Quantitative Textual Analysis (Student Applications)

I am hugely proud of the many students who have sought my assistance in the process of writing their dissertations and theses. These are just a few examples of the many excellent student research papers.


Hannah Bailey

China’s Two-Faced Rhetoric: Mobilising Citizens and Calming Foreigners? LSE Undergraduate Dissertation (2018)

This paper measures the extent to which China depicts in its rhetoric a peaceful or an aggressive rise to power, focussing on the period from 2010 to 2017. In particular, I examine the extent to which China, under the Presidency of Xi Jinping, employs rhetoric towards its domestic citizenry in a systematically different way to the rhetoric it employs towards foreigners. The Mandarin and the English language versions of the People’s Daily Online are used as data reflecting rhetoric to domestic and foreign audiences, respectively. The methodology is twofold: a Latent Dirichlet Allocation topic model is used to capture the content of the two versions of the publication, and an Ordinary Least Squares regression is then employed on key topic proportions as they shift over time. The findings strongly support a bifurcation in Chinese rhetoric regarding its rise to power, where the content of the People’s Daily Online towards foreigners is largely peaceful, and that toward domestic citizens is more aggressive.

*Contact me for this paper, as it is transitioning


James E. Sanders

Emphasis Framing in UK Parliamentary Debates: Uncovering Framing Behaviour in Large Textual Datasets LSE Undergraduate Dissertation (2018)

Political debates are an innately public activity, and as such, an opportunity for politicians to influence the electorate's preferences. One potential method is framing, the process of emphasising certain evaluative dimensions of an object over others. Within the context of the 2010-2015 Liberal Democrat-Conservative UK government and the broader framing literature, this paper attempts to address three objectives. First, by extrapolating the results of previous behavioural experiments that study the effectiveness of framing, I attempt to justify why politicians would frame at all during parliamentary debates. Secondly, I formulate a number of hypotheses surrounding expected framing behaviour in parliamentary debates, based on previous literature. Finally, I formulate a novel methodology to identify the factors which influence framing behaviour in large unstructured textual datasets. I do so through computer-assisted quantitative textual analysis, specifically latent semantic analysis and regression analysis. I find that there is a small negative association between the ideological distance of two consecutive speakers' parties and the similarity of the frames contained within their speeches. I also find that senior-ranking cabinet and shadow cabinet members engage in more framing behaviour than their lower ranked colleagues. This effect is amplified when cabinet and shadow cabinet members are replying to one another, dwarfing the effect of ideological distance on framing. I postulate that the majority of framing behaviour in parliamentary debates stems not from its effectiveness, but rather from the whip and party cohesion.

*Contact me for this paper, as it is transitioning


Giulio Lisi

Essays on Central Bank Transparency, PhD Dissertation (LSE) (forthcoming)

Paper 1: Legislative oversight and bureaucratic transparency: The case of the FOMC

Abstract. Does legislative oversight affect the transparency, and hence the accountability, of public sector agencies? I study how political scrutiny affects the disclosures of the Federal Open Market Committee (FOMC). To link oversight to transparency levels, I develop a version of the model of costly communication put forward by Dewatripont and Tirole (2005). An agency takes a decision which outcome is uncertain and must decide whether to exert effort to communicate it to a supervisor, whom in turn must exert effort to monitor the agency’s action. I show that, under reasonable assumptions, the agency prefers to communicate its decisions to signal its ability rather than remaining silent; moreover, communication efforts increase with political attention (‘visibility’). The empirical study tests model predictions using automated content analysis to extract the amount of information released in FOMC minutes together with the exogenous timing of committee hearings in US Congress and daily data of media coverage of macroeconomic policy to identify effects. Results indicate a positive and robust correlation between predictors of legislative attention and the amount of information the FOMC discloses to the public, even after controlling for competing mechanisms, in a way that is consistent with the research hypotheses. I conclude with implications for the study of communication and accountability in monetary committees and public sector agencies more broadly.

