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.     

Firm age and performance

Amid increasing interest in firm age and its effects on firm performance, this special issue offers an exhaustive review of the literature and a novel collection of evidence on the effects of firm age on performance, including a special focus of interest on innovation performance, financial performance, exports, survival and growth. This editorial positions the theme in the extant literature, and provides key definitions and challenges ahead in the field of evolutionary economics. It introduces the collection of articles composing the special issue. The papers offer a diversity of country contexts, as well as analytical approaches and methods. They include an exhaustive review of the literature on age and firms’ performance, and present original empirical studies focusing on the effects of age on firms’ economic outcomes on the one hand, and on innovation outcomes on the other hand. While most of the papers use econometric analysis, the level of analysis ranges from firm to individual.     

Bivariate and multivariate scaled association models. An application to homogamy of social origin and education in Hungary between 1930 and 1979

In this article we compare bivariate and multivariate models for homogamy of social origin and education to test whether bivariate models of homogamy lead to biased results. We use data on Hungarian couples married between 1930 and 1979 and loglinear models of scaled association. The results indicate some differences between bivariate and multivariate analyses. At each point of time bivariate models overestimate homogamy, both with respect to education and social origin. However, results on trends in time do not differ much between the two analyses. The exception is the period 1940–1959, in which bivariate analysis showed decreasing educational homogamy, and multivariate analysis showed an increasing trend. The latter finding can be explained by declining homogamy of social origin, as well as the weaker reproduction and cross-effects in this period.     

The risk-return trade-off in human capital investment

In this paper, we analyze investments in human capital in a way which is standard for financial assets, but not (yet) for human capital assets. We study mean-variance plots of human capital assets. We compare the properties of human capital returns using a performance measure and by using tests for mean-variance spanning. Fields differ strongly not only in common rates of return, but also in return per unit of risk. We identify a range of educations that are efficient in terms of investment goods, and a range of educations that may be chosen for consumption purposes.     

Killing off cohorts: Forecasting mortality of non-extinct cohorts with the penalized composite link model

Mortality forecasting has crucial implications for insurance and pension policies. A large amount of literature has proposed models to forecast mortality using cross-sectional (period) data instead of longitudinal (cohort) data. As a consequence, decisions are generally based on period life tables and summary measures such as period life expectancy, which reflect hypothetical mortality rather than the mortality actually experienced by a cohort. This study introduces a novel method to forecast cohort mortality and the cohort life expectancy of non-extinct cohorts. The intent is to complete the mortality profile of cohorts born up to 1960. The proposed method is based on the penalized composite link model for ungrouping data. The performance of the method is investigated using cohort mortality data retrieved from the Human Mortality Database for England & Wales, Sweden, and Switzerland for male and female populations.     

Impacts of an EC carbon/energy tax and deregulating thermal power supply on CO2, SO2 and NOx emissions

Emission of CO₂, SO₂ and NO_x are all closely linked to the burning of fossil fuels. Here we report on simulations done by linking a Sectoral European Energy Model (SEEM), covering energy demand in nine Western European countries, with the emission-transport-deposition model RAINS developed by IIASA. The study analyses emissions of CO₂, SO₂ and NO_x, deposition of sulphur and nitrogen and the extent of areas where the critical load for sulphur is exceeded in year 2000 under four different energy scenarios. Two different sets of future behavioural patterns for the thermal electric power production sector are considered. In one regime, called the plan-efficient regime, the sector is assumed to follow official plans with regard to investment in new capacity. In the other regime, called the cost-efficient regime, the thermal power sector is assumed to behave in a cost minimizing manner. The effects of the proposed EC carbon/energy tax are studied under both regimes, giving rise to altogether four scenarios.In both regimes the effect of the EC tax is to reduce emissions by between 6 and 10 per cent in year 2000 relative to the scenarios without the tax. A change of regime, from the regulated, plan-efficient regime to the market-based, cost-efficient regime, will, by itself, reduce emissions of CO₂ and NO_x by approximately 3 per cent, while SO₂ emissions are reduced by 13 per cent. The EC tax will reduce sulphur deposition by more than 5 per cent in the nine model countries under the plan-efficient regime. A change of regime further reduces the total deposition by 9 per cent. The area where depositions exceed the critical load is reduced by approximately 6 per cent in year 2000 by the tax in both regimes. Changing from the plan-efficient to the cost-efficient regime has a similar impact.Although the emission reductions due to the EC tax may seem modest, they are shown to have a sizeable effect on the technological abatement costs of reaching targets like those prescribed in the Sofia protocol on the stabilisation of NO_x emissions, and the Helsinki protocol on SO₂ emission reductions. This is part of what can be considered to be secondary benefits of the EC carbon/energy tax.     

Sequencing and the Adding-up Property in Contingent Valuation of Endangered Species: Are Contingent Non-Use Values Economic Values?

This paper presents tests of the theoretical validity of the contingent valuation (CV) method. The validity of CV is especially a concern when involving environmental goods with a predominant non-use value. One test of theoretical validity is the adding-up property that implies that a specific good should be equally valued irrespectively of it is being valued directly or built-up sequentially. In this CV study four independent sub-samples stated willingness to pay for the same composite good, or package, using different sequences. One sub-sample valued the composite good directly, while two sub-samples faced built-up sequences valuing first subsets of this composite good. A fourth sub-sample valued the composite good from a dividing-out approach, facing first the valuation of a larger multi-package. Theoretically expected sequencing effects were observed; the subset goods obtained higher values earlier in a sequence, and the dividing-out approach decreased the stated value for the composite good. Most importantly, these CV data did pass the tests of the adding-up property.     

Environmental quality indicators: Background,principles and examples from Norway

The need for a simple and general overview of the development in the state of the environment has led to work on environmental indicators in several countries and international organizations. Except for the basic common requirement of providing an overview, the works vary a great deal with regard to final aim, target groups, classification and, hence, choice of indicators. This paper gives an overview of the work carried out in the Central Bureau of Statistics of Norway within this field, with emphasis on the basic principles behind the choice of indicators. A hierarchical system of indicator sets is proposed, and examples of indicators are presented. Also, a brief overview of work carried out elsewhere is included.     

The World Is Not Enough! Small Open Economies and Regional Dependence

In this paper, we explicitly introduce regional factors into a global dynamic factor model. We combine new open economy factor models (emphasizing global shocks) with the recent findings of regional importance in the business cycle synchronization literature. The analysis is applied to a large panel of domestic data for four small open economies. We find that global and regional shocks explain roughly 30 and 20 percent, respectively, of the business cycle variation in all countries. While global shocks have most impact on trade variables, regional shocks explain a relatively large share of the variation in cost variables.