Network externalities on retail payments market: evidence from russian merchants

Finally, in a number of robustness checks I also control for the shadow economy and economic crimes intensity, as Russia was included in top-5 countries with the greatest shadow economy level (about 33.7% of GDP according to IMF survey). Previous studies hypothesized that the grey activity in a sector may affect probability of cashless payments acceptance. On the one hand, merchants are more likely to refute transparency offered by cashless payments (Malphrus, 2009). On the other hand, regional governments may stimulate cashless payments acceptance if tax evasion practices are widespread and region suffers from shadow economy segment (Chernikova et al., 2015; Chizhikova, 2013; Krivosheya et al., 2015). Two control variables were considered. The first one is the logarithm of the average size of additional fees resulting from checks, measured in thousand rubles. The second one is the logarithm of the number of economic crimes in region. Data is taken from Rosstat.

3.6 Estimation method

Previous literature used binary choice models (e.g. Loke, 2007; Bounie et al., 2015, 2017) to model the probability of merchant's card acceptance. I use a standard binary probit model as it allows measuring the strength of influence of the explanatory variables on the dependent variable and is appropriate for determining core factors that influence probability of card acceptance by merchants. This study follows Bounie, Franзois, and van Hove (2016) and uses probit to estimate merchants' acceptance probability. Besides, probit model is undertaken following Krivosheya and Korolev (2018), who used the same dataset as in this study and used probit to estimate the selection equation before implementing a two-step Heckman section model. Probit and logit models provide similar results but the probit model might provide better estimation of probability at the expense of computational complexity. Unlike linear probability model, the resulting probability estimates do not exceed the range from 0 to 1. Also, probit allows for non-constant marginal effects across sample. Economic significance is analyzed using the marginal effects calculation at average values of independent variables. Robust standard errors are used in the analysis to control for potential problems with errors such as heteroscedasticity.

Descriptive statistics for the variables used in the analysis are provided in Table 1, outlining the minimum and maximum values, as well as standard deviations and mean values. Multicollinearity is one of the potential problems for the parametric regression because it leads to inconsistency of standard errors and hinders the power of regression. To account for it, correlation matrix, presented in Table 2, was analysed. Cross-correlations of the variables included in the analysis indicate that there are some signs of multicollinearity among the variables chosen for the models, but the exclusion and inclusion of the correlating variables and the series of the variables doesn't change the results of regression. Key variables, however, do not exhibit the signs of multicollinearity. I have decided to leave those variables that could potentially lead to multicollinearity, as all of them are theoretically important for the explanation of why merchants may or may not accept cards for payment and, therefore, mitigate omitted variable bias. Most of the specifications do not include highly correlated variables simultaneously.



4.Results

4.1 Unilateral tests

At first, statistical tests are performed to examine the relationship between the probability of acceptance and the explanatory variables. Although simple correlation coefficient might provide the misleading results in case of a binary variable, some insights are unveiled by looking at the correlation between the probability of merchant's card acceptance and the perception of the merchant about competitors acceptance. It is positive (0.40) and statistically significant at any reasonable significance level. The correlation between the probability of card acceptance and average acceptance level is also positive (0.24 for the Regional level and 0.14 for the Federal Region level) and statistically significant at any reasonable significance level. Therefore, there is some correlation between my key explanatory variables proxying the network effects and the fact that merchant accepts payment cards.

To test it more formally, I conduct the comparison of means test using the sub-samples of accepting and non-accepting merchants. First, I compare the means of the perceived share of competitors accepting payment cards. The corresponding t-statistic for the comparison of means test is -12.57, which means that the perceived share of competitors accepting payment cards is larger across the accepting merchants. Similar is also true for the other variables proxying the direct network effects. T-statistic for the equality of means of the actual (federal) regional average acceptance rate is -7.01 (-4.02). Hence, direct network effects variables are indeed larger for the sub-sample of accepting merchants, which supports the hypothesis H1.

