Variables

3.2.1 Dependent Variable: Collaboration with fintechs and new partnerships

Merger and acquisition, as well as a product-based collaboration and a partnership were considered. The collaboration was coded as 1 if the bank made a partnership/ collaboration/ majority stake/ minority stake/ merger/ acquisition with a fintech in a particular year and 0 if the bank did not.

In addition to that, the variable Number of new collaborations which was used to identify the number of collaborations the bank was involved in any given year was analyzed as well. For the collaboration with fintechs, the researcher ran an extensive search using the official websites of the identified banks to find press releases, Crunchbase database to identify fintechs that formed partnerships with banks and FactSet, particularly a section for the global news monitoring. There are several requirements for this search. First, banks must be located in one of the 4 countries identified earlier, yet fintech can be anywhere in the world. Secondly, the collaboration between banks and fintechs must be announced in the last 8 years between the fiscal year 2013 and 2020.

3.2.2 Explanatory Variable: Purpose

In this paper, the purpose is measured by CSR performance that is proxied by Factset's Truvalue Labs ESG score identified by SASB (Sustainability Accounting Standards Board) through the FactSet database. Different ESG scores as proxies of CSR performance were used in academic literature before (Deng et al., 2013). For instance, Gomes (2019) used the Thomas Reuters ASSET4 database as the measure of CSR initiatives. These scores provide comprehensive information on the environmental, social and governance dimensions of the company. The researcher used Truvalue Labs over the others since it covered 120,000 public and private companies across the globe and different industries including banking. Large quantities of unstructured data are analyzed and sorted by deploying machine learning algorithms into 26 categories outlined by the SASB materiality framework to create Truevalue ESG scores (Factset, 2022). SASB categories include Environment, Social capital, Leadership and Governance, Human Capital and Business Model Innovation. The most of information comes from 100,000 data sources in thirteen different languages. They include

company-provided materials, analysts' perspectives, advocacy groups, and government and law regulators as reported by independent media. Because of the diversity of data sources in terms of the different languages, the researcher assumes that scores for all the countries represented in the sample get a full scope of the analysis. The scale of the score lines in 5 different levels: laggard, below average, average, above average and leader, each given score of 1, 2, 3, 4 and 5, respectively. For the sake of the analysis scores from 1 to 3 were recorded in 0 and 4 and 5 to 1. The ESG score quantifies the sustainability of CSR initiatives and allows to run a regression model to get statistical results on the relationship of the variables.

Figure 2. 5 dimensions of FactSet Truevalue Labs ESG Score based on SASB categories

3.2.3 Control Variables

Previous studies in the banking literature suggest incorporating the control variables for the increased validation of the research (Peng et al. 2017). Control variables in banking literature include the financial indicators, such as equity ratio, natural log of total assets, return on average assets, and banks' features like the age of the bank, whether it is publicly listed or not. Control variables in banking literature are usually used to control for the size of the bank, the financial capital structure, organizational structure, management decisions, and age (Peng et al., 2017).

Leadership and Governance

Business Model Innovation Human

Capital

Environment

Social Capital

Thus, the natural logarithm of total assets is used to control for the size of banks presented in the dataset, usually, scholars use a natural logarithm to normalize the distribution and to achieve a constant variance that allows comparing different banks in one dataset. With the big size in terms of total assets, firms are more likely to pay higher price premiums to acquire another company (Moeller et al., 2004). It can be also connected to the partnerships where the fintechs might be forced to enter the alliance for a big monetary incentive. Moreover, big banks collaborate with fintechs more often than smaller ones (Hornuf et el., 2020). Based on the literature, this control variable is included in the model to fix the effect of the bank’s size on the probability of collaboration.

The equity ratio is used for identifying the capital structure of the firm. In the case of a small equity ratio, it means that the most of assets were acquired through the debt and indicates that there is higher pressure in terms of profit results since there is a substantial amount of debt obligations to fulfil that can in return influence the interaction between purpose and collaboration variable in both directions. On one hand, focusing on profit maximization instead of stakeholder maximization can drive the purpose of the company downwards. On the other hand, willingness to collaborate with fintechs might increase since they are a source of innovations and new revenue streams.

Organizational structure is reflected in whether a bank is publicly listed or not. Companies are usually more pressured regarding reporting their activities in the case of the former, which can potentially have an influence on the variables. In order to check the efficiency of the management team and their decisions, in banking literature (Peng et al., 2017), it is common to use Return on Average Assets. It reflects the performance outcome that is directly connected to the potential decision about the collaboration, which is one of the dependent variables. The researcher uses them in order to make sure that the independent variable has a significant effect while control variables are held on a constant level. The data on the financial performance and indicators were derived from the ORBIS database and Crunchbase database.

According to Hofmann and Gavin (1998), for the sake of the interpretation of the results in a meaningful way all the continuous variables used in a regression analysis must be centered around

their mean instead of the zero. Consequently, equity ratio, natural log of total assets, natural log of banks’ age and ROAA were centered by subtracting the mean of each of these variables from the initial value of individual observations. The data points were not centered before the logarithmic transformation since some of the values were negative and according to logarithmic rules, it is not possible to transform negative values.

Table 3 provides the variables that are going to be used in this research.

Variable Explanation Data Source

Collaboration (d) Binary variable coded as 1 or 0:

1 in case of at least the partnership/collaboration/M&A

Press releases, annual

Purpose (d) Binary variable coded as 1 or 0:

1 in case of the purpose-driven business model

Truevalue Labs ESG score from FactSet database

Equity Ratio Ratio of the equity in the bank FactSet and Orbis databases

Total Assets Natural Log of the total assets to control for the bank’s size

Orbis database

Listed (d) Binary variable coded as 1 or 0: 1 in case of the publicly listed

FactSet and Orbis databases

ROAA Ratio of the return on average assets of the bank in the year

FactSet and Orbis databases

Age Natural logarithm of age of the bank to control for the age FactSet database, official websites

Table 3. Overview of the variables used in this research. d indicates the dummy variable