Research

Publications:

1) Does Common Ownership Concentration Affect Corporate Payouts? (Working paper)

I analyze the impact of common ownership concentration on corporate payouts. I hypothesize that when a set of investors owns significant equity stakes in two or more firms within the same industry (common ownership), this decreases competition, raises markups and increases payouts. Consistent with my hypothesis, I find a statistically significant increase in both, corporate payouts and acquisitions, following an increase in industry-level common ownership. To establish causation I rely on the use of an instrumental variable based on a major mutual fund scandal, propensity score matching, and machine learning methodology. I further show that the effect of increased common ownership concentration on payouts is larger in industries that face relatively less competition from Chinese imports.

2) Evaluation of the Modernization Hypothesis in Post-Socialist European Economies (Working paper)

After the socialist system collapse in the Eastern Europe in 1989, the formerly socialist European economies underwent initial democratization. Since then while some of these economies gained in terms of both income and democracy, others experienced a decline in democratization. I consider the initial political uncertainty as a major initial condition that determined subsequent development trajectories of these countries. Initial political uncertainty (or ‘initial political disruption’) is a concept from political science that indicates the disruption due to the lack of formal institutions in the period between the socialist party collapse and emergence of a new non-communist parliament. The key contribution of my study is to suggests a quantitative measure of initial political uncertainty, and then to use it as a mediating variable in the regression of democracy on per capita income. My findings indicate that higher initial political uncertainty resulted in lower democratization and prevented positive effect of income growth on democracy.

3) Gray Divorce: Its Causes and Consequences (Working paper)

Over the past three decades divorce rates among the elderly have more than doubled and exceeding the divorce rates of younger cohorts. This could have negative impact on welfare and health outcomes of the elderly. The study explores the effect of ‘gray divorce’ on wealth of splitting couples and on their individual health outcomes. The results indicate that ‘Gray divorce’ leads towards faster depletion of financial resources and also has negative health implications. The negative effect has its highest magnitude at the time of the divorce, but its effect weakens afterwards. Machine learning methodology, including random forest and sequential neural network classifiers, were used to predict 'gray divorces'. This showed that the key predictors of a divorce include household wealth and differences between the spouses in levels of their fitness, age and education.

4) Resilience of the Russian Mafia: An Empirical Study, with Federico Varese and Jakub Lonsky (The British Journal of Criminology, 01/2021. Vol. 61-1, pp. 143–166, https://doi.org/10.1093/bjc/azaa053)

Criminal organizations constantly face challenges that threaten their existence. What makes an organization survive amidst such threats and confrontations? The paper begins with a discussion of the effects that state repression and state transformation might have on criminal organizations, and how such organizations might respond. We then turn to the case of the Russian mafia, known as the vory-v-zakone. We identify the key challenges faced by the vory, examine how the Russian mafia adapted to such threats. We conclude that the most significant threat occurred at the end of the Soviet Union and show that the Russian mafia adapted to new circumstances and changed elements of its admission ritual without significantly changing its organizational structure. We also show that the Putin era has not been as damaging to the criminal fraternity as some observers have argued. We conclude that the biggest threat came not from state policies but state transformation. The paper is based upon a new and unique dataset we constructed containing biographical information of more than 5,000 members of the vory fraternity. The paper contributes to the study of organizations, the effect of state policies on mafia groups, and the history of the Russian mafia.

5) Estimating marginal effects in machine learning models (revision in progress, available on request)

The use of machine learning models has gained its popularity among researchers and practitioners due to their ability to achieve high forecasting performance and their focus on the out-of sample prediction. Still there is lack of applications of marginal effects in relation to machine learning models. Marginal effects can be relevant for decision makers when evaluating contribution of a given factor towards some outcome. As machine learning models capture significant non-linearity among the variables, marginal effects are different at different points of variables distribution. The current study suggests an approach, and develops a Python module, to estimate marginal effects at different points of the feature distribution, together with their standard errors. Simulation analysis is performed to evaluate the performance of the suggested method.

6) Evaluation of Constrained Stock Pairs-Trading Strategies (revision in progress, available on request)

The study investigates performance of a pair-trading stock investment strategy with and without short selling restriction. Pair-trading strategy is an application of statistical arbitrage, it aims to identify a pair of closely related stocks, then take long position in the temporarily under-priced stock combined with a short position in the temporarily over-priced one. While performance of pair-trading strategies was previously explored, there is lack of analysis of the strategy performance under the no short selling restriction. The analysis was performed based on the S&P500 constituent stocks, and its results indicate solid performance of the strategy even under the no short-selling constraint.

7) Stock Trading Recommender System: Development and Testing Stock Trading Algorithm, with Turki Alenezi and Yu Wang (Working paper)

Efficient market hypothesis suggests that past information cannot be used to consistently predict future stock prices. Our study uses machine learning methodology to predict future stock price movements (returns), based on the past information in relation to the stock returns, volatility of returns, trading volumes and stock sentiment. The latter was estimated by textual analysis of Twitter posts regarding each considered stock. Performance of various machine learning algorithms was compared based on their ability to correctly classify the sign of the future stock returns. Also, based on estimated probabilities of the positive future returns for each stock, a relevant investment strategy was developed and back-tested. Our results indicate that past information does have some power to predict future stock returns, which is not in line with the efficient market hypothesis.