In order to pre-test the hypothesis H2 I conduct similar equality of means test using the indirect network externalities measures. Simple correlation between the probability of card acceptance and average card holding is positive (0.09 for the Regional level and 0.11 for the Federal Region level) and statistically significant at any reasonable significance level. T-statistic for the equality of means test of the average cardholding rate variables is -2.21 (-3.13) at (federal) regional level. The correlation coefficient between the probability of card acceptance and average card usage is also positive (0.08 for the Regional level and 0.11 for the Federal Region level) and statistically significant at any reasonable significance level. T-statistics for the equality of means across the sub-samples are respectively -2.17 and -3.1 at the regional and federal region levels for the usage variables. These results are in line with the mechanisms developed in the theoretical framework section and support hypothesis H2. However, to test these hypotheses properly I conduct the multilateral tests set-up in the previous section.

4.2 Multilateral tests

In order to answer my main research question I estimate the card acceptance probability model using the probit estimation method as outlined in the empirical set-up. I first test the effect of the direct network externalities in order to test hypothesis H1. The results are presented in Table 3.

At first, I introduce a model without the inclusion of any network externalities. Model (1) presents the results of the regression estimation for this baseline model, which reveals the relationship between the probability of card acceptance by the merchant and control variables. The explanatory power of the model can be measured by the pseudo-R-squared, which is equal to 11.8%. It is similar to that found in the literature (e.g., Arango & Taylor, 2008a; Bounie et al., 2016; Carbу-Valverde et al., 2012; Krivosheya & Korolev, 2018). Among the significant controls found in the regression model, shop type affects probability of card acceptance: supermarkets are more likely to accept cards, whereas stalls, kiosks and micro-retailers are overall less likely to accept them. Retail network affiliation also positively influences the probability of card acceptance, which can altogether be explained by the economies of scale, higher patency of buyers and reputational issues. Overall, all of the controls are in line with the previous literature and are of expected sign (e.g., the significance of controls coincides with Krivosheya & Korolev (2018)). Since the controls have expected influence I can proceed with tests for my key hypotheses and use the sample in order to extend findings from existing literature.

In model (2) I extend the baseline model by adding the direct network externalities variable, measured by the merchant's perception of competitors' cashless payments acceptance level. Controlling for the shop type, merchant characteristics, product assortment, city size and regional characteristics, the merchant's perception of competitors acceptance is positively and significantly (p < 0.01) associated with the probability of merchant's card acceptance. Higher level of acceptance rate among the competitors perceived by merchant increases the likeliness that the merchant will start accepting cards as a method of payments (everything else being equal). The result is robust to the addition of economic crimes controls, as can be seen from model (3). From the economic point of view, if a merchant perceives that more than half of the competitors accept payment cards, the probability of merchants' acceptance increases by 35.97 percentage points (compared to the situation when less than 50% are expected to accept payment cards).

Models (4) and (5) extend these results using the alternative measures for the direct network externalities, i.e. average acceptance level for the region and federal region accordingly. Both Models (4) and (5) show that higher average acceptance level leads to higher probability of card acceptance by a merchant on both regional and federal region levels. The variables are significant at 1% significance level. From the economic point of view, one standard deviation increase in the federal region average acceptance rate increases the probability of card acceptance by 7.4 percentage points. Model (6) also controls for the economic crimes. Federal Region average acceptance remains significant.. The variable positively and significantly affects the probability of cards acceptance, though now at 5% level, possibly, because of the potential multicollinearity between shadow economy variables. When all of the regional level controls are excluded in the unreported robustness checks, the results of all the direct externalities variables remain significant and of the same sign.

Overall, models (2) - (6) test the H1 hypothesis of the presence of direct network externalities on Russian retail payments market. All of the listed above models confirm positive and significant effect of the direct network externalities. This result is also robust to changes in the composition of controls and changes in the explanatory variable estimation method. Mechanisms outlined in the theoretical framework persist on Russian market despite the role of the shadow economy and the habits of using cash. Hence, net benefits of each particular merchant increase as a result of increased activity of other merchants and, thus, lead to higher probability of cashless payments acceptance.

