**4. Results**

#### *4.1. Results from Mesurement Model*

The analysis of the measurement model was done in [7] and quantitative results are summarized in Tables A1–A3 of annex. Table A1 shows basic descriptive statistics of our sample. Results in Table A2 show that factorial analysis detects only one dimension in all scales. Bartlett sphericity test has always a *p*-value (<0.0001), whereas Kaiser-Meyer-Olkin statistic, was always ( ≥0.5). The percentage of variance explained by the factors was in all dimensions (>70%), which confirms scales suitability. Regarding the evaluation of the measurement model, Table A2 of annex suggests that results on this concern are correct.

Constructs present a composite reliability and Cronbach's alpha always >0.7, confirming so that reliability was correct (see Table A3 of annex). Average variance extracted (AVE) in all scales is greater than 0.5. Therefore the convergen<sup>t</sup> validity criterion was thus met. The HTMT values were correct in all cases (<0.9). Likewise, the square root of the AVE was higher than the correlations between constructs, i.e., discriminant validity criterion has been also accomplished (see Table A4).

#### *4.2. Results from Fuzzy Set Comparative Qqualitative Analysis*

A previous step to develop fsQCA implies implementing so-called necessity analysis [73]. It consists in stating cons and cov in (7) and (8) between each individual input (in affirmed and negated form) and the output (also affirmed and negated). Results are in Table 3. This analysis lets stating the degree in which and individual factor that is affirmed/negated is necessary to induce the output (or the negated output).


**Table 3.** Necessity analysis on IU and ~IU.

Table 3 suggests that there is an asymmetry in the explanation of IU and ~IU. Regarding IU are "almost always necessary or sufficient" PE, EE, FC and FL. When analysing ~IU, again negated PE seems to be the main cause (cons and cov > 0.9) but also ~SI (e.g., news about bad experiences with cryptos) presents grea<sup>t</sup> cons and cov. Notice that SI did not present high cons to produce IU. PR (and not ~PR) has a cons > 0.8 in IU. Then, the presence of perceived risk can incentive some people to use crypto. These persons act as risk lovers, e.g., they buy cryptos as a method for betting, or as an investment with a high expected return due to its grea<sup>t</sup> volatility. This fact might sugges<sup>t</sup> rejecting H5. However, the empirical weight of that finding (cov < 0.5) diminishes its relevance. When analyzing the influence of PR on the negation of IU, it can be checked that PR (and not ~PR) causes also ~IU (cons > 0.8, cov > 0.7). Therefore consistency and coverage of PR on ~IU outline PR as a relevant factor to reject the use of cryptos.

Table 4 shows configurations of QCA-IS and QCA-PA of model (1). To generate intermediate solution we have supposed for non-covered Boolean configurations of outputs (19 over 26 = 64 possible configurations) that PE, EE, SI, FC and FL cause IU only when they are present. It is according to our hypothesis, findings in literature described in Section 2 and with necessity analysis in Table 3. So, due the contradiction between H5 and necessity analysis of PR, we suppose that either presence or absence of PR may cause IU. Analysis of intermediate solution lets appreciating that:



**Table 4.** QCA-IS and QCA-PS for the model IU = f(PE, EE, SI, FC, PR, FL).

The results of fsQCA over IU = f(PE, EE, SI, FC, PR, FL) sugges<sup>t</sup> the acceptance of H1, H2, H3, H4, and H6. On the other hand, the assessment of H5 must be nuanced since its rejection would not imply that PR does not influence IU. Some causal configurations sugges<sup>t</sup> that PR may stimulate IU to some users when it is present and to others if it is absent.

Table 5 shows configurations of QCA-IS and QCA-PA in (2), i.e., for the explanation of ~IU. To generate intermediate solution in ~IU we have supposed for non-covered Boolean configurations that PE, EE, SI, FC and FL cause IU only when they are absent. It is congruen<sup>t</sup> to that we done for intermediate solution in Table 4 and with the results from necessity analysis in Table 3. The same argumen<sup>t</sup> applies to justify supposing that PR can be either present or absent to cause ~IU. After checking Tables 4 and 5 we find that the explanation of IU and ~IU by causal recipes of input factors is far to be symmetrical. So, QCA-IS for ~IU shows that:



**Table 5.** QCA-IS and QCA-PS for the model ~IU = f(PE, EE, SI, FC, PR, FL).

So, it seems again that, with the exception of H5, we cannot reject any hypothesis in Section 2.

