Looking at all available price & network hashrate data going back to August 2010, once and for all we can put to rest (statistically speaking) of the chicken and egg of what causal relationship there is between $price of BTC and network computational power. Using a Granger causality test (in EViews) we find:

(

http://en.wikipedia.org/wiki/Granger_causality)

Pairwise Granger Causality Tests

Sample: 8/16/2010 11/22/2013

Lags: 4

Null Hypothesis: Obs. F-Statistic Prob.

H

_{0}A) LOG(HASH) does not Granger Cause LOG(PRICE) 1182 0.30372 0.8756

H

_{0} B) LOG(PRICE) does not Granger Cause LOG(HASH) 8.96203 4.E-07

What this says is that A) we Fail to reject the Null hypothesis that a change in hash does Not cause a change in price, because the F-statistic is very small. This means that the statement "hash does not change price" is likely very true to a high degree of certainty.

On the other hand in B) we see that the F-statistic is very high (significant) for the Null hypothesis that a change in price does Not change the hashrate. This means we should reject this null hypothesis with a high degree of certainty, and accept the alternate hypothesis: "A change in prices DOES cause a change in hash".

(I used natural log of each variable to normalize them a bit as they both have increased very rapidly lately. If you don't use the log, the results are even MORE significant).

This sort of throws a wrench in the labor theory of value in mining bitcoins, but I think the debate isn't 100% over yet on that front.

cheers.