Global economic trends in the context of investor relations and strategic management. A longitudinal analysis of trends utilized in annual reports of german blue chip corporations

Resumen

Besides being confronted with fierce competition, today’s corporations face macroeconomic driving forces that change the configuration of existing and future markets that are also called “global economic trends” (GET) or “megatrends”. Currently, research does not provide a coherent view or theory for the assessment of these trends, which often leads to misinterpretation in business and academics. Literature such as publications provided by governmental institutions, multinational corporations, or consulting companies provide numerous examples where economic trends like globalization or the over-aging of western societies are depicted as global megatrends in the context of strategic management and economic research. The rather arbitrary use of the terms in literature leads to the idea that trend terms are vehicles to transport subjective assumptions about future developments that stem from foresight and innovation activities. Authors like Groddeck and Schwarz (2013) or Rust (2008) criticize the vague definition of these types of trends and question the information value and validity of “global economic trends” and “megatrends”. A growing body of literature points out that web search data is an ideal foundation for econometrical analysis and forecasting, because behavioral data of online user activity allow researchers to make inferences about the current state of economic decision-making of users, which is also called nowcasting (cf. e.g. Askitas and Zimmermann, 2009; Choi and Varian, 2012; Vosen and Schmidt, 2011; Dimche and Davcev, 2014). So far, there has been little to no work on trends that implements the tool Google Trends, which is the breading ground for the empirical research of this thesis. To get an insight view into the utilization of trends, a pilot study revealed that (1) corporation use of the terms “global economic trend” and "megatrend" in annual reports from 2008 to 2012, and (2) web search data from Google Trends correlates with the geographical location of multinational enterprises and economic wealthy regions measured in GDP. In the empirical part of the thesis, a long-term study from 2004 to 2014 explores how German stock market listed corporations (DAX) adopt trends in their business strategy, and how these trends are represented in web searches. Based upon a mixed-methods research approach, the thesis examines in detail what trends are utilized in annual reports and investigates these trends based on qualitative methods from the field of foresight. Then the results are analyzed with the use of web search data quantitatively. The qualitative analysis of annual reports revealed that: • In total, 5,920 passages that contained the term “trends” were identified in the population (n=330) of annual reports from 2004 to 2014. Included in the empirical research were 2,012 trend passages, whereof 392 trends were categorized as direct trend passages (TP), and whereof 1,620 were classified as indirect TPs. • The use of terms like “global economic trends” and “megatrends” grows from 2004 (117) to 2014 (273). However, only a few corporations used this term directly, but describe the effects indirectly with the use of other terminology. In 2005, Siemens AG firstly introduced the term “megatrends”. • The analysis revealed that directly mentioned TPs are more likely to be depicted as an opportunity, rather than a risk in annual reports with an odds ratio of 6.63. The 2,012 trend passages found in annual reports were categorized with an existing system that provides the attributes social, technological, economic, environment, political, and value (STEEPV). In addition, an individual categorization system (ICS) was developed as a tailor-fitted solution for the analysis of trends used in annual reports. The following outcomes were produced when both systems were applied to the data: • The STEEPV categorization system is capable of categorizing trend passages from an ex-post perspective. The final distribution of categories has a strong qualitative appeal and the application of a STEEPV category is ambiguous. The category “Economic” (74%) dominates in the overall population. • The implementation of the newly developed system reduced the overall use of the category “economic” was reduced to 45%. For quantitative analysis, (1) an indicator called Regional Index (RI) was created to analyze geographical information of web search data and (2) an indicator called confidence ranking index (CRI) was developed that measures the reported confidence of a corporation towards each trend mentioned in the annual report that is used for further examination. First, the regional index was used for correlation analysis between geographical information of web search data and GDP data, and for creating a visualization module developed with the statistical software R that provides geographical maps of web trend searches. It could be shown that: • Google Trends data is not provided globally. Scarce data from countries like China and Russia shows that the use of Google is restricted in these countries. • The regional index correlates well GDP with .591 (p < .001) on a global level, which indicates a linear relationship. A regression analysis reveals that the global RI index is able to explain 35% of the total variance. • The regional index has an even higher correlation for Germany with .841 (p < .001) on a local level. In the regression model, the local RI is able to explain 70% of the total variance. This undermines the results of the pilot study. Assuming that financial KPIs might have an influence on the CRI index, two regression models were developed that incorporate operational income and shareholders’ equity as explanatory variables. One model interprets the overall population as cross-sectional data and uses regression analysis, and the other model uses a generalized estimated equations (GEE) approach. The results indicate that: • The cross-sectional model has an R2 of .103 and adjusted R2 of .097. Approximately only 10 % of the total variation is explainable by the model, which is considered as interim results in the exploratory research. • The GEE model only contains operating income as an independent variable. With an R2 of 0.174 and an adjusted R2 of .143, the model showed little improvement in comparison to the cross-sectional model. The rather weak influence indicates that financial KPIs have a rather low influence to the CRI index, which motivated further steps of inquiry. At this phase, web search data from Google Trends was implemented and tested for its explanatory capacity. An automated correlation analysis based on the qualitative assessment of annual reports was developed to identify those Google Trend time series with the best model fit. It could be demonstrated that, • Nine hundred forty-one trend terms were found in the annual reports that were the foundation for extracting Google Trends data from 2004 to 2014 with the regional setting global and local for Germany. Globally 315 trends were returned, of which 87 trend series had a high significance (p < .01) to the CRI index. One hundred twenty- two (39%) series correlated significantly with CRI (p <0.05), and 106 (34%) showed no correlation. Thirty-six trends were returned on the local level, whereof six were highly significant (17%), 19 were significant (53%), and 11 (31%) had no significance. • Each dataset was used as a foundation for regression analysis with the confidence ranking index (CRI) as the dependent variable. The individual qualities of the ordinary least square (OLS) models measured by the coefficient of determination (R2) range from .36 to .85 on a global level and from .37 to .75 on a local level. • To optimize the previously developed OLS model with CRI as the dependent variable, the annual mean of the obtained global and local web search data was added. Google Trend data could optimize the regression measured in R2 to .151 for global data based on search terms like “Innovation trends” or “Corporate responsibility, and to .149 for local data based on search terms like “Social media” or “RFID”. • On the contrary to OLS, an optimization of the developed GEE model required the creation of an index per annual report that sums up the local and global Google web search results individually per for each trend used per annual report. A comparison of different working structures based on the Akaike's information criterion reveals that an one-period autoregressive correlation or AR(1) has the best model fit. The thesis contributes to the theoretical discussion about the information value of GETs and shows empirically that trend terms like “GETs” and ”megatrends” are rather artificial and have a strong subjective character. This was also confirmed by the longitudinal trend analysis that was based on the methodologies of mixed-method research, which combined and extended qualitative and quantitative methods. In addition, the thesis takes a unique approach optimizing the quality of multivariate models. By implementing web search data into linear regression and generalized estimation equation models, the overall performance of the models could be improved.