Using the text of 200 million pages of 13,000 US local newspapers and state-of-the-art machine learning methods, we construct a novel 170-year-long time series measure of economic sentiment at the country and state levels. It expands existing measures in the time series availability (by more than a century) and is the first to provide cross-sectional (regional) variation. Our corpus includes approximately 1 billion newspaper articles, a major increment over the Wall Street Journal corpus, a popular source of text data in economics and finance, that contains about 1 million articles.

To measure text-based economic sentiment, we customize a new state-of-the-art machine learning technique. We create a fully automated topic-specific dictionary by leveraging Word2vec, a neural-network-based algorithm, that allows us to capture the meaning of words and phrases from the context in which they are used. Furthermore, instead of using a binary positive/negative connotation, we produce a continuous measure of sentiment for each word and phrase in the dictionary. As a result, our method automatically overcomes many common challenges faced by simple word-count techniques (e.g., detecting negation or measuring word/phrase intensity). Empirically, our measure is highly correlated with the outcome of observable survey expectations (e.g., Michigan Consumer Sentiment survey), yet significantly extends their available time span for analysis and provides a source of cross-sectional variation.

Our measure predicts future economic growth even after controlling for current macroeconomic fundamentals: One standard deviation increase in sentiment corresponds to 2% additional annual growth in GDP per capita during 1850-2017. Over the recent sample with observed quarterly data (1947Q1 - 2019Q4), a one standard deviation increase in sentiment leads to 0.29% of additional GDP growth one quarter ahead (corresponding to 1.1% annualized growth). This predictability remains even after controlling the slope of the yield curve, past GDP dynamics, and the consensus forecast, implying that our measure captures important information not spanned by leading predictive macroeconomic indicators.  

We then examine the sub-components of GDP that our measure is able to predict. We show that our results are mainly consistent with the labour channel rather than the capital channel of sentiment propagation in the economy, as our measure predicts employment, consumption, and services but neither investment nor industrial production. Similarly, we also show that economic sentiment operates through the real economy and does not affect inflation.

Next, we evaluate the extent to which sentiment affects monetary policy decisions. To this end, we quantify the importance of sentiment in explaining the changes in the fed funds rate relative to what the forward-looking Taylor rule proposed by Romer and Romer (2004) would imply. We find that sentiment has a large influence on the key policy rate: a one-standard-deviation decrease in sentiment over the past two quarters leads to a 25 basis point (5 basis point) decrease in the policy rate during recessions (expansions). Furthermore, we find that sentiment has significant predictive power for the Fed Funds Rate during recessions even after controlling for its predictive power on ex ante (Tealbook projections) and ex post (realized) GDP growth.

Our data also allows us to measure local sentiment, for example, at the state level. This variation is important, as it reveals significant heterogeneity in sentiment across states, with the common component driving only approximately 35% of the state-level sentiment. Local sentiment predicts state-level GDP growth even after controlling for national sentiment and both national and state-level fundamentals. Furthermore, using the dispersion in sentiment across states as a measure of heterogeneity, we find that higher dispersion is a significant predictor of low economic growth at the national level.

Overall, our results indicate the importance of sentiment for understanding business cycles (both globally and locally), and provide a set of robust empirical facts that could indicate potential channels of its impact. 

Mayukh Mukhopadhyay is a PhD student in Financial Economics at the London Business School. He holds a BA in Economics and an MPhil in Economic Research from the University of Cambridge. Mayukh’s research interests include the use of big data and machine learning methods for the prediction of asset returns and macroeconomic forecasting. Prior to joining London Business School, Mayukh worked for the International Finance Division at the Bank of England. 

Research Interests:  Asset Pricing, Macroeconomics, Big Data, Machine Learning

Reputation is an important aspect that has a major impact on the performance of banks and the overall banking system, including correspondent banking. However, it can be difficult to measure and evaluate. Reputation and trust are particularly crucial elements of (not only) the entire financial world. One important player that contributes to the reputation of correspondent banking is the media, especially newspapers and economic periodicals. This paper analyses the media image of the different actors of correspondent banking and the relationship between it and other important influences in the US between 1990 and 2022.

The study analyzes hundreds of articles from the most influential periodicals over the past 32 years using Natural Language Processing (NLP) techniques. The focus of the analysis is on content analysis, specifically sentiment analysis, and a comparison with the performance of correspondent banking. Archival sources, which include digital versions of newspapers, magazines, financial statements, annual reports, and earnings calls, are processed through OCR, and analyzed using Python programming and libraries such as NLTK, TextBlob, VADER, or Beautiful Soup. Major publications such as Reuters, Bloomberg, The Washington Post, The New York Times, The Herald (Everett), and The Capital are among the included sources.

The study also shows the strengths and limitations of NLP text data analysis. The principal advantages of the approach are the collection and analysis of enormous amounts of data. On the other hand, there are also risks of quantitative methods within historical sciences such as imprecision, incompleteness, lack of data, or confusion of correlation and causality. The findings, which are presented both in written and graphical form using Tableau, will provide valuable insights into the relationship between media image and cross-border financing.

Keywords

NLP, Sentiment Analysis, Correspondent Banking, Economic Analysis, SWIFT, Cross-border payments

Petr Štěrba is a PhD student specializing in economic history at the Prague University of Economics and Business. He is also engaged in research in the field of text mining and alternative data analysis as part of his research at the Central Bank of the Czech Republic. Petr holds a master’s degree in Economic and Social History from Charles University, and has also completed a study visit at Northumbria University in the UK.

