This paper considers the fate of Norwegian firms in their first decade after entry. The underlying dynamics of entry and exit play an important part in the growth and development of an economy. On the one hand, there is net entry of firms in some industries or sectors and down-sizing in others to adapt to changing economic realities, such as demand shifts and relocation of production. On the other hand, there is a considerable excess turnover of firms within industries, which some researchers have explained as productivity-based sorting. This means that the entry and exit of different firms are intrinsically linked. The survival and growth of some comes at the expense of others.
This paper will consider the following problems: 1) What is the duration dependence of the exit hazard of an individual firm? Is the duration dependence positive or negative? 2) Can we observe any differences between the duration dependence of different types of exits? 3) What factors other than age can contribute to an explanation of why some firms exit while other survive? Are there differences in how these factors impact the risk of different types of exit?
Economic theory provides several different mechanisms that may account for differences in firm exit rates. Vintage theories emphasize the age of capital and rate of technological innovation. These are also known as theories of creative destruction or Schumpeterian growth. They predict an increasing exit hazard in the age of capital. Theories of learning can be divided into passive “learning about ones relative quality through market feedback” and active “learning-by-doing” (improvement in quality with time and experience). Passive learning is treated explicitly in this analysis, by both taking into account the observed differences and modeling the unobserved differences that exist between firms. These differences mean that firms are of different quality and have differing exit risks. This creates a sorting process. Since "low quality" firms tend to exit earlier than higher quality firms, we get a selection process that makes the “gross” observed exit hazard (i.e. ignoring differences in quality) decline in firm age. In addition to the two groups of theories mentioned above, business cycles and sector shifts have effects on the entry and exit of firms that differ by year and by industry/sector. These effects are not the focus of my analysis, but are controlled for by year and sector dummies.
The analysis has a particularly rich categorization of the different types of entries and exits. Using a comprehensive set of register data, we are able to distinguish between the plant and the firm level. This allows us to control for different types of entries, and identify those cases where the plant survives an exiting firm. We are also able to distinguish bankruptcies from other exits. The identification of different types of firm exit is exploited in two models of competing risks. Using a model with competing risks, we are also able to identify differences in the factors associated with different types of exit.
To decompose the gross exit rate I use a series of proportional hazard models with parametric and non-parametric assumptions regarding the unobserved heterogeneity. One of the models assumes that the unobserved heterogeneity in the sample is gamma-distributed. This model is estimated in a Stata program developed by Jenkins (1997). The other models are non-parametric in that they estimate a discrete distribution of the unobserved heterogeneity of firm. These models are estimated by an R-program developed by Simen Gaure at the Frisch Centre. By explicitly modeling the unobserved quality differences between the firms, we are able to separate out the selection/sorting effect from the duration dependency facing individual firms. The results turn out to be similar for the model with a gamma-distributed unobserved heterogeneity and the non-parametric model.
The analysis is performed in a number of stages. The observed hazard rate of firms is declining in firm age. . Taking differences in the observed heterogeneity of firms into account, the remaining duration dependency of the exit hazard is not significantly related to firm age. Estimating a model with unobserved heterogeneity turns the duration dependence strictly positive, which indicates that the exit hazard of firms increases with firm age when selection effects are taken into account.
Introducing competing risk models, I distinguish first between the cases of firm exits where the corresponding plant also shuts down ("full exit") and the cases where the plant continues under a new firm ("half exit"). I find that the duration dependency of half exits is more positive than that of full exits, meaning that a continuation of the plant after firm exit is relatively more common if the firm is older at time of exit. In the second competing risks model, bankruptcy is distinguished from other firm liquidations. The results indicate that bankruptcies differ significantly from other liquidations. First, the bankruptcy hazard has no discernible duration dependency, our results are not significantly different from one where the risk of bankruptcy is constant with firm age given firm quality. This suggests that the positive duration dependency of the total exit hazard is driven solely by the non-bankrupt exits. Second, the observed differences between firms, such as employment and debt-to-equity in the year of entry, are associated in starkly differing ways with patterns of bankruptcy hazard compared to liquidation hazard. For example, a firm with negative ratio of debt-to-equity has a substantially higher risk of bankruptcy than a firm with moderate debt-to-equity, but seems to have lower risk of a non-bankrupt exit.
Some of the results were significant in interesting ways. Interestingly, a high share of female employees in the year of entry is associated with a lower risk of bankruptcy. The results in the single risk model indicates no relationship between the initial employment size and the exit hazard. This surprising finding can be explained in the competing risks model, where the employment size has opposite effects on bankrupt and non-bankrupt exits. High employment is associated with a reduced risk of bankruptcy and a higher risk of liquidation. The competing risk models demonstrates that ignoring the distinctions between types of exits obscures underlying differences in the duration dependencies for different types of exits.