Aktuelle Publikationen

Measuring sentiment in social media text has become an important practice in studying emotions at the macroscopic level. However, this approach can suffer from methodological issues like sampling biases and measurement errors. To date, it has not been validated if social media sentiment can actually measure the temporal dynamics of mood and emotions aggregated at the level of communities. We ran a large-scale survey at an online newspaper to gather daily mood self-reports from its users, and compare these with aggregated results of sentiment analysis of user discussions. We find strong correlations between text analysis results and levels of self-reported mood, as well as between inter-day changes of both measurements. We replicate these results using sentiment data from Twitter. We show that a combination of supervised text analysis methods based on novel deep learning architectures and unsupervised dictionary-based methods have high agreement with the time series of aggregated mood measured with self-reports. Our findings indicate that macro level dynamics of mood expressed on an online platform can be tracked with social media text, especially in situations of high mood variability.

Since the Fukushima nuclear accident in 2011 (3/11), Japan has witnessed a nationwide upsurge of antinuclear demonstrations as well as protests against diverse concerns. Why did this upsurge of protests occur beyond the antinuclear concerns? The exact mechanism that caused this general upsurge of protests has not been explored in detail. Given the limited number of first-time participants, this phenomenon can be fully explained only through an analysis of network-building processes among social movement organizations (SMOs). Based on the first nationwide survey of SMOs conducted in Japan by our team in February 2018, covering 308 groups, we analyzed the constellation of the SMOs’ networks after 3/11, their logic of coalition building, and their network effects on mobilization. We observed that the new characteristics of the constellation of the SMOs’ networks are twofold. The first feature is that the networks of various SMOs were bridged by antinuclear groups. Antinuclear organizations served as the hub of SMOs’ networks, enabling the mobilization to cross different concerns. The second feature is the connections between the citizen groups and labor unions. Labor unions increase the participation during events by mobilizing their partner unions. Both new connections contributed to the general upsurge in large-scale demonstrations for various concerns after 3/11. Our paper contributes to the general discussion on the relation between coalition of SMOs and mobilization by focusing on the dyadic level of networks, and analyzing its effect for mobilization through egocentric network analysis.

Background: Research has repeatedly shown that exposure to suicide-related news media content is associated with suicide rates, with some content characteristics likely having harmful and others potentially protective effects. Although good evidence exists for a few selected characteristics, systematic and large-scale investigations are lacking. Moreover, the growing importance of social media, particularly among young adults, calls for studies on the effects of the content posted on these platforms.
Objective: This study applies natural language processing and machine learning methods to classify large quantities of social media data according to characteristics identified as potentially harmful or beneficial in media effects research on suicide and prevention. Methods: We manually labeled 3202 English tweets using a novel annotation scheme that classifies suicide-related tweets into 12 categories. Based on these categories, we trained a benchmark of machine learning models for a multiclass and a binary classification task. As models, we included a majority classifier, an approach based on word frequency (term frequency-inverse document frequency with a linear support vector machine) and 2 state-of-the-art deep learning models (Bidirectional Encoder Representations from Transformers [BERT] and XLNet). The first task classified posts into 6 main content categories, which are particularly relevant for suicide prevention based on previous evidence. These included personal stories of either suicidal ideation and attempts or coping and recovery, calls for action intending to spread either problem awareness or prevention-related information, reporting of suicide cases, and other tweets irrelevant to these 5 categories. The second classification task was binary and separated posts in the 11 categories referring to actual suicide from posts in the off-topic category, which use suicide-related terms in another meaning or context.
Results: In both tasks, the performance of the 2 deep learning models was very similar and better than that of the majority or the word frequency classifier. BERT and XLNet reached accuracy scores above 73% on average across the 6 main categories in the test set and F1-scores between 0.69 and 0.85 for all but the suicidal ideation and attempts category (F1=0.55). In the binary classification task, they correctly labeled around 88% of the tweets as about suicide versus off-topic, with BERT achieving F1-scores of 0.93 and 0.74, respectively. These classification performances were similar to human performance in most cases and were comparable with state-of-the-art models on similar tasks.
Conclusions: The achieved performance scores highlight machine learning as a useful tool for media effects research on suicide. The clear advantage of BERT and XLNet suggests that there is crucial information about meaning in the context of words beyond mere word frequencies in tweets about suicide. By making data labeling more efficient, this work has enabled large-scale investigations on harmful and protective associations of social media content with suicide rates and help-seeking behavior.

