Based on these ARDL bounds testing results, we regarding cannot provide a Hypothesis 2. Instead, these results call for further analysis using more disaggregated location-specific mining cost data, given that the effect of electricity prices in North America is in line with expectations, as derived in Equation (1), whereas the opposite relationship was found for electricity prices in China. robust answer

and Increasing the difficulty of mining reduces the incentives for mining and in the presence of learning by mining increases the concentration of mining activities, as such reducing the blockchain security. Our estimates suggest that negative network externality dominates positive network externality. Our results are in line with those of Parra-Moyano, Reich, and Schmedders (2019), who found that the probability of winning a mining contest increases with the miner size. This motivates miners to join mining pools to increase their probability of winning the mining contest and receiving the reward.10 Indeed, our competition proxy variables were constructed based on the observed number of miners but not on the number of members within mining pools. Since greater competition may imply fewer miners (because many individual miners join mining pools), the implied actual long-

Competition and network externalities The variables measuring the miners competition level had a negative impact on the security outcomes of the bitcoin blockchain; all estimated coefficients were statistically significant. These results suggest that a permanent increase in competition intensity exercises downward pressure on the mining computer power in the long-run. Digital distributed ledgers, such as blockchains, are subject to a number of network externalities. When new miners enter the mining of blockchains, two types of direct network externalities related to blockchain security arise: one positive and one negative. The positive network externality implies that blockchain security increases with the number of miners because each additional node reinforces the chains security. Consistent with the previous literature (Waelbroeck 2018), negative network externality occurs because each individual miner invests in the mining-computing power, which increases both the individual

Hardware efficiency. In line with the theoretical model, the hardware efficiency variable has a statistically significant negative impact on the mining computer power in models M1.1, M1.2 and M3.1. The estimation results imply that an increase in the efficiency of mining equipment (decrease in the input units of the mining hardware efficiency per security output unit) leads to an increase in the security outcomes of the bitcoin blockchain in the long-run. The estimated elasticities vary between 0.6 and 0.8, implying that a 1% permanent increase in the efficiency of mining equipment increases the mining computer power in the long run by between 0.6% and 0.8%. Hence, the bitcoin blockchain mining security is dependent on the mining technology available at a given point in time. Time preference The 10-year-treasury variable, which is a proxy for the discount rate, had a statistically significant positive impact on mining computer power.