A credit conditions index for Norway

Fernandez-Corugedo and Muellbauer (2006) represents a novel paper where a credit conditions index (CCI) is estimated for the UK. The CCI is intended to capture supply-side shifts that are due to structural changes and other shocks to the financial sector, controlling for e.g. changes in the interest rate or the level of output. This index will therefore allow researchers to control for the institutional development in their empirical work and take into account how some of the mechanisms in the credit market can have changed over time. The index of Fernandez-Corugedo and Muellbauer seems to capture the main developments in the UK including the extensive deregulation of the credit market during the 1980s. Succeeding works (e.g. Aron, Muellbauer, and Murphy (2008)) have proven that the including the CCI improves the performance of some econometric models. If the methodology of credit condition indices is to gain increased relevance and usage more empirical evidence is needed. The main objective of this thesis is therefore to estimate a CCI for Norway, following the setup and method of Fernandez-Corugedo and Muellbauer (2006). To do so I go through several steps. Chapter 2 outlines how the regulation of Norwegian credit markets has developed since the 1970s. I have combined various sources in order to provide a relatively detailed summary of the deregulation process that took place. As far as I am aware of, there exists no other unified presentation of this kind, making this chapter important in its own respect. This also motivates that a more detailed version of Chapter 2 will be made available in Krogh (2010, forthcoming). In Chapter 3 I consider what tendencies in credit conditions that can be drawn from the mortgage survey conducted by Finanstilsynet. The purpose of Chapters 2-3 is to provide sound qualitative evidence for the structural development of the Norwegian credit markets since the 1970s. The requirement for a sensible estimate of the CCI must be that it is in line with this qualitative information. Chapter 4 contains the model I will use to estimate the CCI and also sketch what I think the CCI will look like. I formulate a system of two equations to explain the development in secured and unsecured debt relative to income. Chapter 5 explains how the model is estimated using a maximum likelihood approach and it also provides some considerations of practical problems that may arise when the model is estimated. The end result of this chapter is a Stata command that is tailor-made to estimate the model. The description of the maximum likelihood framework applied and the derivations that accompany it can be useful for other researchers that want use their own maximum likelihood codes to estimate a nonlinear SUR model. This can be relevant if their model entails a variation that is not permitted by the standard commands that exist. The estimates are presented in Chapter 6. I detect long-run relationships that have fairly reasonable coefficients, but with some exceptions. The implied CCI has a shape that matches most of the ex ante expectations that I have, given the evidence in Chapters 2-3, but both the CCI and the demographic variable seem to get too large coefficients in the equation for unsecured debt. I judge my results to represent a very useful first step, but I do think that the model probably ignores an important interaction between secured and unsecured debt that has led to a shift from the latter to the former type of debt. I describe how the weaknesses of my results point in this direction. Chapter 7 concludes. All estimations have been performed with the statistical software Stata (StataCorp, 2007, Rel. 10)