Despite recent attention to the economic and political consequences of automation and technological change for workers, we lack data about concerns and policy preferences about this structural change. We present hypotheses about the relationships among automation risk, subjective concerns about technology, and policy preferences. We distinguish between preferences for compensatory policies versus “protectionist” policies to prevent such technological change. Using original survey data from Spain that captures multiple measures of automation risk, we find that most workers believe that the impact of new technologies in the workplace is positive, but there is a concerned minority. Technological concern varies with objective vulnerability, as workers at higher risk of technological displacement are more likely to negatively view technology. Both correlational and experimental analyses indicate little evidence that workers at risk or technologically concerned are more likely to demand compensation. Instead, workers concerned about technological displacement prefer policies to slow down technological change.

When separation is a problem in binary dependent variable models, many researchers use Firth's penalized maximum likelihood in order to obtain finite estimates (Firth, 1993; Zorn, 2005; Rainey, 2016). In this paper, I show that this approach can lead to inferences in the opposite direction of the separation when the number of observations are sufficiently large and both the dependent and independent variables are rare events. As large datasets with rare events are frequently used in political science, such as dyadic data measuring interstate relations, a lack of awareness of this problem may lead to inferential issues. Simulations and an empirical illustration show that the use of independent “weakly-informative” prior distributions centered at zero, for example, the Cauchy prior suggested by Gelman et al. (2008), can avoid this issue. More generally, the results caution researchers to be aware of how the choice of prior interacts with the structure of their data, when estimating models in the presence of separation.

Sierra Leone is a successful case of postconflict democratization, but citizens' trust in parliament is low. Using qualitative and quantitative data, this article examines the relationship between MPs and their constituencies. Given the fierceness of the competition at the level of party primaries, aspiring MPs make promises they cannot keep. Confronted with unrealistic expectations, they reduce their presence in the constituency to a minimum. As a result, dissatisfaction increases, and disappointing MPs are voted out of office. The high turnover rate weakens the institutional memory and law-making capacities of parliament, thus creating even more disillusionment. In the long run, the continuation of this cycle may break the links between voters and political representatives.

Though negativity in political debates influences citizens’ attitudes towards legislative institutions, research on how Members of Parliaments (MPs) use negative language remains scant. This study shows how the gender of speakers and the context of debates influence the level of negativity in parliamentary speeches. We argue that female MPs use less negative language than male colleagues due to gender differences in socialisation and stereotypical expectations. Applying sentiment analysis with word embeddings to 20 years of plenary speeches in the Austrian parliament, we find that speeches by women MPs are less negative on average compared to those of their male colleagues. A more balanced gender distribution within a party group decreases differences in tone by lowering the negativity of male speakers. A growing share of women in parliament can thus change the tone of debates, which might enhance the legitimacy of political institutions and the quality of democracy.

With this work, we present a publicly available data set of the history of all the references (more than 55 million) ever used in the English Wikipedia until June 2019. We have applied a new method for identifying and monitoring references in Wikipedia, so that for each reference we can provide data about associated actions: creation, modifications, deletions, and reinsertions. The high accuracy of this method and the resulting data set was confirmed via a comprehensive crowdworker labeling campaign. We use the data set to study the temporal evolution of Wikipedia references as well as users’ editing behavior. We find evidence of a mostly productive and continuous effort to improve the quality of references: There is a persistent increase of reference and document identifiers (DOI, PubMedID, PMC, ISBN, ISSN, ArXiv ID) and most of the reference curation work is done by registered humans (not bots or anonymous editors). We conclude that the evolution of Wikipedia references, including the dynamics of the community processes that tend to them, should be leveraged in the design of relevance indexes for altmetrics, and our data set can be pivotal for such an effort.

dc.title:

The livelihoods of rural populations in Africa are closely tied to small-scale farming. In recent years, private investors as well as governments have shown a growing interest in large-scale acquisition of arable land across the continent. While researchers have started to analyze the local economic and environmental impacts of such investments, their socio-political as well as psychological consequences remain poorly understood. This paper investigates how changes in land ownership patterns caused by large-scale land acquisitions affect the level of interpersonal trust among rural communities. We maintain that the transition from community and individual-smallholder land ownership into large-scale investor property has a negative impact on local levels of trust. Furthermore, we assume that the deterioration of trust caused by large-scale land investments is stronger among women than men. To test our claims, we connect circa 71,000 respondents from Afrobarometer surveys to georeferenced information on the location of land deals from 33 African countries. Relying on a difference-in-differences type of empirical strategy as well as an instrumental variable approach, we show that large-scale land investments indeed disrupt local social fabrics by reducing interpersonal trust. Our results suggest that trust in relatives is particularly affected by large-scale land acquisitions. In addition, we find that land deals reduce personalized trust among women but not necessarily among men.

