Lars

Introduction#

Eerlijk gezegd vond ik het beschrijven van het te gebruiken model erg lastig. Daarom hoop ik dat we hier tijdens de feedback sessie iets meer aandacht aan kunnen besteden en ik de methodologie verder kan uitbreiden.

Central research questions:

• Does leasing crowd-out existing debt funding (is it a substitute) or is it rather a complement to existing debt funding?

• What is the effect on $\text{Leasing} \Rightarrow \text{Debt Financing}$ of the new IFRS (2019) legislation?

The explanation for the found discrepancy between substitution and compliment theory is found in an identification problem which is present in prior research. In prior used models the relation between leasing and other forms of debt financing is a potentially unidentified mix of both the true relation and other simultaneously affecting factors. This endogeneity problem is the most likely reason causing the wide array in results between prior studies. To tackle this problem a generalized method of moments technique can be used as this enables the ability to remove the endogeneity problem. As such providing a better explanation for the relation between leases and other forms of debt financing.

This is a little bit vague: what other factors? Spelling this out more precisely might introduce the intuition that leads you to the usage of (presumably) the Arellano-Bond estimator (the GMM estimator you are talking about?)

• Your data set is at the firm-country-year level, right?
• The AB estimator requires lagged data to instrument contemporaneous data

Structure#

The methodology section lacks structure:

I would start firstly with the data, sample, variable definitions and model.

• It is effective to make a table with variable definitions in your case.
• In your robustness analyses, you should make sure that the results you obtain are robust to changes in the variable definitions (which are to some extent arbitrary).

It is not clear to me why you would use tobit (or OLS) to find out whether debt and leasing are substitutes. What is the model used in these studies? Is it:

$$Debt_{it} = \alpha_i + \beta \cdot \text{Leasing}_{it} + \epsilon_{it}$$

(With $\beta > 0$ in case of complements and $\beta < 0$ in case of substitutes?) And are you planning to estimate:

\begin{align*} Debt_{it} = \alpha_i + \beta_1 \cdot \text{Leasing}_{it} + \\
\beta_2 \cdot \text{DumIFRS}_{it} + \\
\beta_3 \cdot \text{Leasing x DumIFRS}_{it} + \epsilon_{it} \end{align*}

This technique enables testing of the true relation between leasing and other forms of debt in isolation as it is able to control for firm fixed effects as well as correlation between both variables.

Correlation between debt and leasing should be there, right? There is some dynamic aspect in the data (e.g. changing corporate culture, or other unobservables that should be correlated with the outcome (debt?), in order for the FE-estimator to be inconsistent. That should be the reason why you should use the Arellano-Bond estimator. Try to be specific about the Yen (2006) study, and how they (concretely) motivate the usage of the AB estimator.

Sometimes you still include some theory in your methods section:

Since growth firms on average have less internal funds, overinvestment tends to be less of a problem. Therefore growth firms should make less use of both leasing and other forms of debt financing (Yan, 2006).

I think it should be motivated in the theory section, and you should just state the model with the control variables (possibly briefly recapitulating the logic behind their inclusion) in the methodology section.

My intuition is that you can test all of these theories by including an interaction effect-based design:

$$ Debt_{it} = \alpha_i + \beta \cdot \text{Leasing}_{it} + \\
\gamma \cdot \text{Distress}_{it} + \delta \cdot \text{Leasing x Distress}_{it} + \\
\phi \cdot \text{DumIFRS}_{it} + \epsilon_{it} $$

Possibly, you can still use Tobit-based logic (censored regression, for firms that have 0 debt), if there are censored observations in your data set.