Machine learning for financial transaction classification across companies using character-level word embeddings of text fields

An important initial step in accounting is mapping financial transfers to the corresponding accounts. We devised machine-learning-based systems that automate this process. They use word embeddings with character-level features to process transaction texts. When considering 473 companies independently, our approach achieved an average top-1 accuracy of 80.50%, outperforming baselines that exclude the transaction texts or rely on a lexical bag-of-words text representation. We extended the approach to generalizes across companies and even across different corporate sectors. After standardization of the account structures and careful feature engineering, a single classifier trained on 44 companies from 28 sectors achieved a test accuracy of more than 80%. When trained on 43 companies and tested on the remaining one, the system achieved an average performance of 64.62%. This rate increased to nearly 70% when considering only the largest sector.     

Testing the Empirical Relevance of the ‘Saving for a Rainy Day’ Hypothesis in US Metro Areas

The joint implication of the consumption Euler equation and cointegration between income and consumption is that savings predict future income declines, the ‘saving for a rainy day’ hypothesis. The empirical relevance of this hypothesis plays a key role in discussions of fiscal policy multipliers, and it holds under the null that the permanent income hypothesis is true. We find little support for this hypothesis using time series data for the 100 largest US Metropolitan Statistical Areas for the period 1980q1–2015q4. Our approach is to test for cointegration and weak exogeneity between income and consumption, and by exploring the direction of Granger causality between the two time series. We find that income more often predicts consumption and saving than the converse. We also give evidence that house price changes played a role in US income and consumption dynamics, before, during and after the Great Recession.     

Predicting the equity market with option-implied variables

We comprehensively analyze the predictive power of several option-implied variables for monthly S&P 500 excess returns and realized variance. The correlation risk premium (CRP) and the variance risk premium (VRP) emerge as strong predictors of both excess returns and realized variance. This is true both in- and out-of-sample. Our results also reveal that statistical evidence of predictability does not necessarily lead to economic gains. However, a timing strategy based on the CRP leads to utility gains of more than 5.03% per annum. Forecast combinations provide stable forecasts for both excess returns and realized variance, and add economic value.     

Revisiting the expatriate failure concept: A qualitative study of Scandinavian expatriates in Hong Kong

This article advances the understanding of expatriate failure, which remains a contested social phenomenon in international work life as well as scholarly research. The study challenges the definition of expatriate failure and its inherent biases, i.e., the epistemological primacy of the firm level and the failure/success binary. We argue that this qualitative study of 51 Scandinavian expatriates in Hong Kong can contribute to advancing theory on the expatriate failure concept by asking individual expatriates what constitutes failure to them. By applying social constructionist and social anthropological ideas to the expatriate failure concept debate, we develop the internationality thesis which demonstrates a discrepancy between the expatriates’ perceptions of successful international assignments and the actual nature of their lived lives; many expatriates desire to enrich their lives through experiencing an international/intercultural and adventurous lifestyle, but, in fact, living lives with limited intercultural exposure and interaction. We conclude by proposing a reconceptualisation of expatriate failure in terms of offering both a new definition and approach to researching expatriate failure in which time/duration, context, and geographical location need to be taken into account. We believe the new approach can overcome some of the empirical unsoundness of mainstream definitions.     

Machine Learning and Expert Judgement: Analyzing Emerging Topics in Accounting and Finance Research in the Asia–Pacific

In this paper, we focus on the question to what extent machine learning (ML) tools can be used to support systematic literature reviews. We apply a ML approach for topic detection to analyze emerging topics in the literature—our context is accounting and finance research in the Asia–Pacific region. To evaluate the robustness of the approach, we compare findings from the automated ML approach with the results from a manual analysis of the literature. The automated approach uses a keyword algorithm detection mechanism whereby the manual analysis uses common techniques for qualitative data analysis, that is, triangulation between researchers (expert judgement). From our paper, we conclude that both methods have strengths and weaknesses. The automated analysis works well for large corpora of text and provides a very standardized and non-biased way of analyzing the literature. However, the human researcher is potentially better equipped to evaluate current issues and future trends in the literature. Overall, the best results might be achieved when a variety of tools are used together.     

Predicting Household Poverty: A Methodological Note with a Kenyan Example

Household surveys provide data that is used for identifying and measuring the poverty status of households and individuals. However, carrying out such surveys is expensive, especially in poor developing countries. Thus it is important to make maximum use of the available survey data in developing countries, especially in sub‐Saharan Africa, where such data are expensive to collect and analyse. This paper develops a simple method for using poverty indices derived from survey data for a given year, to predict poverty rates for subsequent periods without having to conduct a new household survey. We illustrate the workings of the method with data from Kenyan household surveys for 1994 and 1997.     

The Local Whittle Estimator of Long-Memory Stochastic Volatility

We propose a new semiparametric estimator of the degree of persistencein volatility for long memory stochastic volatility (LMSV) models.The estimator uses the periodogram of the log squared returnsin a local Whittle criterion which explicitly accounts for thenoise term in the LMSV model. Finite-sample and asymptotic standarderrors for the estimator are provided. An extensive simulationstudy reveals that the local Whittle estimator is much lessbiased and that the finite-sample standard errors yield moreaccurate confidence intervals than the widely-used GPH estimator.The estimator is also found to be robust against possible leverageeffects. In an empirical analysis of the daily Deutsche Mark/USDollar exchange rate, the new estimator indicates stronger persistencein volatility than the GPH estimator, provided that a largenumber of frequencies is used.     

Time Inhomogeneous Multiple Volatility Modeling

Price variations at speculative markets exhibit positive autocorrelationand cross correlation. Due to large parameter spaces necessaryfor joint modeling of variances and covariances, multivariateparametric volatility models become easily intractable in practice.We propose an adaptive procedure that identifies periods ofsecond-order homogeneity for each moment in time. To overcomethe high dimensionality of the problem we transform the multivariateseries into a set of univariate processes. We discuss thoroughlythe implementation of the adaptive technique. Theoretical andMonte Carlo results are given. We provide two applications ofthe new method. For a bivariate exchange rate series we comparethe multivariate GARCH approach with our method and find thelatter to be more in line with the underlying assumption ofindependently distributed innovations. Analyzing a 23-dimensionalvector of asset returns we underscore the case for adaptivemodeling in high-dimensional systems.