Difference-in-differences (DD) analysis is highly popular for empirical research that tries to establish a causal link between a treatment (often it is a policy) and a variable of interest. This method uses two samples - the treatment group that was influenced and the control group - to calculate the effect of treatment. Thus, for such analysis researchers use panel data (data about different cross-sections across time). Nevertheless, there are some limitations of this approach that became the topic of the study by Marianne Bertrand, Esther Duflo, and Sendhil Mullainathan.

Authors highlight several limitations of DD estimations including the endogeneity of interventions, but they mostly focus on other problems:

1. DD in literature relies on quite long time series;
2. Frequently used dependent variables are serially correlated;
3. Treatment variables change very little within the state over time.

All these factors lead to the fact that standard error for the estimated coefficient of interest could understate its standard deviation. And, according to the authors, many papers ignore adjusting for these issues.

To assess the scale of the problem authors look at how DD performs on placebo laws, which were randomly assigned to states and years. The authors use the dataset of female surveys about wages testing the null hypothesis of no effect. If OLS were to provide consistent standard errors, we would expect to reject the null hypothesis of no effect approximately 5% of the time, but it is not the case.

After making exercises with placebo treatment on wages data and the Monte Carlo studies (state-level empirical distribution of the survey data and AR(1) model with normal disturbances), authors show that conventional DD estimation leads to gross over-estimation of t-statistics and significance levels.

Then, (Bertrand & Duflo & Mullainathan, 2002) suggests possible solutions to described problems:

1. Specify an auto-correlation structure for the error term, estimate its parameters, and use these parameters to compute standard errors.
Results: This technique does little to solve the serial correlation problem: the rejection rate stays high at 24 percent (instead of the expected 5).
Why it does not work? The downward bias in the estimator of the auto-correlation coefficient, parametric correction may be that we have not correctly specified the auto-correlation process.

2. Block bootstrap (maintains the autocorrelation structure by keeping all the observations that belong to the same group together).
Results: work well for data with 50 groups (6.5% of rejections in survey data and 5% for CPS+ AR(1) data). But the quality of performance decreases with a lower number of groups (13% for 20 groups, 23% for 10)

3. Ignore the time-series information when computing standard errors. It suggests averaging data before and after treatment.
Results: work for laws passed at the same time. However, if it is not true, one may modify the technique: regress Y_st on group fixed effects, year dummies, and any relevant covariates, then divide the residuals of the treatment states into two groups: residuals from years before the laws, and residuals from years after the laws. The estimate of the laws’ effect and its standard error can then be obtained from an OLS regression in this two-period panel. The authors state that it performs well for a small number of groups (5.3% for 10 states). The disadvantage is that power is low and declines with sample size.

4. Empirical variance-covariance matrix.
Results: poor performance for small sample size (8% for 10 groups, 15% for 6). The limitation is that it is consistent under the assumption of cross-sectional homoskedasticity, which often is violated.

5. Arbitrary variance-covariance matrix (generalization of previous method).
Result: rejection rates increase significantly above 5% with the decline of the number of groups (11.5 % with 6 states, 8 % with 10 states) that is comparable with the empirical variance-covariance matrix correction method for small samples, less extreme than with block bootstrap, but higher than with the time series aggregation.

Thus, (Bertrand & Duflo & Mullainathan, 2002 ) highlight the problems of DD’s implementation that were also present (and, in many cases, were not addressed) in literature before. They explain the nature of over-rejection that takes place when researchers use the DD approach and provide empirical evidence for it via implementing placebo treatment on the data. Also, they propose 4 approaches to deal with described issues: block bootstrap, ignoring series information (aggregation), usage of empirical variance-covariance matrix, or arbitrary one.

Marianne Bertrand, Esther Duflo & Sendhil Mullainathan (2002) How Much Should We Trust Differences-in-Differences Estimates? NBER Working paper 8841 DOI 10.3386/w884, https://www.nber.org/papers/w8841
2003 version: https://economics.mit.edu/files/750

The Climate Casino is devoted to the question of climate change, in particular, W. Nordhaus discusses the science frontier on this topic as well as provides an analysis of possible solutions to problems related to the forecasted environmental situation. Overall, this book is a concise representation of the thesis about climate change from climate and social science, that helps to create a comprehensive vision on the issue. I will highlight some of them following the book’s chapters.

1. The first part describes the genesis of climate change - greenhouse gases, including CO2, accumulate in the atmosphere that leads to the warming of land and oceans. Change in temperature influences precipitation patterns, storm location and frequency, snowpacks, river runoff, water availability, and ice sheets. From this, we can expect impacts on biological and human activities that are sensitive to the climate. The important definition here is tipping points - thresholds of temperature level after which sudden or irreversible changes occur. Accumulation of GHG is related to emissions, or, more generally, anthropogenic interference.


