The "100 companies is a representative sample" is a big assumption. The linked article says those 100 companies were all Fortune 500: that implies that there's a very large likelihood that they are bigger than the other companies in the Fortune 1000. Bigger companies naturally have more job openings.
Even if this is true, it doesn't necessarily mean the economy's doing better. It's pretty likely that the job openings require skills that the newly unemployed workers do not have. Otherwise, they'd be filled. You can't hire a former real estate agent at Google and expect him to be immediately productive. (Hell, you usually can't hire a software engineer at Google and expect him to be immediately productive, but at least you can hire him.)
The issue with recessions is misallocation of capital, and it can't be fixed with a magic wand. You generally have to wait for people to retrain in new, productive areas of the economy, and for capital to be liquidated and entrepreneurs to find new uses for it. That's why it takes ~4 years instead of ~4 months.
The samples were taken from a random selection of 100 of the top 1000 firms. The real question is how to make that selection of 100 firms so that statistically the sample size is a representative sampling. The goal is not to determine whether or not the jobs are available to those folks who may have just been laid off. The goal is to determine just how many job opening do exist in the Fortune 1000 without needing to hit every single job board of every one of those 1000 companies... surely there is a statistician who can solve that one? ;-)
"Yesterday we embarked on a research project to find current job openings. We started with 21 firms (with 18,839 open positions) and now we're up to data from 100 of them. (All Fortune 500)"
The chance that a random sample of 100 firms are all in the top half of the sample universe is 2^100 = about a decillion. Doesn't really look random to me.
Thanks, nice catch and that post needs to be changed... momentarily... The universe was truely from the Fortune 1000 from the list publish online... not the top 500 only...
So here is the methodology:
1. Hit the site above.
2. Randomly pick a company from the list... for instance:
3. Pick Paychex Systems... company 921 on Fortunes list...
4. On their site they list 137 job openings...
5. Add that to the total...
6. Repeat until you have a representative sample size
Question: Using an appropriate random number generation between 1 and 1000, what is the sample size required to then estimate the total job openings on the 1000 sites?
Even if this is true, it doesn't necessarily mean the economy's doing better. It's pretty likely that the job openings require skills that the newly unemployed workers do not have. Otherwise, they'd be filled. You can't hire a former real estate agent at Google and expect him to be immediately productive. (Hell, you usually can't hire a software engineer at Google and expect him to be immediately productive, but at least you can hire him.)
The issue with recessions is misallocation of capital, and it can't be fixed with a magic wand. You generally have to wait for people to retrain in new, productive areas of the economy, and for capital to be liquidated and entrepreneurs to find new uses for it. That's why it takes ~4 years instead of ~4 months.