*3.2. Results from the FIGARCH*

Both BECI UB d and BECI LB d indicated a completely stationary process with long memory (see Table 3). The BECI average likewise exhibits a long memory process. However, no cases of short memory (*d* = 0) were detected. Logically, the effect of long memory changes with the sliding window. The results are consistent in terms of the Hurst exponent (except for one instance, where BECI LB showed anti-persistence). BECI UB displayed a persistent pattern in 17/17 cases, whereas BECI LB showed persistent values in 16/17 cases. The BECI average showed a clear persistent pattern in 17/17 cases with the H value being higher than 0.5. A notable 88% of the BECI UB indicated extremely long memory (H > 0.85), whereas BECI average stood second, with 82%, followed by BECI LB, with 76%. Given that hardware costs are substantial and electricity prices are not constant (globally), upperbound-based calibration is sensitive to various cost matrices [13]. Accordingly, the BECI UB cannot be taken too seriously when making policy decisions. The BECI LB can be considered a better proxy of electricity consumption, and thus suitable for measuring the carbon footprint of the Bitcoin market.