2. The second is devoted to the impact on humans and other living systems. W. Nordhaus introduces the distinction between managed, where societies try to use resources in an efficient and sustainable way, for example, indoor living, and unmanaged systems that operate without human interactions, such as oceans. Agriculture, being an important sphere for humanity, could be highly managed. The climate change impact on it is unambiguous: from one hand, we can inspect growth in harvest under warmer conditions, but, on the other hand, it is true within some limits of temperature rise, also farmers should adapt to new conditions that also require investments.
Climate change could impact human health via increasing environmental stress, indirectly through lowering living standards due to an increase of the geographical range of some infectious diseases.
Regarding unmanaged threats, the author mentions sea-level rise (SLR), ocean acidification, hurricane intensification, and ecosystem losses.


3. The ways to slow climate change are discussed in the third part of the book. Three approaches are presented:

— Adaptation: learning to cope with climate change, but it won’t solve the problem;

— Geoengineering: introducing cooling systems - partially efficient way, but there are dangerous and uncertain consequences, therefore W. Nordhaus consider this approach as salvage therapy;

— Mitigation: reducing the concentration of GHG emissions (reduce energy consumption or storing emissions).

The important part here is the discussion of costs - new electricity could be more expensive as well as systems for storing, and if costs are high enough, then such policies won’t be effective. W. Nordhaus provides a detailed discussion on discounting and future value.


4. The practical side is located in the fourth part that discusses the issues of policy design and implementation. First, since climate change is a global problem, international policies should be adopted, but there are no successive examples in history so far due to the obvious opportunistic behavior of countries. As a solution for ensuring cooperative behavior, W. Nordhaus suggested the incorporation of trade sanctions for deviation from climate agreements. By the way, partially we see it in EU Carbon Border Tax. Also, there is a discussion of national policies, in particular - carbon tax, cap and trade, and intermediary measures. From pure economics, we expect the same outcome in equilibrium (raise carbon price), but there are some interesting differences, for example, establishing a market in allowances is to ensure that emissions are used in the most productive manner, but under this system, we expect higher volatility, but limited quantity, while carbon tax ensures stable price and unlimited emissions.

5. The last chapter captures issues related to public opinion on climate change and critiques the science of climate change. The authors emphasize the meaning of scientific consensus and how society’s opinion on this topic forms and changes over time. Also, some obstacles for climate policy implementation are mentioned: present bias, economic self-interest, unequal distribution of losses and gains.

To sum up, the topic of climate change is full of uncertainties, but there is scientific consensus that the rise of temperature takes place, and it relates to human actions. Due to its impact on living systems, it becomes a policy that requires proper measures and international cooperation for mitigating risks of most dangerous scenarios.


The Climate Casino: Risk, Uncertainty, and Economics for a Warming World
By William Nordhaus
2015
https://yalebooks.yale.edu/book/9780300212648/climate-casino

Uncommon Sense: economic insights, from marriage to terrorism.
(G. S. Becker & R. A. Posner, 2010)

From my experience, it can be difficult to find good popular scientific literature on economics. Uncommon Sense by G. S. Becker, R. A. Posner is a good example of a book that provides insightful economics perspectives on various social issues. The book is a set of authors’ blog posts accompanied by afterthoughts for each part. I want to highlight several features that I especially enjoyed:

1. The variety of topics covered.
In 7 parts authors discussed different facets of economic science providing clear arguments which draw on classic economic theory. There are themes of each part with some subtopics in parenthesis: gender issues (gay marriage, population balance), property rights (organs sales, files sharing, patents), universities (tenure, plagiarism), incentives (libertarian paternalism, taxes), labor economics (judicial term limits, income inequality), environmental economics (federalism, global warming, discounting), crime and terrorism drunk driving, preventive war), and global economics (international aid, microfinance). As one can see, the book comprises both classical economics questions and topics that are on the frontier.

2. Illustrative narration.
Since this book is popular science, there are not many literature references. Nevertheless, the authors provide multiple examples from real-world cases which allows the reader to look through the economic framework at real events. They also generally mention empirical literature as the evidence to their thoughts and suggest ideas for further research.

3. Discussion format and blog size.
Each question is discussed by both authors. In most cases, they agree with each other’s arguments and just add some thought and details to the provided arguments, but there are posts where they take opposite sides, and it is interesting to get a comprehensive opinion on a question from different perspectives. Also, each theme is discussed within the Also, each theme is discussed within the length of the blog post, which is pretty short. It makes authors concentrate only on the most important arguments leaving readers with a concise topic overview.

Overall, it is quite an interesting popular economic book that will be relevant to those who want to get a general overview of different economic topics or have a light (compared to scientific articles) but interesting reading.

One of the main problems that we face speaking about fighting with reduction of CO2 emission is the lack of coordination between countries in terms of policies. Due to the nature of the global warming issue international cooperation is a significant condition of success. Generally, people pay attention to direct restrictive measures, but J. S. Shapiro shed a line on how import tariffs and nontariff barriers (NTBs) affect the emissions of CO2. In particular, he showed that tariffs and NRBs are lower on dirty than on clean industries. This bias creates an implicit subsidy to CO2 emissions in internationally traded goods and so contributes to climate change.