In order to test hypothesis H2 I proceed with testing the effect of indirect network externalities separately from the analysis outlined above. The corresponding models are presented in Table 4. Firstly, I analyze regional level indirect network effects and then proceed to the federal region level ones. Baseline model is the same as the one used in table 3. Model (7) reveals that the regional average level of cardholding positively and significantly (p < 0.05) affects the probability of cashless payments acceptance by a merchant.

Although cardholding does affect some of the mechanisms explained in the theoretical framework, card usage is observable by merchants and lead to higher convenience benefits (e.g., Krivosheya & Korolev, 2018). Hence, I add the regional level card usage variable in order to address the robustness of results. Model (8) presents the results. Regional average usage of cards positively and significantly correlates (p < 0.05) to the probability of card acceptance by a merchant.

Similarly, models (9) and (10) measure indirect network externalities effect at the Federal Regions. Although there is enough variation at the regional level, the sample was constructed in a way to represent federal regions and, thus, results using federal regions may provide better effect estimates despite some loss of variation in the explanatory variable. Model (9) exploits average holding variable and shows that it positively and significantly affects (p < 0.05) the probability of card acceptance by a merchant. In Model (10) I undertake the average usage approach and find that Federal Region average usage of cards positively and significantly affects (at 1% significance level) probability of card acceptance by a merchant. Economic significance of the result is also calculated using the marginal effects at average values of all of the variables. One standard deviation increase in average federal region usage rate of payment cards increases merchant acceptance probability by 7.04 percentage points.

Finally, I also control for the economic crimes in Model (11). The results are robust to the inclusion of these additional controls for the average size of additional fees resulting from checks and number of economic crimes in region. Besides, when all regional variables are excluded from the model to mitigate potential multicollinearity problem, the effect of indirect network externalities stays unchanged. This further highlights the robustness of the regression results.

Models (7) - (11) outlined above test the H2 hypothesis of the presence of indirect network externalities on Russian retail payments market. All of the listed models confirm the positive and significant effect of the indirect network externalities. Hence, individuals' activity at retail payments market indeed affects the net benefits of the retailers, which in turn increases the probability to accept cashless payments.

After analyzing the direct and indirect network externalities separately, I include both of the explanatory variables simultaneously to measure the total effect of the network externalities on the probability of merchant's card acceptance. Table 5 reveals the outcomes of this analysis.

In model (12) direct externalities are measured by the merchant's perceptions about his competitors' acceptance share, while indirect externalities are presented in the form of regional average usage of cards. The effect of direct network externalities remains to be significant (p < 0.01) and positively affect the dependent variable, while indirect externalities become insignificant at any reasonable significance level. When I use the federal Region average usage of cards instead of regional one in model (13), the indirect network externalities once again become significant at 5% significance level and positively affect the probability of card acceptance by the merchant. Model (13) leaves the direct network externalities measure as in the previous model changing only the measure of the indirect externalities. Direct externalities effect remains positive and significant (p < 0.01). Such a result can be explained by the sample specificity, as it is representative for federal regions rather than lower level regions. The effect of control variables remains the same as in baseline model.

In further models I change the variable proxying the direct network externalities and use the actual acceptance level instead of perceived one. First, I investigate the total network externalities effect on regional and federal region levels separately. The explanatory variables in Model (14) are: regional average acceptance level for direct externalities and regional average usage of cards for indirect externalities. Once again, the direct externalities turn out to be significant at 1% significance level, while the indirect externalities are not significant at any reasonable significance level. Similarly, the indirect network effects are insignificant in model (15) when I use the same explanatory variables at federal region level. This can be explained by the problem of multicollinearity (correlation between these variables are, respectively, 50% and 77%). Multicollinearity can be resolved by introducing the principal component analysis (PCA), which was outlined in empirical set-up. The PCA used in this study includes just two variables: usage and holding rates at regional and federal region levels. The respective components are, then, constructed. Besides, the aggregated component will allow examining the aggregate effect of network externalities, which was partially investigated in previous literature in the context of other geographic markets. Further Models (16) - (18) use PCA of Region/Federal Region network effects as explanatory variables.

