4. Data
4.2 Firm data
4.2.3 Financial data
The various financial data chosen for the estimation reflects characteristics of consumers that the switching costs could be affected by. Rather than just financial figures, the variables should be seen as characteristics of the firms. Each variable is reviewed below. The database includes financial reports of firms that are not active anymore, as well as historical financial reports. The bank connections are not registered with a date, so they are all assumed to be the current bank
connections. It is not clear how the bank connections are registered, but it is unlikely that they are all collected on the same day and updated daily. The financial data used, has to be the latest available to reflect the current bank connections, and the time the switching costs are estimated. The financial report also has to be for 2011 or later, to ensure that the characteristics of the firm are current.
The dependent variable of the estimation is the consumer switching costs estimated empirically from the firms’ bank connections. All customers at each bank thus have the same switching costs. The variable will be extensively reviewed in section 5.2.
The balance and capital of the firms are included in the estimation to model the size of the firms.
Both figures are non-‐negative and have a high amount of variance across individual firms, and is often highly dependent on the industry of the firm as well as the firms’ financial policy. The balance of the firm is sensitive to accounting methods, and items that may not be relevant for the size of the firm can also be included in the balance. Capital of the firm can also vary between otherwise similar firms. One can be undercapitalized while the other can be overcapitalized. The high variance across otherwise identical firms can decrease the predictive power of the variables. The variance across time is on the other hand not very large compared to other financial figures that may change
significantly from one year to another. Both of the variables are skewed across firms, due a small amount of very large firms with very high balance and capital. To reduce this skewedness, both variables are logarithmically transformed. Alternative measures of market size could be output of the firm, market capitalization or revenue, which may be more suitable for some subgroups of firms.
Output and market capitalization is however not available for all firms in the dataset. Revenue is registered for around 50% of firms, but it is optional for firms if they want it published, so it is not included in the estimation.
Firms’ profit are available through the financial reports and possibly has influence on the switching cost of the firm. Firms’ profit provides some indication of the current financial situation of the firm, and thus its ability to pay back its loans. Banks are generally not interested in acquiring new
customers with high credit risk. The profit itself is not particularly interesting, when it is not related to another financial figure that contains information about the base upon which the profit is earned.
The hypothesis being examined is concerned with the switching costs of good and bad customers, so it is not interesting to estimate exactly how much switching costs change when the profit change.
Therefore a dummy variable that is 1 if the firm earned a positive profit and 0 otherwise is used instead of the original variable. The profit of firms also varies considerably across firms, most likely more than the balance and capital, but also over time. A firm’s profit, especially in a distressed economic environment as the one the firms in this thesis operate in, can vary significantly from year to year. Using the dummy variable, instead of the actual profit, should reduce this effect.
The equity of firms is included in the estimation because it is a central financial figure. It is
dependent on the assets and liabilities, as well as the size of the firm and the profit over time. It is therefore possible that the amount of equity a firm has is correlated with some of the other variables in the estimation. While it is related to many factors, the amount of equity a firm has is a financing decision, and can vary from firm to firm. Large outliers characterize the data, as the balance and capital figure, so a logarithmic transformation is applied to reduce the effect the of outliers. Equity can be negative, and there are both positive and negative outliers. Since the logarithm of negative numbers or zero is not defined, the lowest equity in the dataset plus one is added to all equity figures. It obviously does not make sense to interpret the parameter value, but since the concern of
this estimation is only the dependencies and not the actual parameters, there is no loss of generality due to the logarithmic transformation.
The total amount of debt each firm has is also included in the estimation. It is calculated as the sum of the short-‐term and long-‐term debt for each firm. The debt is related to equity and balance by the accounting identity that says debt and equity must equal assets. Including debt could thus possibly cause multicollinearity, which could affect the parameters estimations of individual variables.
Multicollinearity cannot reduce the predictive power or reliability of the estimation, but should be avoided to reduce the standard errors of the parameter estimates. In the real world, the accounting identity is not that simple. There are other instruments such as subordinated loan capital and other instruments that are not short-‐ or long-‐term debt and not equity. These instruments are therefore not included in the dataset variables used for the estimation. Including the debt in the estimation should therefore not cause multicollinearity. Like many of the other variables, the calculated variable debt is skewed to the right with many large outliers. A logarithmic transformation of the variable is therefore used in the estimation.
The solvency ratio is the only financial ratio included in the estimation. It is defined as equity divided by total assets. The solvency ratio is an expression for a firm’s ability to incur losses. It measures the percentage of capital the firm can lose before the more senior financing is affected. The solvency ratio is negative if the equity is negative, which is a clear sign of financial distress. The solvency ratio has a tendency to be very negative when it is negative. In the database the solvency ratio is capped at -‐999 and 999, which is relevant for some firms in the dataset. The solvency ratio is very sensitive in its nature, with many outliers and high variation, so the solvency ratio is logarithmically transformed in the estimation to reduce these effects. The solvency ratio can be negative, so the database cap plus one is added to all solvency ratios. It could be preferred to use a dummy variable to proxy firms with a good solvency ratio and a bad solvency ratio. It is however not obvious what a high and low
solvency ratio is for the entire population of firms, so a continuous scale seems like the better choice3.
3 Using a dummy variable yield equivalent results.
The dataset used for the estimation contains all available bank connections of firms that satisfy the requirements. If firms have more than one bank connection, then they appear more than once in the dataset. It is interesting to examine if it has an effect on the firms’ switching costs. A simple variable that contains the number of bank connections each firm has in the dataset is therefore included in the estimation.
The last variable that is included in the dataset is the year of establishment. It is not a continuous variable and there are too many years to include each year as a dummy. Therefore the seven groups used in Figure 5 are used. The groups are included in the estimation as six dummy variables, with the group of firms established before 1950 as the default group. Each dummy thus represents the
switching costs of the firms that were established in the time period, compared to firms that were established before 1950.