In this paper J. S. Shapiro, first, shows this phenomenon, and, second, investigates political economy explanations for the covariance of upstreamsness and trade policy involving lobbying competition. According to the author, firms may lobby for high barriers on their own outputs and for low tariffs on the goods they use as inputs to decrease production costs. So politicians give the least protection to the upstream industries (dirtiest) and the greatest protection to the downstream industries (cleanest).

Analyzed the relationship between pollution and trade policy via econometric techniques. J.S. Shapiro introduced a more complex model that includes the following important features: pollution could influence utility, pollution creates transboundary damages, countries may have preexisting and suboptimal trade policy on any goods, tariffs generate revenue that is lump-sum redistributed, and industries are connected through input-output links, so dirty industries can be upstream. In the model, we have two symmetric countries and two industries (clean and dirty). Solving this model allows us to study counterfactuals both from theoretical and empirical perspectives.

Quantitative general equilibrium model was adapted to assess the effect of counterfactual trade policies on CO2 emissions and social welfare. The author analyzed the set of 6 counterfactual policies and got the following results:

1) If each country sets a single tariff per trading partner for all industries and implements a similar reform for NTBs, CO2 emissions will decrease while global real income remains unchanged or slightly increased which is the similar effect on CO2 as from the EU Emissions Trading System and the U.S. Waxman-Markey Bill.

2) If only the EU adopts the described policy, it will decrease global CO2 emissions by half the amount of the global policy with no significant changes in real income.

3)-4) Changing tariffs and NTBs to equal either the baseline level of the cleanest third or dirtiest third of industries decreases global CO2 emissions by several percentage points.

5) If every country adds a tariff proportional to goods’ CO2 intensity, we will get modest environmental benefits.

6) If countries completely eliminated tariffs and NTBs, both global CO2 emissions and real income would rise.

Also, the paper contains the detailed estimation of effects by region, fossil fuel, and decomposition of the change in CO2 emissions (the change in real output (scale), the change in the share of global output from each industry (composition), and the emissions intensity of each industry (technique)).

Thus, this article shows the presence of bias in tariffs between ‘clean’ and ‘dirty’ industries which can influence the pollution level. It provides the political explanation of such a phenomenon, mentioning the lobbying efforts of upstream industries. Apart from providing evidence of existing bias, this work contains a theoretical model that allows investigating the influence of trading policy on emissions level and social welfare. Applying this instrument to estimate contrafactual policies, J.S. Shapiro shows how social welfare and emission levels differ under different trade measures, therefore this work could have important policy implications.

Micromotives and Macrobehavior T.C. Schelling (1978)


T.C. Schelling won the Nobel Prize in 2005 for his contribution to the understanding of conflict and cooperation via game-theory analysis. His ideas on these topics are discussed in The Strategy of Conflict (1960) and Arms and Influence (1966). In 1969 and 1971 he published articles devoted to racial dynamics and segregation originating from people’s preferences for neighborhoods. In Micromotives and Macrobehavior T.C. Schelling provides generalized models describing how people making individual choices could come to the particular equilibrium distribution in place of living, school attendance, etc.

The first chapter starts with an example of people choosing seats in the auditorium skipping the first rows. Here authors discuss that such behavior could be driven by various individuals' motives. Then, the chapter provides a description of the market and non-market social interaction and introduces the concept of equilibrium analysis that will be used throughout the book.

In the second chapter, T.C. Schelling elaborates on patterns that do not depend on the individual choices, such patterns are realized due to the characteristics of the aggregate as a whole: there always will be people who will be standing in the musical chairs game.

The third chapter discusses the usage of models in social science and comprises several models that can be applied to various topics: self-fulfilling prophecy, critical mass, the commons, the market of lemons, acceleration principle.

— Self-fulfilling prophecy stands for situations when expectation induces behavior that cause the expectation to be fulfilled. The basic examples are run on banks lending to banks’ insolvency or inflation. There are several types of self-fulfilling prophecy: self-equilibrating expectation (if we believe that everyone will bring food, we might turn around and bring drinks), self-confirming expectation (if smokers believe that mentholated cigarettes are in green packages, manufacturers may use such package), self-displacing prophecy (if everyone expects same average behavior, they will displace the average from where they thought it would be), self-negating prophecy (if everyone believes that an event is overcrowded and stays home, it would not be overcrowded), etc.

— Critical mass could be exemplified by the contrast between a research seminar that fades after several meetings and a volleyball game gathering enough people to maintain its existence. T.C. Schelling defines two subclasses of critical mass phenomena: tipping and lemons. Tipping is a concept that manifests itself in migration processes, occupation, colleges, public beaches, etc. It describes how people react to new entrants, for instance, the coming of a few members of a minority can cause the departure of a formerly homogeneous population, which frees new space for more minority members. Thus, we could obtain a new allocation of households. In the lemons market described by Akerlof, the market could disappear without the introduction of institutional arrangements like guarantees.

— Commons is a paradigm referring to the situation in which people impinge on each other in pursuing their interests, but they could be better off collectively if they could restrain. Nevertheless, no one gains individually by self-restraint. One example is water-saving in summer.


