direction time series process, a fact with implications for several key intra-day characteristics of markets. We suggest that breaks arise as a consequence of daily variation in order flow direction independently of intra-day events and as a consequence of a natural and widespread daily periodicity in the timing of investment decisions. Empirical implementation of our short memory AR model with daily level shifts captures the striking long horizon predictability of trade direction, performs better out-of-sample than the standard long memory ARFIMA alternative and is computationally easier to estimate.

Explaining Zipf's Law for US Cities

I develop an explanation of Zipfs law that is consistent with the observed marked heterogeneity in the growth of US cities. The explanation is that heterogeneous growth results in heterogeneous size distributions across cities, with the heaviest tailed distributions being Zipf and dominating the cross-sectional mixture distributions tail. I demonstrate in the context of a popular model that this explanation is consistent with observed growth heterogeneity and other key stylized facts about city size and growth. This explanation has significant policy implications for controlling size distributions and the size of the largest cities in a country.

An algorithm for computing estimators that optimize step functions

The average of a large number of random step functions produces a discontinuous surface with a large number of local optima although it may converge to a smooth surface with a unique optimum as the number of step functions tends to infinity. Such a function arises when certain types of econometric estimators are used, including variants of the maximum score estimator. I propose an algorithm for computing the optimum of such a surface, where standard gradient-based optimization methods are inapplicable. This algorithm replaces the discontinuous surface with a sequence of easily optimized continuous surfaces that converge to it. Sufficient conditions for the algorithm to converge to a global optimum are given and the algorithms performance is evaluated in a simple but relevant application.

The Sign of A Mean Regression: Characterisation, Estimation and Applications

There exist a number of important applications in which interest centers on the sign rather than the overall shape of a mean regression. These include certain decision-making problems as well as the problems of calibration and econometric equation inversion. We show that the sign of a mean regression for Y may be viewed as a generalisation of a quantile regression for the sign of Y. This characterisation is used to construct three related estimators that require only a very weak specification condition for the estimated models sign to be consistent for the regressions sign. The relative merits of each estimator are discussed and simulations are used to study one of the estimators.

Gibrat’s Law Does Not Imply Zipf’s Law

This note contests a widely cited, used and derived ‘result’ according to which Gibrat’s law implies Zipf’s law. I show that this result is only true under strict, unrecognized and ad hoc additional conditions some of which are even inconsistent with the economic models that have been used to explain Zipf’s law. The widespread confusion regarding the relationship between Gibrat and Zipf laws derives from the mathematically subtle effect on dynamics of lower boundaries in Gibrat processes. In order to make my argument concrete, I illustrate in the context of a paper widely credited for explaining Zipf’s law.

Exact Computation of Censored Least Absolute Deviations Estimators

We show that exact computation of Powells (1984) censored least absolute deviations (CLAD) estimator may be achieved by formulating the estimator as a solvable linear Mixed Integer Programming (MIP) problem with disjunctive constraints. We apply our approach to two previously used censored datasets and find that standard heuristic/approximate approaches to computation of the estimator can lead to substantial errors and misleading economic conclusions. Additionally, we present a small Monte Carlo study which illustrates that MIP computation using widely available solvers is practical and efficient for data sets of sizes typically encountered in econometric applications.