Overview: This project extracts customer contact preferences to receive USAA-initiated insurance correspondence for a selection of customers across every US state. The project employs the MbrContact R Package to reduce customer preferences for USAA-initiated contact across different purposes to a single, dominant preference. This information helps Contact Center partners adjust resources prior to a policy shift. Additional analysis compares (and explains) the distribution of preferences across each state.

Overview: This project compares USAA-initiated customer contact preferences across states and subgroups. To do so, the project employs the MbrContact R Package to reduce customer preferences for USAA-initiated contact across different channels to a single, dominant preference. For each group (e.g., state), the business could adapt to the different distribution of preferences in preparation for a policy shift.

Overview: MbrContact is a R package that extracts a single customer preference for USAA-initiated contact. The data output provides the preferred mode of USAA-initiated contact (i.e., preference) using a rule-based algorithm that extracts preferences across different purposes. Since preference extraction is a frequently requested analytical exercise from business partners, the R Package makes it possible to increase efficiency in decision-making and to establish consistency over time in those analytical decisions. Knowing contact preferences helps the business to distribute resources for updating customers following a policy shift.

Overview: This analysis demonstrates that customers employ different communication channels for specific tasks. We should not expect consistent preferences for different directions of communication—USAA to Member and Member to USAA—and we should not expect consistent preferences across the different reasons for contact. Preferences should be allowed to shift based on the purpose for contact. Motivating these observational outcomes is a fundamental interest for members to control communication with USAA. With a deeper analysis, the business would know the factors that explain the count of phone call and email contacts, making it possible to dynamically adjust resources to efficiently balance market demands.

Overview: Missing values are a common feature in industry data, within and outside of USAA. The consequence of missing values could either introduce bias that results in misguided conclusions or be randomly absent to have no impact on confident conclusions about business behavior. When data is missing completely at random, incomplete data will be inconsequential to our ability to adequately assist our business partners. Because there is no concrete test to resolve this analytical hurdle, this analysis proposes a strategy to help researchers diagnose missing data and assess the opportunity to still provide confident conclusions.

Overview: This descriptive analysis explains to the business partners an analytical strategy to uncover customer preferences for contact from USAA. It suggests a strategy to infer preferred mode of contact for different company reasons for initiating contact to customers.

Overview: Since USAA delivers multiple letters following missed payments, this analysis explores the amount of revenue paid from delinquent customers for the 90 days that follow the final Debt After Cancellation letter. This letter is distributed 140 days after the initial missed payment. The analysis integrated multiple data sources using observational data and managed missing data. Although revenue exceeds costs, attributing the effect to the letter is impossible due to multiple confounding factors. I recommended and designed an experiment to isolate the impact.

Overview: This essay describes the empirical relationship between legislative assembly size and the population of the state. Based on this formula, the US House of Representatives has been undersized since adopting the Apportionment Act of 1911. This means that individual legislators have a greater workload than prior legislators due to representing a larger average number of people in their districts than before.

Overview: This essay describes political behavior in nonpartisan elections. Voters tend to rely on candidates’ gender and ethnicity as informational cues to make vote decisions, or they simply abstained altogether. Although party names are absent from the ballot and voters must overcome additional hurdles to make vote decisions, the differences in outcomes are minimal between partisan and nonpartisan elections once officials are in office.

Overview: This essay describes features of online voters registration and the various implementation across US states. Although electronic data processing minimizes the frequency of erroneous registration, any online system is susceptible to security vulnerabilities from external attempts to infiltrate data.

Abstract: Democratic accountability relies on citizens to anticipate future governing behavior. We explore the strategic incentives for parties to shape voter expectations by generating vague or concrete campaign statements. Using an English-language dictionary we scale electoral statements from all industrialized English-speaking nations to develop a measure of concreteness. Concrete statements can create electoral risks from unfulfilled expectations. Therefore, political parties have incentives to use concrete statements to clarify reputation uncertainty associated with unclear informational cues. Political context shapes these incentives. Incumbent parties tend to dictate concrete statements to balance attributed responsibility for government outcomes and signal that they are competent managers. Strong government performance, however, reduces the incentive for incumbents to be concrete as favorable outcomes reveal competent management. Opposition parties are unconstrained from these demands. The research reveals how political parties actively manage and balance the information that voters use in order to adjust the informative value of party reputation.

Abstract: In this paper we present TopFish, a multilevel computational method that integrates topic detection and political scaling and shows its applicability for a temporal aspect analysis of political campaigns (preprimary elections, primary elections, and general elections). It enables researchers to perform a range of multidimensional empirical analyses, ultimately allowing them to better understand how candidates position themselves during elections, with respect to a specific topic. The approach has been employed and tested on speeches from the 2008, 2012, and the (ongoing) 2016 US presidential campaigns.

