I would be cautious about drawing conclusions about startups from this study. Startups are such a small subset of newly founded businesses that unless a study is very careful about how they select their sample, they're likely to get mostly ordinary businesses. And judging from what the people who did this study say about how they selected their sample (see page 8), it sounds like that's what happened. They say they "extracted randomized records by region." Randomization is not the way to select a population who by definition are outliers.