Electoral systems provide distinctive accountability mechanisms in democratic polities and thereby affect government responsiveness to citizens. In this article, we concentrate on the effects of proportional vis‐à‐vis majoritarian electoral rules. We expect members of parliament to be more responsive under majoritarian rule, because these MPs have a direct mandate from their local constituency, are less dependent on their party, and can be held directly accountable by voters. We exploit Germany's mixed‐member system and test MP’s responsiveness using behavioral data generated within a two‐round field experiment. The experiment observes concrete interactions between voters and representatives. In the experiment, real voters sent emails about a policy issue to their MPs. We show that MPs who were elected via the majoritarian tier are almost twice as likely to respond to a voter request than MPs elected via PR. Our results deliver novel evidence that electoral institutions cause distinct behavioral responses from elected officials.

What kind of welfare state do voters of populist radical right parties (PRRPs) want and how do their preferences differ from voters of mainstream left- and right-wing parties? In this paper, we draw on an original, representative survey of public opinion on education and related social policies in eight Western European countries to measure (1) support for social transfers, (2) support for workfare and (3) support for social investment. Challenging the view that PRRPs turned into pro-welfare parties, our results indicate that their voters want a particularistic-authoritarian welfare state, displaying moderate support only for ‘deserving’ benefit recipients (e.g. the elderly), while revealing strong support for a workfare approach and little support for social investment. These findings have important implications for contemporary debates about the future of capitalism and the welfare state.

This article adopts a novel regulatory perspective on the conditions that facilitate and obstruct economic equality between migrants and natives. Regulation scholars have long emphasized that regulatory interventions need to be geared toward the needs of regulatory targets. We contribute to this research by examining the fit between regulatory instruments and targets' human capital skills. We develop a theoretical framework that captures how economic integration regulations (EIRs) influence economic equality by supporting or restricting migrants in the labor market and as entrepreneurs. We argue that EIRs foster economic equality when they are responsive to the professional needs of specific types of regulatory targets (in terms of language and education skills). We apply the framework in the context of OECD countries. A fuzzy‐set qualitative comparative analysis reveals how the specific configurations of EIRs in 26 OECD countries coincide with either high or low economic equality between migrants and natives. Our approach contributes to the conceptual understanding of a pressing regulatory problem: the successful economic integration of migrants.

Stricter regulation of food waste reduction is widely presumed to increase food prices, which could render its implementation politically unfeasible. Here we empirically tested whether specific policy framing, design and feedback could help ensure public support despite potential food price increases. We used survey experiments with 3,329 citizens from a high-income country, Switzerland. A combined framing and conjoint experiment shows that messages emphasizing national or international social norms in favour of reducing food waste (policy framing) can increase public support for more ambitious reduction targets. Also, most citizens support food waste regulation even if this leads to substantial increases in food prices, but only if such policies set stringent reduction targets and are transparently monitored (policy design). Finally, a vignette experiment reveals that voluntary industry initiatives do not crowd out individuals’ support for stricter governmental regulation, but potentially crowd in support if industry initiatives are unambitious (policy feedback).

Researchers now have greater access to granular georeferenced (i.e., spatial) data on social and political phenomena than ever before. Such data have seen wide use, as they offer the potential for researchers to analyze local phenomena, test mechanisms, and better understand micro-level behavior. With these political event data, it has become increasingly common for researchers to select the smallest spatial scale permitted by the data. We argue that this practice requires greater scrutiny, as smaller spatial or temporal scales do not necessarily improve the quality of inferences. While highly disaggregated data reduce some threats to inference (e.g., aggregation bias), they increase the risk of others (e.g., outcome misclassification). Therefore, we argue that researchers should adopt a more principled approach when selecting the spatial scale for their analysis. To help inform this choice, we characterize the aggregation problem for spatial data, discuss the consequences of too much (or too little) aggregation, and provide some guidance for applied researchers. We demonstrate these issues using both simulated experiments and an analysis of spatial patterns of violence in Afghanistan.