New Advancements of Scalable Statistical Methods for Learning Latent Structures in Big Data

Constant technology advances have caused data explosion in recent years. Accord- ingly modern statistical and machine learning methods must be adapted to deal with complex and heterogeneous data types. This phenomenon is particularly true for an- alyzing biological data. For example DNA sequence data can be viewed as categorical variables with each nucleotide taking four different categories. The gene expression data, depending on the quantitative technology, could be continuous numbers or counts. With the advancement of high-throughput technology, the abundance of such data becomes unprecedentedly rich. Therefore efficient statistical approaches are crucial in this big data era.

Previous statistical methods for big data often aim to find low dimensional struc- tures in the observed data. For example in a factor analysis model a latent Gaussian distributed multivariate vector is assumed. With this assumption a factor model produces a low rank estimation of the covariance of the observed variables. Another example is the latent Dirichlet allocation model for documents. The mixture pro- portions of topics, represented by a Dirichlet distributed variable, is assumed. This dissertation proposes several novel extensions to the previous statistical methods that are developed to address challenges in big data. Those novel methods are applied in multiple real world applications including construction of condition specific gene co-expression networks, estimating shared topics among newsgroups, analysis of pro- moter sequences, analysis of political-economics risk data and estimating population structure from genotype data.

Institutional Persistence in Eastern Europe: Economic and Political Legacies of Empires and Communism

This dissertation seeks to advance our understanding of the roles that institutions play in economic development. How do institutions evolve? What mechanisms are responsible for their persistence? What effects do they have on economic development?

I address these questions using historical and contemporary data from Eastern Europe and Russia. This area is relatively understudied by development economists. It also has a very interesting history. For one thing, for several centuries it was divided between different empires. For another, it experienced wars and socialism in the 20th century. I use some of these exogenous shocks as quasi-natural social experiments to study the institutional transformations and its effects on economic development both in the short and long run.

This first chapter explores whether economic, social, and political institutions vary in their resistance to policies designed to remove them. The empirical context for the analysis is Romania from 1690 to the 2000s. Romania represents an excellent laboratory for studying the persistence of different types of historical institutional legacies. In the 18th and 19th centuries, Romania was split between the Habsburg and Ottoman Empires, where political and economic institutions differed. The Habsburgs imposed less extractive institutions relative to the Ottomans: stronger rule of law, a more stable and predictable state, a more developed civil society, and less corruption. In the 20th century, the Romanian Communist regime tried deliberately to homogenize the country along all relevant dimensions. It was only partially successful. Using a regression discontinuity design, I document the persistence of economic outcomes, social capital, and political attitudes. First, I document remarkable convergence in urbanization, education, unemployment, and income between the two former empires. Second, regarding social capital, no significant differences in organizational membership, trust in bureaucracy, and corruption persist today. Finally, even though the Communists tried to change all political attitudes, significant discontinuities exist in current voting behavior at the former Habsburg-Ottoman border. Using data from the parliamentary elections of 1996-2008, I find that former Habsburg rule decreases by around 6 percentage points the vote share of the major post-Communist left party and increases by around 2 and 5 percentage points the vote shares of the main anti-Communist and liberal parties, respectively.

The second chapter investigates the effects of Stalin’s mass deportations on distrust in central authority. Four deported ethnic groups were not rehabilitated after Stalin’s death; they remained in permanent exile until the disintegration of the Soviet Union. This allows one to distinguish between the effects of the groups that returned to their homelands and those of the groups that were not allowed to return. Using regional data from the 1991 referendum on the future of the Soviet Union, I find that deportations have a negative interim effect on trust in central authority in both the regions of destination and those of origin. The effect is stronger for ethnic groups that remained in permanent exile in the destination regions. Using data from the Life in Transition Survey, the chapter also documents a long-term effect of deportations in the destination regions.

The third chapter studies the short-term effect of Russian colonization of Central Asia on economic development. I use data on the regions of origin of Russian settlers and push factors to construct an instrument for Russian migration to Central Asia. This instrument allows me to interpret the outcomes causally. The main finding is that the massive influx of Russians into the region during the 1897-1926 period had a significant positive effect on indigenous literacy. The effect is stronger for men and in rural areas. Evidently, interactions between natives and Russians through the paid labor market was an important mechanism of human capital transmission in the context of colonization.

The findings of these chapters provide additional evidence that history and institutions do matter for economic development. Moreover, the dissertation also illuminates the relative persistence of institutions. In particular, political and social capital legacies of institutions might outlast economic legacies. I find that most economic differences between the former empires in Romania have disappeared. By the same token, there are significant discontinuities in political outcomes. People in former Habsburg Romania provide greater support for liberalization, privatization, and market economy, whereas voters in Ottoman Romania vote more for redistribution and government control over the economy.

In the former Soviet Union, Stalin’s deportations during World War II have a long-term negative effect on social capital. Today’s residents of the destination regions of deportations show significantly lower levels of trust in central authority. This is despite the fact that the Communist regime tried to eliminate any source of opposition and used propaganda to homogenize people’s political and social attitudes towards the authorities. In Central Asia, the influx of Russian settlers had a positive short-term effect on human capital of indigenous population by the 1920s, which also might have persisted over time.

From a development perspective, these findings stress the importance of institutions for future paths of development. Even if past institutional differences are not apparent for a certain period of time, as was the case with the former Communist countries, they can polarize society later on, hampering economic development in the long run. Different institutions in the past, which do not exist anymore, can thus contribute to current political instability and animosity.