Do Powerups and Power Downs Matter?
How much does powering up Hive contribute to the price of Hive? What about users powering down their Hive? Does it have a meaningful impact and should we be concerned with it? Finally does combining power ups and power downs into net power-ups make any difference? In this post we will examine those questions.
This is the third post in a series where Granger Causality Testing is used to examine the factors that could contribute to changes in the price of Hive.
In the first post we looked at user activity on Splinterlands as a possible causal contributor to the price of Hive, and there was an explanation of the methodology. In the second post we looked at the number of users making posts and comments, to answer the question of whether the price of Hive drove user activity or if the reverse could be true as well.
Preparing the Data
The datasets we will examine today are the daily amount of Hive powered up, the daily amount of Hive powered down, and the net (which we get by subtracting powered down from powered up). We must first determine that these datasets are stationary. As with the last two posts, we will be testing them against daily changes in the Hive price (which has already been determined to be stationary).
The chart below shows daily amounts powered up in blue, powered down in orange (shown as negative to put it opposite power ups), and the net of both in grey.
Examining by eye, it is not immediately clear if this qualifies as stationary. There is no obvious trend, but there may be a seasonality factor, particularly in power downs, as the largest power downs often contribute to a 7-day cycle. So we examine all three with the Augmented Dickey Fuller test to be sure.
According to the results, all three datasets qualify as stationary. Presumably then, any 7-day cycle present in power downs was not strong enough to matter.
Granger Tests
Below are the Granger Test results for each dataset vs changes in the price of Hive.
Recall from the Splinterlands post how to interpret the data. In short, F represents the 'strength' of the effect, 'p-value' represents statistical significance, the probability of a finding occurring purely by chance. Statistically significant findings are highlighted in yellow.
Conclusions
Power Ups
Power Ups show some ability to predict changes in the price of Hive. Between 3 and 6 days later, there is a statistically significant impact on the price of Hive. The strength of the impact is stronger than the impact of a change in Splinterlands user activity, but more short-lived. It's comparable to the immediate impact of a change in the number of posts and comments, but again short-lived.
Changes in the price of Hive have no statistically significant ability to predict changes in the numbers of power ups.
Power Downs
Power Downs have no ability to predict the price of Hive (at least on their own, we'll get back to that). In fact, this is the weakest relationship of all examined so far, in all posts. It is however plausible that the impact of power downs is too spread out over time to be identified by this analysis, but the lack of clearly identifiable impact means it's probably not worthwhile to be overly focused on users who choose to power down their rewards. It's possible those users who are powering down their rewards are still offsetting this by their downstream impacts of user activity and the fact that powering down means more curation rewards and voting power for those who keep their Hive powered up.
Changes in the price of Hive have some predictive value in power downs, from 8 to 26 days after a change. Perhaps users choose to start power downs after a price rise, and we see the impact of that in the second, third and fourth weeks after. I have not examined the power down transactions themselves in this post.
Net Power Ups
Net Power Ups have a statistically significant impact on the price of Hive from 4 to 30 days after. The impact is of a similar level to a change in the number of Splinterlands players.
The price of Hive also has a modest ability to project net power ups about 4 weeks in advance.
When I started this analysis, I did not expect summing power ups and power downs to give any more information than those two elements alone, but I wanted to include it for completeness. As it turns out, net power ups have the biggest and longest lasting impact overall - meaning power downs do matter but only with the added context of how much is being powered up as well. Still, power ups appear to have the more clear impact of the two.
My statistics and analysis posts take many hours each to research, chart and write, so if you find them valuable and of interest to other Hivers, I appreciate your support in sharing, commenting, and/or upvoting my work. If you're interested in these kinds of stats posts, click the 'follow' button on my profile, or subscribe to the Hive Statistics Community which features daily Hive stats posts from 