Paper 2: Disclosure requirements and communication in organisations: Evidence from the FOMC

Abstract. What is the effect of disclosure requirements imposed by legislatures on public sector agencies? Do they work, as intended, to reveal policy-relevant information to legislators? Or do they hinder internal discussions by bureaucrats wary of political repercussions? I address these questions by focusing on the communication behaviour of the Federal Open Market Committee (FOMC) following the unexpected decision to release verbatim transcripts of policy meetings in November 1993. I draw on specific features of this reform together with computational linguistics methods (particularly cosine similarity) to derive consistent measures of the amount of information released in the meeting minutes and alleviate methodological concerns. I analyse the responsiveness of the committee’s announcements to measure of public salience of its decisions before and after the reform, showing a positive and robust effect of the increased visibility on FOMC transparency. The release of the transcripts appears to have made committee’s disclosures more responsive to newspaper coverage of its decisions, particularly around the time in which FOMC hearings in Congress are scheduled. I do not find significant changes in the content of the transcripts before and after the reform. The findings suggest that ex-post release of deliberation records can be an effective way to increase public accountability of government agencies.

Paper 3: Constraints on the executive, exchange rates, and central bank transparency

Abstract. How do domestic political institutions influence central banks’ communication policy? I study how political factors affect the transparency of monetary authorities in both industrialised and emerging economies. My main argument is that, because transparency serves also as a cred- ibility device for monetary policymakers, communication should be more open in contexts when the central bank has greater control over policy, as well as when global economic factors make information transmission more important as a tool to anchor expectations. I start by evaluating how characteristics of the state’s political regime affect transparency, suggesting that central banks should communicate more in contexts where constraints on the executive power are higher. I then argue that this effect should be more significant when exchange rates are more flexible, and when a country’s financial account is more open. I test these predictions using panel data on transparency for 98 central banks in the period 1998-2010, finding strong support for the research hypotheses. In addition, I show that effects of transparency on inflation are similarly conditioned by institutional factors. The paper generates new insights into the relationship between central bank transparency, the institutional contexts in which monetary authorities operate and the overall effectiveness of communication as a policy instrument.

*Contact me for this thesis, as it is transitioning


James E. Sanders

Similarities and Differences in the Argumentative Characteristics of the Official Brexit Campaigns, LSE Undergraduate Political Review (January 2018): 1-21 (http://blogs.lse.ac.uk/lseupr/lse-undergraduate-political-review-%E2%80%A2-volume-1-%E2%80%A2-january-2018/)

This paper adds to the growing empirical literature surrounding the UK's vote to leave the European Union. Specifically, a series of quantitative and qualitative textual analysis tools are implemented on a corpus consisting of the websites of Vote Leave (VL) and Britain Stronger in Europe (BSE). By breaking down argumentative text into two components, this paper attempts to characterise how the two official campaigns differ in the information they choose to convey (or “focus"), and the style by which this information is conveyed. To analyse variation in focus, a structural topic model and thematic analysis of elementary context units are conducted with the inclusion of document-level metadata. This is then compared to survey data and their potential effectiveness is considered. To study the style of information transmission, an analysis of sentiment is used to calculate sentence-level polarity scores. An unambiguous thematic divide is uncovered with BSE employing a “focussed" approach by singling out topics related to the economy, whereas VL chose a “scattershot" approach by spreading their resources across a broader range of themes. The thematic analysis uncovers little reciprocity in most major areas — a notable exception being public services, which acted as a battleground. BSE's focussed approach allowed it to target the most influential topic for the electorate, but despite this, VL's approach led to a greater targeted proportion. A sentiment analysis yields two results: (A) the variability in sentence-level polarity scores was consistent across campaigns, and (2) BSE's website had a significantly greater mean score.


Joel Pearce

A Politics of the People: Comparing the Use of Populist Discourse in the 2016 US Presidential Election, LSE Undergraduate Political Review (January 2018): 22-57 (http://blogs.lse.ac.uk/lseupr/lse-undergraduate-political-review-%E2%80%A2-volume-1-%E2%80%A2-january-2018/)

The 2016 US presidential election saw Donald Trump and Bernie Sanders both being described as “populists”, despite running for different parties and coming from different political traditions. This paper empirically assesses the validity of this claim by conducting computer-assisted thematic analysis of their speeches during the presidential primaries. It explores the puzzle of populism being associated with diverse political positions by mapping out the candidates’ discourse, finding that both used populist themes but in strikingly different ways. Whilst Trump presented a divide between the American people and the perceived threats of Islam and immigration, Sanders contrasted the people with economic elites. They had a different approach to the campaign, with Trump framing it as a battle between him and his opponents and Sanders as an opportunity for people to come together against the powerful. Most interestingly, their discourse on trade showed little overlap: Trump presented it in both nationalist and populist terms whereas Sanders associated it with a broader theme of lost opportunities for young Americans. Existing theories account for different aspects of this contrast but none provide a comprehensive explanation of varieties of populism on their own.