Abstract: General political topics, such as social security and foreign affairs, recur frequently in electoral manifestos across countries. The Comparative Manifesto Project aims to collect and manually code manifestos of political parties from all around the world, detecting political topics at sentence level. Since manual coding is expensive and time consuming and allows for annotation inconsistencies, in this work we present an automated approach to topical coding of political manifestos. We first independently train three sentence-level classifiers – one for detecting the topic and two for detecting topic shifts – and then combine their predictions in a global optimization setting using a Markov Logic network. Experimental results show that the proposed global model achieves high classification performance and significantly outperforms the local sentence-level topic classifier.

Abstract: This article explores the question of why coalition partners negotiate and publish coalition agreements before entering into a cabinet and why the content of these agreements varies so widely. Some scholars suggest that coalition partners draft agreements for electoral purposes, while others suggest that coalition agreements can be used to commit to policy negotiations. Although both sides of the debate have uncovered supportive evidence, the literature remains in disagreement. This article provides new organization of previous work on agreements and develops two alternative theoretical arguments about the crafting of coalition agreements. It is argued here that coalition partners consider both electoral and policy motivations during the drafting of agreements and that the dominance of one of these motivations is conditional on the degree of issue saliency and division between partners. Empirical support is found for the theoretical argument that coalition partners include low saliency issues in the coalition agreement on policy dimensions on which they are less divided, and that coalition partners include high saliency issues in the coalition agreement on policy dimensions on which they are more divided.

Abstract: We examine the proposition that incentives for legislative organization can be explained by the nature of electoral competition. We argue that legislators in environments where parties are competitive for majority status are most likely to have delegated power to their leadership to constrain individualistic behavior within their party, which will in turn increase the spatial predictability of individual voting patterns. Using roll call votes and district-level electoral data from the U.S. state legislatures, we show empirically that increased statewide interparty competition corresponds to much more predictable voting behavior overall, while legislators from competitive districts have less predictable behavior.

Overview: This paper describes patterns of party competition and party development during the 2013 Ecuadorian legislative and presidential elections. The overview also provides a background to major political and economic themes during the prior Correa administration and describes the institutional structure of the presidential and legislative electoral systems. See our blog post on The Monkey Cage (click here) for a more focused evaluation of party nationalization overtime in Ecuador.

Overview: Invited book review of From the Mines to the Streets: A Bolivian Activist’s Life.

Abstract: Electoral, cabinet, and floor-voting coalitions reflect differences in the credible commitments of coalition partners to the executive, largely due to the tension between the electoral imperative to develop partisan reputations and the policymaking necessity to act collectively. We argue that variation in partners’ commitments via coalition design should result in predictable differences in policymaking efficiency. Using success rate of executive-initiated bills and an original panel database of Latin American government coalitions, we show how coalition type and control over the legislature impact passage rates. Legislative success is highest under single-party majority government, followed by electoral coalition government, cabinet coalition government, and single-party minority government with floor-voting coalitions. The empirical precision of our models improves when incorporating democratic quality in model heterogeneity, and our findings are robust across over 29,000 specification permutations. Results suggest that some partisan solutions are more effective than others at resolving institutional problems of presidentialism.

Abstract: Uncovering how political parties manage negotiated outcomes in multiparty democracy affects the source of informational cues that structure viable political competition. Consolidated democracy requires distinct informational cues between government and opposition parties. This manuscript isolates the effect of multiparty democracy on government-opposition distinctions. In it, I argue that shared government responsibility, though favorable to democratic transitions, is detrimental to political competition. Informational distinctions between government and opposition blur as the number of parties in government increase and the nature of the governing coalition is less coherent. To do so, I develop a measure using automated content analysis of legislative debates. It captures the coherence of legislative speeches among government party members and determines if speeches are consistently different from legislators of the opposing parties. I apply this measure to Bolivia since the 1982 transition to democracy until the second single-party majority government in 2010, where varying compositions of the governments in a presidential democracy—with fixed terms—reveals the negative effects of multiparty government arrangements on viable political competition.

Abstract: Political parties communicate policy intentions to voters before elections. These statements can be specifically engineered to signal competency and to balance negative government performance that is attributed to the incumbent party (Eichorst & Lin forthcoming). Approximating vague and concrete language in political communication, however, can be susceptible to error. Two dominant methods exist for coding political statements: human-coders and computer-assisted. Each strategy has a tradeoff. Human-coding incorporates human intuition for processing messages. Thus, it appears valid. Overuse of human coders can lead to unintended consequences since exhausted coders can generate errors. Though there are many forms of computer-assisted classification, we focus on dictionary methods. Dictionary methods, alternatively, are reliable for massive text processing but can have issues with validity. Since we rely on a dictionary method in our own research, this raises an important question: Do human-coders interpret political statements consistently with dictionary methods? We explore the consistency between these two methods using the LIWC dictionary and two survey samples with respondents recruited using Amazon MTurk. Our results suggest that human coders perform consistently with dictionary methods