In the next two chapters, the author turns to the issues of mixing and segregation both on discrete characteristics like race or language and continuous ones - age, income, etc. Here we have a model explaining how people organize themselves depending on the proportions of different groups and preferences on the neighborhood's composition.

The sixth chapter looks at the hypothetical situation of the possibility to choose genes for kids. Parents' individual choices lead to the shift in demographics for the society on a macro level. The possible government intervention aimed at the return to balance is discussed.

The concluding chapter discussed uniform multi-person prisoner’s dilemma showing how different equilibrium could be achieved under different preferences conditional on the behavior of others.

To sum up, the book covers some fundamental models that describe social equilibrium derived from individual choices. The latter depends on the actual behavior of others or the beliefs about their behavior since people comprise the social aggregate. Having different size groups and individual preferences we can obtain various equilibria, but not all of them will be efficient. In such cases, the introduction of institutional mechanisms could be socially beneficial.

Governmental initiatives are considered important for solving environmental problems, therefore one can be interested in the assessment of such measures' environmental impact and economic efficiency. D. A. Keiser and J. S. Shapiro analyzes the Clean Water Act showing its impact on water quality, allocation of funds, and Act’s impact on housing values.

Clean Water Act is a federal law in the USA regulating water pollution. It introduces wastewater standards for industries, water quality criteria, etc. The paper is aimed to solve two controversies around the Clean Water Act: its impact on pollution levels and the relationship between its cost and benefits. To do this the authors use a comprehensive set of data: atlas mapping all U.S. surface waters, panel description of the country’s wastewater treatment plants, historic extract of the Grants Information and Control System describing each Clean Water Act grants the federal government gave to cities; the Survey of Water Use in Manufacturing, a confidential plant-level data set of large industrial water users; and around 50 million water pollution readings at over 240,000 pollution monitoring sites during the years 1962–2001

From econometric analysis including difference-in-difference and triple-difference models, the following results are obtained:


1. Most variables that show the level of water pollution declined over the period 1962–2001 with a slowing rate.

2. Using triple difference regression (before vs after investments, upstream vs downstream of recipient plants), the authors show that receiving the grant decreases the probability of violating the standards for being fishable by half a percentage point. Concerning the flow from federal to municipal budgets on fighting pollution, it was found that $1 of a federal grant led to about $1 more of municipal sewerage capital spending.

3. In the analysis of house values around rivers and streams, D. A. Keiser and J. S. Shapiro estimates the effects of grants on home values are about 25% of their costs. Nevertheless, for absolute values, the analysis of effects on housing demand leads to the conclusion that the value of homes increases, but the effect is modest and not significant in most specifications.

Possible explanations are the following: people may have incomplete information about changes in water pollution and their welfare; these numbers exclude nonuse values; grants may increase sewer fees; these estimates abstract from general equilibrium effects; and they exclude the 5% of most distant recreational trips.


David A Keiser, Joseph S Shapiro, Consequences of the Clean Water Act and the Demand for Water Quality, The Quarterly Journal of Economics, Volume 134, Issue 1, February 2019, Pages 349–396, https://doi.org/10.1093/qje/qjy019

The Church was a highly powerful institution in the Middle Ages and later, and one of the colonization’s rationale was proselytizing mission. Priests that came to Latin America patronized the local population and kept records that helped to get information about the colonization process. Also, it is worth remembering that religion at those times played first fiddle in the educational process. Thus, religious missions influenced the further development of areas where they were located. F. V. Caicedo analyzed whether it is a case, and what impact the Jesuit order’s mission had on the current human capital in some regions of Argentina, Brazil, and Paraguay.

The author uses archival records, census data, and household surveys to analyze this question. The explanatory variable is the distance to the nearest mission with controls for geographic features and weather, and the dependent variables are schooling years/literacy level and income, if available, or multidimensional poverty index as a substitute. The strong and significant effect is shown for both education and income: there is a 10% reduction in illiteracy when moving 100km closer to a mission, poverty is also lower for locations close to missions. PCA combining income and education variables provide the same results.

To prove the causation, F. V. Caicedo performs two tests. First, the author uses abandoned missions to address the potential endogeneity of missionary locations. She conducts placebo-type tests and finds no consistent effects on modern education and income. Second, F. V. Caicedo studies the Franciscan Guaranı missions, which was aimed to maximize the number of souls converted to Christianity and did not stress the formation of human capital in contrast to Jesuits. The results also have no significant influence on Franciscan Guaranı missions on modern education.

Then, the author investigates the persistence of human capital. Using the earlier censuses, F. V. Caicedo shows that the effect is larger for intermediate historical periods. Also, the literacy gap is higher for women that corresponds to historical evidence of female instructions.

The last part of the article provides and tests mechanisms of persistence of differences in human capital and incomes. The author focuses on the following mechanisms:

— Occupational specialization: individuals that attended religious missions, receiving instruction and technical training, moved away from agriculture to start a proto-artisan class. The empirical tests suggest that structural transformation is a mechanism of persistence for missionary effect.