Piotr Kowalczyk

Greener economies, better lives? A political economy of decarbonisation policies in the EU: An assessment of policy coordination in the EU through the lens of Discursive Institutionalism applied to the EU's decarbonisation discourses, from the launch of the Lisbon Strategy to the start of the Europe 2020 Strategy, London School of Economics MSc Thesis (2016)

My study focuses on policy coordination in the European Union (EU) and assesses how these process have changed over a 10-year period. Policy coordination has been considered a key subject of political economy and political science research as it is essential to ensure policy coherence. Coordinating policy has proven especially problematic for the EU due to its multi-level nature. As a result, various efforts have been made to increase the effectiveness of the EU's intra and inter-institutional policy coordination since 2001. Using a discourse analysis aided by ALCESTE, a textual analysis software, and predicated on Discursive Institutionalism as it is applied to decarbonisation policies, my investigation demonstrates that the EU's efforts have not optimised the coordination of different policy frames in its overarching discourse. I further show that only the European Commission has improved the coordination of its discourse, reflecting efforts within this institution to increase intra-institutional coordination since 2001. My results substantiate policy coordination debates applied to the EU and support a rethinking of current decarbonisation policies.

PDF/Adobe Acrobat


Jack Winterton

Inheritance in PMQs: Assessing the Development of the Prime Minister’s Rhetoric during the Questions to the Prime Minister Debate, LSE Undergraduate Dissertation (2015)

This paper empirically assesses the development of the Prime Minister’s rhetoric during the Questions to the Prime Minister debate (PMQs). It offers an exploratory analysis of all 4621 occasions that the Prime Minister (PM) rose to speak during the PMQs debates in the 2010-2015 Parliament. Using computer-assisted textual analysis software (Alceste), this paper presents a quantative breakdown of the key themes in the Prime Minister’s responses. In short, Alceste attempts to divide textual data into classes according to the co-occurrence of lexical forms (Valles William, 2014). Stepping into a more detailed analysis, this paper then focusses on the Prime Minister’s rhetoric that explicitly make reference to the previous Labour administration. I aim to build up an understanding of how this particular rhetorical tool changes over the course of a parliamentary term. This analysis finds that (1) the Prime Minister spent 35% of his responses at PMQs explicitly talking about the failures of the previous administration (2) while the rhetorical framework changes, the Prime Minister’s responses across all years consistently reference the previous Labour administration (3) the Prime Minister gave his most partisan responses in 2010 and 2011.

PDF/Adobe Acrobat


Derek Valles

The Contours of Political Discourse: Deliberating the Patient Protection Affordable Care Act, PhD Thesis (London School of Economics, 2014)

This thesis takes an interdisciplinary approach to the study of political deliberation by fusing together the normative dimensions of political theory with the empirics of positive political science. It proceeds by offering three interrelated contributions to what it argues are three interrelated literatures.

First, it evaluates the current state of deliberative democratic theory. It recognizes that, in the midst of their ‘empirical turn,’ deliberative democrats are interested in methodological approaches that can evaluate the extent and quality of deliberation. In doing so, it is argued that they have turned a deaf ear towards deliberative content. In response, a comprehensive theoretical framework is offered for empirically approaching deliberative content. At its core, it claims that we can access the essential deliberative elements—political concepts and ideologies—by examining the discursive contents of deliberative discourse.

Second, the thesis defines a novel and empirically grounded methodological approach to studying deliberative discourse. Recognizing the need for principled, reliable, and valid methods, it provides a framework for using quantitative text analysis to create objective mappings of the thematic patterns inherent in political text. This takes the form of an excavation by quantitatively revealing underlying patterns in the co-occurrences of terms within textual data. It offers a framework for engaging in exploratory content analysis using Alceste, and provides a comprehensive overview of its methodology and philosophy.

Third, the thesis carries out a quantitative text analysis of the floor debates held in the U.S. Senate during the consideration of the Patient Protection & Affordable Care Act of 2010. The empirical analysis bridges the divided between normative political theory and positive political science by demonstrating new technologies that statistically measure the thematic content, discursive flow, legislator impact, and role of coalitions within a deliberative discourse, with findings that reveal where there are cleavages and convergence in the political reasoning of legislators.