Model (16) exploits one common explanatory variable to reveal the total network effects - PCA of region network effects. Leaving the control variables unchanged, the model shows that network effects positively and significantly (p < 0.01) affect the probability of card acceptance by merchants. In Model (17) I use PCA of Federal Region network effects instead, which gives similar results to the regional level PCA. To check for the robustness of the results I also add controls for the economic crimes into the later specification in Model (18). The results have not changed after the introduction of the two controls for shadow economy.

As a result, when I test the H1 and H2 together I find that the indirect externalities can appear insignificant in some frameworks. The problem that leads to such an outcome is the presence of multicollinearity. When I overcome this econometric issue by introducing of the results of the principal component analysis, the result becomes positive and significant at 1% significance level. These findings are in line with previous studies (Loke, 2007; Bounie et al., 2016) as well as with the theoretical mechanisms outlined in this study.

Overall, models (2) - (11) show support for the mechanisms identified in the theoretical framework. Hypotheses H1 and H2 of the presence of direct and indirect network externalities respectively cannot be rejected. The presence of only direct, only indirect or both network externalities at the same time positively and significantly affects the probability of card acceptance as a method of payments by any particular merchant. The results are robust to the changes in control variables.

4.3 Marginal Effects

However, it should be understood that coefficients obtained in the probit model show only the sign of the effect, but not the strength of the influence on the dependent variable. In order to investigate the magnitude of the network effects I additionally calculate the marginal effects for the last three specifications presented in the previous sub-section.

Network externalities have strong significance in economic sense as well. The results of the marginal effects analysis at the average values of independent variables for the last three models are presented in Table 6. One standard deviation increase in component of the regional network effects leads to a 10.50 percentage point increase in the probability of cards acceptance by a merchant. Similarly, one standard deviation increase in PCA of Federal Region Network effects leads to a 7.74 percentage point increase in the probability of cards acceptance by a merchant. When economic crimes are controlled for, the magnitude of the effect becomes smaller: one standard deviation increase in the factor reflecting average federal region usage rate leads to 6.7 percentage point increase in the probability of cashless payments acceptance by a merchant.

Overall, approximately 11.6% of the probability of card acceptance is explained by the network externalities effect. Thus, there are 88.4% remaining that are attributed to the social-demographic profile of the merchants, regional characteristics and other factors that may be directly affected by the CB and other market players via the obligatory changes in fee levels, acceptance subsidies, mandatory acceptance policies, taxes regulations and other stimulating mechanisms (e.g., Krivosheya et al. (2015) provides the list of stimulating measures for cashless economy development).The above analysis of the marginal effects confirms the economic significance of the network externalities. H1 and H2 cannot be rejected, as the results are not only statistically, but also economically significant.



Conclusion

This research examines the role of network externalities in card acceptance by merchants on the retail payments market in Russia. The main finding of this study is that the probability of cards acceptance by merchants increases with the presence of direct and indirect or both types of network externalities, controlling for a large set of control variables, including merchants' characteristics and other location-specific differences between retailers. The results are robust to the changes in measure of network externality. The effect persists when regional level explanatory variables are used instead of the federal region ones and after the introduction of controls for economic crimes. The findings are significant both statistically and economically. From the practical point of view, understanding the magnitude of influence of the network externalities might explain the extent to which the government, commercial banks, payment systems and other market participants can influence the probability of card acceptance by merchants.

The article contributes to the small but rising literature on the determinants of card acceptance demand by merchants (C. Arango & Taylor, 2008a; Bounie et al., 2016; Carbу-Valverde et al., 2012; Hayashi, 2006; Krivosheya & Korolev, 2018; Loke, 2007; Rochet & Tirole, 2011). The role of the network externalities have been established in the theoretical studies and have often been hypothesized to influence the cashless payments usage and acceptance, but there is a lack of empirical studies evaluating the magnitude of the network effects at the retail payments market. Moreover, to the best of my knowledge, none of the studies separate between the direct and indirect the network effects. Also, there is a lack of empirical studies regarding cashless payments acceptance on Russian market, where the role of cash has historically been high and the end-users behavior habits are yet forming. This article fills these gaps by providing the empirical analysis based on the survey of 800 traditional (offline) merchants from all Russian regions, and shows estimates of the effect of both direct and indirect network externalities for the merchants' card acceptance probability at Russian retail payments market.