For this part of the analysis, I have the following concern: it seems like there is no control for natural resources ( soil, minerals), distance to big cities which are a sales market, and other variables which can potentially explain the specialization. In my opinion, it could be a limitation for the results.

-- Technology adoption: the author uses the adoption of GE soy seeds in Brazil and tests whether areas with higher human capital—close to Jesuit missions—adopted this new agricultural technology faster. Empirical evidence supports the explanation.

-- Alternative mechanisms: population density, investments in infrastructure, and migration. It was found that places close to Jesuit missions are less dense today. There is a significant effect of missionary distance on modern road network density and the opposite effect on railroad density, reflecting the few tracks on the Brazilian coast. Analyzing health as a human capital investment, the author shows negative and significant coefficients on distance from a mission for the Brazilian health index. Concerning migration, it does not seem that people are sorting themselves into former missionary locations during modern times.


Felipe Valencia Caicedo, 2019. "The Mission: Human Capital Transmission, Economic Persistence, and Culture in South America," The Quarterly Journal of Economics, Oxford University Press, vol. 134(1), pages 507-556.

Labor unions and shifts in inequality

David Card, Thomas Lemieux, and W. Craig Riddell studied the impact of unionization on wage inequality exploiting the survey data from the US, UK, and Canada. The choice of these countries is explained by similar organization of union systems and the large share of the non-unionized labor force which is needed for comparisons. They measure the inequality as a variance of wages within a skill group that was approximated by the observed characteristics such as age and education from the data. Also, they provide a comparison of unions’ effect between male and female workers. Here I would summarize some findings of the paper:

1. Unions tend to “flatten” wage differentials across skill groups. One explanation of this effect is related to unobserved skill differences between union and nonunion workers in different age-education groups. Union workers, as was shown in previous literature, are more skilled on average. Union wage structures do not fully compensate high-productivity individuals. Then, to reduce the burden of above-market wages, employers tend to be more selective. This induces a selectivity bias that manifests as the true union wage premium. Also, it leads to overestimation in “between-group” equalizing the effect of the union in the proposed theoretical framework.

2. Considering the effect on female workers, the authors show that the union wage gaps for women are roughly constant. Since the rate of unionization of women is rising across wages distribution, the absence of a “flattening” effect of unions on female wages implies that covariance between the nonunion wage and the union wage gain for a particular skill group is either zero or positive. Thus, the effect of unions on female inequality is limited.

3. Discussing the trends of unionization, the authors mention that for males the level of unionization has declined sharply while it is not the case for women. The reason is the shift of unionization from the private to the public sector where women, and unionized women, in particular, are concentrated.

4. The union wage gap is typically larger for women than for men. That can be explained by the fact that unionized women are more highly concentrated in the upper end of the skill distribution than unionized men. Thus, controlling for the skills reduces the union wage gap far more for women than for men.

5. Looking at the inequality trends, Card et. al. conclude that unions reduce the variance of wages for men. By contrast, the effect for women is pretty small and slightly positive, which means that unions raise inequality. Considering inequality in a broader context, the authors provide evidence that union wage compression effects help explain a reasonable fraction of the growth in male wage inequality and cross-country differences in male wage inequality. Nevertheless, the effect is not observed in Canada. They do not consider female workers since unions do not affect inequality for them much as was shown.

Card, David, Thomas Lemieux and W. Craig Riddell. “Unions and Wage Inequality.” Journal of Labor Research 25 (Fall 2004)

The origin of state: bandits in Congo

Literature in political science suggests several theories on the states’ origin. The main issue of these theories is their testability, we could rely on historical evidence but it is not enough. Nevertheless, we can test some of them in the context of states which emerge now. Raúl Sánchez de la Sierra studies the situation in Eastern Congo providing support that the state’s origin is associated with the emergence of essential function, taxation in this case. [1] Also, his work serves as an illustration of Mancur Olson’s stationary bandit model of government. [3]

De la Sierra shows that the fiscal and collective capacities are associated with the mines of gold rather than mines of coltan where we observe the distraction of the institutions. In Eastern Congo there is no strong state power, therefore its functions are performed by armed forces which control mines. The difference between coltan and gold is the ability to hide, this fact, according to the author, leads to the difference in institutions. Since bandits can easily observe the coltan which was mined, they can directly tax citizens on the way to the village, therefore they just extract income. Gold could be easily hidden, therefore bandits should create indirect taxation to derive income from the villagers. Thus, as De la Sierra suggests, bandits introduce lump-sum taxes to enter the mine and create taxes on transactions in the village, which require the investment in fiscal capacity for accounting as well as the collective capacity to take those taxes. Using the changes in prices of coltan and gold, De la Sierra provides a causal relationship between the easiness to hide the resource and the emergence of the state’s essential functions.