In the past decade, Eurosceptic national populism has posed one of the most significant challenges to European democracy. To better understand this ideology’s emergence, this project synthesises intellectual histories of national populism and neoliberalism with recent advances in the digital humanities. As David Armitage and Jo Guldi have argued, the increasing accessibility of software enables social scientists and historians to move beyond information overload towards improved data-driven analyses and more impactful visualisations.

Social Network Analysis (SNA) is a particularly promising tool for historians seeking to engage further with the digital humanities. Rooted in mathematical graph theory, SNA extracts data from sources and visualises them as graphs of individual points (nodes) connected by lines (edges). By employing these methods, network software can determine the most important nodes (or their “centrality”) and locate distinct clusters within the network (also known as “community detection”). Using SNA, this project systematically maps the individuals, organisations, ideologies, and intellectual traditions that coalesced to form a transnational network that this paper calls the “Eurosceptic Thought Collective”. This research examines the diachronic ideological stability of individuals and organisations within the network and asks how radical Eurosceptic concepts emerged and disseminated across it.

It is proposed that, through the 1990s and 2000s, this network catalysed support for Brexit by hybridising reactionary neoliberal Euroscepticism, immanent right-wing populism, and British nationalist traditions. British Euroscepticism was not an isolated, nativist ideological formation. Rather, it was a local expression of a transnational network of individuals and institutions that sought the secession of member-states from the EU to further encase global markets within a deregulated, low-tax network of spaces that Quinn Slobodian has called “the zone”.

These findings challenge the view that the 2016 Brexit referendum and the election of Donald Trump heralded the end of the neoliberal world order and the rejection of globalisation. Instead, it is increasingly evident that the alliance between national populism and neoliberalism has become a dominant feature of the early 21st century. A social network analysis of the historical roots of Euroscepticism in Britain helps better understand this syncretic ideology that continues to challenge European democracy.

Over the past five years, Sebastian Lowe has been studying the impact of Euroscepticism on British politics. Born in Lausanne and educated in Switzerland, Britain, and America, he graduated from Queen Mary University of London in 2015 with a Ba in History and Politics. After working in accounting and nuclear energy in the United States, he commenced an MSc in Empires, Colonialism and Globalisation at the London School of Economics. Subsequently, he pursued a doctorate at the University of Oxford, first attaining an MPhil from Wolfson College before starting his current DPhil studies at University College. His research aims to synthesise the digital humanities with modern political and intellectual history, specifically studying neoliberalism, nationalism, and populism.

Inspired by Kimberly Kay Hoang’s recent publication Spiderweb Capitalism, this research utilises the non-ferrous metals industry to illustrate structures of capital networks within wartime (1931-1945) Japanese Empire’s dual-use items industries. Non-ferrous metals are essential for producing dual-use items, which are vital to maintain a flexible wartime economy. During wartime, Japan operated numerous mines to manufacture non-ferrous metals, and the production reached its zenith in the twilight of the Empire. Drawing on Japanese, American, British and Chinese military, intelligence, and governmental archives, this research surveys the business operators of the Japanese non-ferrous metals production-related mines at its homeland, colonial Taiwan, Chōsen, Manchuria, coastal China, and its wartime colonies in Southeast Asia, and reconstructs the capital networks of these operators by utilising yearbooks of share companies published by the securities companies. Similar to the contemporary Spiderweb Capitalism, a majority of these mines were controlled by the zaibatsu’s subordinate companies (agents/subordinate spiders), whilst the major stockholders (dominant spiders) like zaibatsu families, aristocrats, imperial household-related bureaucrats and national policy companies (国策 会社) hidden behind the ‘star network’, which were formed through cross ownerships and intermarriages (spider silk). This research uses Gephi, a network analysis and visualisation tool to analyse and present the networks involving more than a thousand individuals and organisations; and to calculate and measure who the central figures are, the intermediaries between the networks, and the influence of each figure. It shows how the wartime Japanese political-military-economic clique used the capital networks to collectively fund and thereby sustain the exploitation of colonial resources and the production of the non-ferrous metals and other dual-use items industries. Through Gephi’s visualisation and computation, this research further explores the personnel and organisations that acted as the capital resources of Japan’s wartime dual-use items industries, opening a window to explore the patterns of how the power elites of the Japanese empire financed their war machine.

Keywords: Non-ferrous metals industry, Dual-use items industries, Japanese Empire’s power elites, Total war, Networks analysis and visualisation

Tsz Ho Wong is currently a PhD student in East Asian Studies at the University of Edinburgh. He earned a BA from the University of Hong Kong and a MSc from the London School of Economics and Political Science. He is particularly interested in the history of modern East Asia with focuses on economic history, intellectual history and histories of science and technology from the mid-nineteenth century to the early Cold War. He has great passion in applying digital tools in his research. His works have been published by Routledge (forthcoming in 2023), the Center for Malaysian Chinese Studies and the Webster Review of International History. Currently, he is working on the patterns of the capital and power elites’ networks of the wartime Japanese Empire, and the life of a sinicised Mongolian female writer Liang Yen (梁琰, a.k.a Margaret Yang Briggs, born as Yang Chiao-Chu 楊巧珠). For more information, please visit his personal website (brianthwong.com)