**8. Conclusions and Contribution**

The crux of Gibrat's Law is that the best prediction one can make about future firm size is current firm size. This is the starting point of Gibrat's Law, whereas the stricter version adds to more hypotheses. Firstly, all deviations from this initial size come from a white noise error term, and secondly, the variance of this error term is independent of size. If all three hypotheses hold, we can say that the firms follow 'pure' random walks.

This is not the case for Norwegian campsites, in contrast to Italian and Dutch campsites, but in line with most other sectors and industries. We find that the size of Norwegian campsites is mean reverting, and its volatility decreases with size. Hypotheses 1 and 3 are thus rejected, the firms converge towards a 'natural' steady state, and they become more stable as they grow. Gibrat's Law does not hold for Norwegian campsites.

Furthermore, we find evidence of a threshold size for the Norwegian campsites, at which point their growth processes switches. At the point of about 25 employees, the distinction between the medium and large sub-samples, the processes change. When the threshold size is reached, there is no longer any gain of increased size in the stability of growth, and the current success/fiasco becomes a predictor of future success/fiasco. In addition, we can no longer reject a random walk after this threshold size. Consequently, hypothesis 2 is rejected for the large Norwegian campsites, whereas hypotheses 1 and 3 are not. This is the opposite result of the general result we obtained for all Norwegian campsites, and more in accordance with previous studies of campsites in Italy and the Netherlands.

As for hypothesis 4, we can see that a depreciation of the Norwegian Krone translates into higher revenue for the sector, as more foreign tourists choose to visit the country, whereas fewer domestic tourists choose to leave the country. The differing degree to which the exchange rate affects the three sub-samples is grounds for further research.

Hypothesis 5 shows that higher leverage leads to higher revenue streams, but not for small campsites. This can be due to an unwillingness or inability to invest or gain the means to do so. Whether the level of debt is positively related to profitability is another issue, investigated by Opstad et al. (2021b).

We used three estimators to obtain the autoregressive and moving average components of Norwegian campsites, the FD-GMM, SYS-GMM, and ML-SEM estimators. Although they have differing strengths and weaknesses, the results were similar. The steady state assumption of the SYS-GMM seems to not cause too many problems, comparing it to the other two. The Monte Carlo evidence against the FD-GMM estimator when the autoregressive component approaches unity would seem to make it inappropriate for testing Gibrat's Law, although our study does not show it conclusively. The ML-SEM estimator, combining the weak assumptions of the FD-GMM estimator with better precision than the SYS-GMM estimator, seems to be the best choice for testing the dynamic properties of firms, according to Monte Carlo evidence. We are not aware of any studies using it to investigate Gibrat's Law in the literature yet, this being an introduction of the estimator to the literature.

To conclude, another novel contribution of our paper to the general Gibrat's Law literature, is the evidence of a threshold size, for the tourism industry at least. We show evidence of this threshold size (non) rejection of hypotheses 2 and 3, and to a lesser degree hypothesis 1. At the point of 25 employees, in the case of Norwegian campsites, size no longer translates into more stability for the firm, but rather into a spill-over dynamic where current success/fiasco is carried over into the next year.