As a continuation of this work, Soeren J. Henn and coauthors examine the incentives to restrain from violence and arbitrary theft by an armed group in the Eastern Democratic Republic of the Congo, the Front de Liberation du Rwanda (FDLR). [2] They use this setting because of two reasons. First, the weakness of the central state enables the FDLR to perform state functions. Second, the FDLR descends from Rwanda and has weak ties with the Congolese population that help isolate the role of the time horizon from simple benevolence. Henn et. al. show that controlling most villages the FDLR used their power to collect taxes, provide protection, and run fiscal and judicial administrations. Also, they almost never attacked the villages they controlled.

To support the causality of their findings, authors conduct interviews that suggest that in response to Kimia II [4], the FDLR lost the ability to permanently tax in their villages. Also, they hid in the neighboring forest of Itombwe from which they performed violent attacks aimed at stealing wealth and food. It could be explained by the fact that they do not internalize the effect of these operations on village growth anymore, since it does not affect them via taxation. To isolate the effect of a long stealing time horizon, the authors use the timing and targeting of Kimia II. Implementing both an event study and a differences-in-differences framework, Henn et. al. compare the FDLR state villages to the rest of the sample before and after Kimia II. They conclude that compared to before Kimia II, FDLR’s violent expropriation operations increased by 350% in FDLR state villages. This finding supports the notion that a long stealing horizon before incentivized taxation over arbitrary expropriations.


[1] Raúl Sánchez de la Sierra. (2020). On the Origins of the State: Stationary Bandits and Taxation in Eastern Congo. Journal of Political Economy. Volume 128, Number 1

[2] Soeren J. Henn, Christian Mastaki Mugaruka, Miguel Ortiz, Raul Sanchez de la Sierra, David Qihang W. (2021). On the Ends of the State: Stationary Bandits and the Time Horizon in Eastern Congo. BFI WORKING PAPER

[3] Mancur Olson, (2000). Power And Prosperity: Outgrowing Communist And Capitalist Dictatorships

[4] Kimia II is a military mission of the government and UN against FDLR. For references, see this or this.

Long-term impact of institutions on culture

Institutional economics literature argues that both institutions and culture matter for economic development but this argument is built on separate research of the influence of each of these factors. Lowes et. al. study the interplay between culture and institutions, particularly, they investigate the impact of formal institutions on the development of social norms. [1] The paper suggests that centralized institutions are associated with weaker norms of rule following and a greater propensity to cheat for material gain.

The authors use the natural experiment stemming from the origin of The Kuba Kingdom. Approximately in the 15th century, there was an influx of migrants in an area near the confluence of the Kasai and Sankuru rivers. Historical evidence states that migrated groups are homogeneous in terms of culture. This fact leaves us with no pre-trend. The second fact is that the boundaries of the Kingdom were determined by an external ruler using the geographical features of the area. Also, these boundaries which coincide with the rivers’ valleys were stable during history, therefore authors exploit the emergence of the Kuba Kingdom as a natural experiment.

Historical evidence suggests that this state was relatively developed: it had a professional bureaucracy, a system of taxation, extensive public goods provision, an unwritten constitution, etc. To examine the long-term impact of those institutional advances on culture, Lowes et. al. compare individuals whose ancestors lived within the Kuba Kingdom to individuals whose ancestors lived just outside the Kingdom. They test for differences in the propensity to follow rules in cases when there is a monetary incentive to not do so. The measure of rule following comes from two sets of behavioral experiments. The first one is the resource allocation game (RAG): a player rolls the dice with different colors on sides and divides the money between her and the other player according to her color identification which she chooses in her head. So, there is no control for cheating, but it could be identified via the deviation from the theoretical distribution. The second experiment is a version of the standard ultimatum game, in which participants physically allocate money in a private setting that allows them to steal money before offering a division.

Lowes et. al. consider three samples. The first sample is the largest and includes all individuals whose ancestors lived inside and just outside the Kuba Kingdom (499 individuals in total). The second sample comprises only the population that was culturally homogeneous before the creation of the Kuba Kingdom, it includes the central Kuba and the Lele - descendants with the same origin (and culture) but leaving outside Kuba. The third sample focuses specifically on the core people of the Kuba Kingdom and compares them to the Lele.

The authors show that Kuba ancestry is associated with more rule breaking and more theft. This is true for both experiments and all three samples of interest. This finding could be interpreted as follows: culture can be shaped by state institutions, and this setting is in line with the predictions of the Tabellini model. [2] Thus, their results show that Kuba state is crowding out internal norms of rule following. This finding contradicts the hypothesis of Elias, Weber, and Foucault, who argues that culture and formal institutions are complements. [3] [4] [5]

Addressing the positive correlation which was found in previous studies, the authors claim that restricted samples produce negative estimates of the Kuba Kingdom on rule following with the larger magnitude compared to the full sample. Then, the restricted sample estimates, which exploit the cultural regression discontinuity, might be better identified than the full-sample estimates, this suggests a positive reverse effect of culture on institutions: groups with stronger norms of rule following are more likely to establish more centralized and formal state structures. This fact leads to the upward bias of the estimates of the effect of state centralization on rule-following norms.