## **5. Discussion and Conclusions**

There are not so much empirical studies about variables influencing cryto acceptance because of the novelty of blockchain techs. This paper contributes to literature on cryptos and employs an original analytical tool in this context, fsQCA. We have found that complementing UTAUT modelling with fsQCA lets discover relations between variables that influence crypto use that we did not in [7] by using conventional PLS. So, whereas [7] only found relevant three variables to explain the use of cryptos, we have discovered that all factors are relevant. Likewise, with the exception of the hypothesis on PR, all other are confirmed. As in [7], PE is revealed as the most decisive variable to explain IU. EE is also

a relevant variable when it is combined with SI and also its absence could be a sufficient condition for ~IU. Similar considerations can be made about FC, that was relevant in [7], but also for FL, that in [7] was not. Let us remark the importance of social influence to explain IU. Its sole presence or absence is never relevant for IU or ~IU. However it is present with the expected "sign" in recipes that allow EE, FC, PR or FL to be relevant. So, it seems that SI acts as a facilitator factor to induce the other input variables to be relevant for IU.

Evidences do not support that the influence of PR on using cryptos is necessary negative. However this fact does not imply that it is not relevant to explain IU. Depending on the context PR may have a positive or negative influence. We find configurations with PR (i.e., its presence affects positively using cryptos) but also with ~PR (i.e., not perceiving risk also influences positively using cryptos), so the configuration in QCA-IS (SI\*FC\*PR) may explain behaviors of people with FL that from information by close people (SI) use cryptos consciously as a risky asset. Likewise the configuration PE\*PR may explain utilizing cryptos as a bet method. On the other hand the configuration EE\*~PR\*FL explains the behavior of persons with financial knowledge that consider cryptos easy to manage and that, in their context, its use have low risk.

The main objective of this paper was complementing conclusions about IU cryptos in [7] by applying fsQCA instead PLS. As any regression methodology, PLS allows quantifying average influence of each factor over IU by means of a coefficient. Likewise, it supposes a symmetrical impact of the presence/absence of an input variable over the presence/absence of output. By using fsQCA we cannot quantify in a coefficient the average weight of one factor into the intention to use crypto and this fact suppose a drawback. Likewise fsQCA is quite sensitive to how membership functions are built up and also to outliers. On the other hand, fsQCA can discern how factors are combined to produce (or not produce) IU, so the application of conventional correlational techniques in [7] led us to conclude that SI was not a significant factor. However, fsQCA allow checking the important role of SI to induce IU since its presence lets EE, FC and FL to be relevant. Likewise fsQCA lets discover asymmetrical relations between variables. This fact allows us detecting that FC alone is not enough to generate IU but its sole absence influences decisively the intention of not using cryptos. Likewise, the use of fsQCA lets us stating that the non-significance of PR in [7] (i.e., the coefficient that quantify the influence of PR on IU was not significantly different from 0) does not mean that PR is irrelevant to explain IU. It may be caused by the balance of responses from people averse to risk and risk-lovers.

We are aware that this study has some limitations. These constraints are an incentive to conduct further research. This paper is focused on a concrete population segment: collegeeducated adults with Internet skills. Future studies should focus on other household segments but also on other economic agents (small business, transnational corporations, institutional investors ... ). Another constraint is that this research is circumscribed to Spain. Conclusions might be slightly different if the survey had a wider geographical extension or if it were answered in another country, so the use of an international database may allow improving the conclusions that we have extracted in this work. Other issue that could be assessed in future research is the sustainability of blockchain mining. In [76] it is stated that the mining process requires intensive computation resources with large energy consumption. Therefore sustainability factors can impact on the development of blockchain technology. It is an emerging technology that is evolving continuously. Therefore, the findings of this research should be interpreted under above considerations.

Lastly, let us point out that the use of alternative analytical tools to fsQCA based on fuzzy sets as fuzzy correlation indexes [77,78] and fuzzy multiple criteria decision making methods [19] might be also object in future application on this matter.

**Author Contributions:** M.A.-O. and J.P.-B. have collected data have built up model to validate and have validated scales. J.d.A.-S. have implemented fsqca. All authors have contributed equally to extract conclusions from data and in literature revision. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the COBEMADE research group at University of La Rioja (REGI 20/40) and Programa RIS3 La Rioja. CAR-PID2019-105764RB-I00.

**Institutional Review Board Statement:** With regard to ethics approval: (1) all participants were given detailed written information about the study and procedure; (2) no data directly or indirectly related to the subjects' health were collected and, thus, the Declaration of Helsinki was not generally mentioned when the subjects were informed; (3) the anonymity of the collected data was ensured at all times; and (4) no permission was obtained from a board or committee ethics approval, it was not required as per applicable institutional and national guidelines and regulations (5) voluntary completion of the questionnaire was taken as consent for the data to be used in research, informed consent of the participants was implied through survey completion.

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** Survey supporting the study can be obtained by demanding it to any author.

**Acknowledgments:** Authors acknowledge helpful comments of anonymous reviewers.

**Conflicts of Interest:** Authors declare having no conflicts of interest.