[1] Lowes S, Nunn N, Robinson JA, Weigel J. (2017). The Evolution of Culture and Institutions: Evidence from the Kuba Kingdom. Econometrica. 85 (4) : 1065-1091.
[2] Tabellini, G. (2008). The Scope of Cooperation: Values and Incentives. The Quarterly Journal of Economics, 123(3), 905–950. http://www.jstor.org/stable/25098921
[3] Elias, N. (1994): The Civilizing Process. Oxford: Blackwell.
[4] Foucault, M. (1995): Discipline and Punish: The Birth of the Prison. New York: Vintage Books.
[5] Weber, E. (1976): Peasants Into Frenchmen: The Modernization of Rural France, 1870–1914. Stanford: Stanford University Press.

Online media and ideological segregation

In his book, Sunstein argues that the Internet facilitates peoples’ convergence to sources that correspond to their initial views that could be harmful to democracy. [2] The urgency of this issue is rising due to the Internet's fast expansion and its probable substitution of traditional media and face-to-face interaction. Nevertheless, M. Gentzkow and J. Shapiro provide evidence against this point. They conclude that online media is less ideologically segregated than national newspapers or face-to-face interaction. [1]

The authors collect data on the consumption of newspapers, magazines, broadcast television, and cable. To measure in-person interaction they use the political views of individuals’ acquaintances and political discussants as reported in the 2006 General Social Survey and the 1992 Cross-National Election Study. For each outlet, they measure the share of conservatives (those who report their political outlook as “conservative” among those who report being either “conservative” or “liberal”). Then, they calculate the individual’s conservative exposure to be the average share conservative on the outlets she visits. As a measure of segregation, the authors use the “isolation index”, which is a standard metric in the literature on racial segregation. It equals the difference between the average conservative exposure of conservatives and the average conservative exposure of liberals. For instance, if conservatives only visit foxnews.com and liberals only visit nytimes.com, the isolation index will be equal to 100 p.p., while if they both get all their news from one source, the isolation index will be equal to 0.

According to this approach, the estimated isolation index for the Internet is higher than that of broadcast television news (1.8), cable television news (3.3), magazines (4.7), and local newspapers (4.8) and lower than that of national newspapers (10.4). Online segregation is higher than that of a social network where individuals are matched randomly within counties (5.9) and lower than that of a network where individuals matched randomly within ZIP codes (9.4). Also, it is significantly lower than the segregation of actual networks formed through voluntary associations (14.5), work (16.8), neighborhoods (18.7), or family (24.3). The Internet is also far less segregated than networks of trusted friends (30.3) and political discussants (39.4).

M. Gentzkow and J. Shapiro discuss two mechanisms for this phenomenon:

— Online news consumption is concentrated in a small number of relatively centrist sites. More ideologically extreme sites account for a very small share of online consumption.

— The majority of individuals consume content from multiple sources. Especially those who visit more extreme blogs overall who spend time online and use more centrist sources or even sites with the opposite ideological slant as well.

One of the caveats of this study is that it does not explain how people translate the content they encounter into beliefs. People with different ideologies see similar content, but they might interpret it differently. Nevertheless, this paper provides a strong argument against previous results suggesting the higher level of segregation caused by internet media. [3]


[1] Gentzkow, M., and J. Shapiro (2011): “Ideological Segregation Online and Offline,” Quarterly Journal of Economics, 126 (4), 1799– 1839.

[2] Sunstein, Cass R. Republic.com (Princeton, NJ: Princeton University Press, 2001).

[3] Theoretical models of media markets predicting in which increasing the number of outlets may lead consumers to become more segregated ideologically:
- Mullainathan, Sendhil, and Andrei Shleifer. “The Market for News,” American Economics Review, 95 (2005), 1031–1053.
- Sobbrio, Francesco. “A Citizen-editors Model of News Media,” SSRN Working Paper, 2009.
- Stone, Daniel F. “Ideological Media Bias.” SSRN Working Paper, 2010.

Also, there are plenty of comments of public officials, commentators, and thinkers suggesting the increasing polarization. For instance, see this or this article from NYT.

Polarization vs antipolarization

Social media is often accused of dividing society apart. Sustain (2017) in his book argues that social media facilitates the creation of echo chambers that expose people to those who are similar to them. This effect makes their initial slant more strong. However, Gentzkow, M., and J. Shapiro (2011) show that the Internet is not as polarized as face-to-face contacts though they do not provide analysis for social media. C. Bail provides experimental evidence on the polarizing effect of exposure. However, these findings are not supported by R. Levy (2021). Indeed, the answer to this question is not obvious since one could come up with alternative mechanisms which justified the decrease in polarization, for instance, lower cost of searching for alternative explanations. I will briefly describe the findings of both papers and provide some arguments to explain this two-sided effect of social media.

Bail et. al. (2018) experimented on Twitter by suggesting users follow a bot that retweets posts opposed to their views. R. They select accounts of elected officials, opinion leaders, media organizations, and nonprofit groups, and regularly randomly select an account to make a retweet. Comparing the surveys before, within, and after the experiment authors make the following conclusions:

— Republicans who were exposed to a liberal Twitter bot became substantially more conservative post-treatment.
— Democrats exhibited slight increases in liberal attitudes after following a conservative Twitter bot. However, these results are not statistically significant.

In his book, C. Bail talks about his research based on in-depth interviews. Here he explained the channel of polarization. First, the polarization after exposure to the opposite opinion if initially, people have more extreme views. Second, he shows that there are discrepancies between online and offline behavior, particularly, people could be restricted to share their opinion online because offline or, on the contrary, be freer on social media compensating for the lack of social approval and admiration in real life. Third, he provides suggestions to improve communications online by creating a new discussion platform where people of different slants were incentivized to discuss various topics.

R. Levy conducted an online experiment suggesting users subscribe to opposing news outlets. He managed to gather data on subscriptions to outlets, exposure to news on Facebook, visits to online news sites, and sharing of posts, as well as changes in political opinions and attitudes from the surveys. This study has four main findings:

1. News sites visited through social media, and specifically, Facebook, are associated with more segregated and extreme news, compared to other news sites. Nevertheless, exposure to opposing news outlets helps to diversify the news consumption of an individual.
2. Exposure to counter-attitudinal news decreases negative attitudes toward the opposing political party, i.e. affective polarization. This is in line with Bail’s finding of exposing people to those with opposite views on the discussion platform.
3. Exposure does not affect people's beliefs and ideological slant.
4. Facebook’s algorithm is less likely to supply individuals with posts from counter-attitudinal outlets, conditional on individuals subscribing to them. Thus, social media algorithms may limit exposure to counter-attitudinal news.

Thus, a comparison of the focuses of these works could help to shed a light on this divergent effect. First, the extreme belief which is emphasized in Bail's works could be the main driver of polarization. Since the share of such people is not high enough, this is not reflected in Levy's findings. The second reason lies in the dichotomy of active interaction and consumption. In my opinion, the first should affect beliefs and affective polarization compared to pure consumption of news since in the interaction with others more tools of persuasion are used. The third potential explanation, which is close to the first one, is the coherence of beliefs. As C. Bail shows in his book and is established in political science literature, people could have contradictory and incomplete beliefs which also could change the way how they are responsive to the exposure of opposite opinions.

This question lies if the area of my research interest, therefore this small post could be treated as a motivation for my thesis. I hope to provide a more rigorous argument on these counter effects later on.

References:

Levy, Ro'ee. 2021. "Social Media, News Consumption, and Polarization: Evidence from a Field Experiment." American Economic Review, 111 (3): 831-70.

C. A. Bail et. al. 2018. Exposure to opposing views on social media can increase political polarization. Proceedings of the National Academy of Sciences, 115(37): 9216-9221

C. A. Bail 2021. Breaking the Social Media Prism: How to Make Our Platforms Less Polarizing. Princeton University Press.

Dynamic polarization

Polarization is a widespread phenomenon in the political sphere, and it affects different groups differently. Particularly, we notice the polarization between elites (parties) and masses (voters). There are great papers that empirically investigate polarization among people (to be reviewed), but there was no one that tries to catch polarization at both levels and provide a theoretical concept to explain polarization dynamics. S.Callander and J. C. Carbajal fill this gap by providing a dynamic voting model with voters who update their beliefs.

They build upon a classic model of electoral competition adding behaviorally justified change to the nature of voter preference. Namely, the voter updates their preferences in a way to justify their choice which is called a causal loop, i.e. they vote for a party that is closer to them and then they become even closer to this party in terms of preferences. Also, unlike other models, they introduce the tolerance of voters which guarantees that voters vote not just for the closest party but the closest party within a particular range. This helps to derive equilibrium with the so-called “missing middle” of the electorate or “disappearing center” Abramowitz (2010) – moderate voters who are left behind because of parties’ polarization.

Since parties face a trade-off between the probability of winning and the policy outcome, the authors find a new type of equilibrium where parties do not compete with each other but with abstention. This strategy is called “no-voter-left-behind”. It means that parties polarize incrementally such that in each step, no voters are lost to abstention. This implies that the speed of polarization is tightly linked to the degree voters update their preferences after voting: the smaller is updating, the more incremental the updating, and the slower and more iterative party polarization. This dynamic reflects a shift in the nature of electoral competition over time: from competition for the swing voter to competition for turnout.

The dynamic nature of this model allows us to conclude that parties would converge faster to their ideal points while voters will update their beliefs until they reach the position of their favorite parties. Note that abstainers do not update their ideals points, so the gap in the middle opens up. This result is supported by empirical literature that finds evidence for elite polarization and very moderate polarization within the masses.


[1] Steven Callander & Juan Carlos Carbajal, 2022. "Cause and Effect in Political Polarization: A Dynamic Analysis," Journal of Political Economy, Vol. 130, N. 4

[2] Abramowitz, A. I. (2010). The Disappearing Center: Engaged Citizens, Polarization, and American Democracy. Yale University Press.