This research complements recent empirical studies by Bounie et al. (2016) and Arango-Arango et al. (2018), which focuses on the role of network externalities at the retail payments market. The former study focuses on the merchant side of the retail payments card market and explains the card acceptance probability in France. Previous studies, however, could not efficiently separate the effect of direct network externalities from the indirect ones due to data limitations (Bounie et al. (2016) have different time periods for individuals and merchants samples). Apart from other gaps, this study fills this gap by investigating the indirect network externalities and the total network effect alongside the direct network effects in the context of Russian market. I also implement more control variables to avoid the potential omitted variable bias following the set of control variables established by Krivosheya and Korolev (2018) for the Russian retail payments market that arise not only from merchant but also from the geographical and economic specificities.

For each type of the network externalities I show evidence for both regional and federal region level proxies. Empirically, all of the introduced explanatory variables appear to be statistically significant, thus, increasing the probability of cards acceptance by merchants. Due to the fact that (federal) regional usage and acceptance levels might correlate I also introduce the principal component of different network effects variables to analyze the aggregate effect of both network externalities on the card acceptance probability. The PCA variables for the total network effects provide the same results at both Regional and Federal Region levels in Russia. These results are in line with the findings of previous literature that investigated the aggregate network effects influence on card acceptance probability at the developed retail payments markets (Bounie et al., 2016; Carbу-Valverde et al., 2012).

All of the results are robust to changes in measures and are also economically significant. One standard deviation increase in average federal region card acceptance increases the probability of acceptance by each particular merchant by 7.4 percentage points. Indirect externalities have similar effect: a standard deviation increase in average federal region usage rate of payment cards increases merchant acceptance probability by 7.04 percentage points. Combined, one standard deviation increase in the PCA factor reflecting both network externalities at the federal region level increases the merchant acceptance probability by 7.74 percentage points. In comparison, additional year of operations contribute to less than 1 percentage point increase in merchant acceptance probability.

From the practical point of view the results of this analysis unveil the extent to which different stimulating measures can affect merchants' card acceptance probability. Network externalities can be perceived as a multiplier for the policies that are aimed at the retail payments market stimulation. The magnitude of the effect of the network externalities reflects the degree towards which an increase in payment activity of cardholders and other merchants influences the acceptance rates by merchants. Hence, any stimulating measure is able to influence the payments market in two ways: directly influencing the acceptance or usage of payment services and indirectly influencing the merchant acceptance via the network externalities. The effect of network externalities cannot be changed immediately by any existing stimulating measures. Therefore, the magnitude of the effect of the network externalities shows the share of the merchants' demand that cannot be altered by any financial market policies implemented by market participants such as the commercial banks, payment systems and central banks.

As any other analysis, this study has certain limitations that can be used to set up the directions for further research. First of all, the sample used in this study does not account for the online merchants. Online retailing market increased actively during the past decades both in Russia and in the world. Merchants have more incentives to accept cards as a method of payments in digital space because online markets have specific nature of competition, which is not dependent on the merchant's physical location, and which is usually more intense compared to offline retail (Au & Kauffman, 2008; Krivosheya & Korolev, 2018). It can, thus, be expected that the effect of the direct network externalities will intensify due to the increased competition. Besides, the increasing stimulating measures by bank-issuers, such as cash-back and discounts incentivize consumers to prefer online card payment rather than cash-on-delivery. Thereby, the inclusion of online-merchants may as well intensify the indirect network effects. Secondly, the sample used includes the data on 2013-2014 period only. Although the correlations between the network effects and the acceptance probability are unlikely to change, it would be interesting to see how the effects of new regulation and technologies have altered the degree of influence of network externalities in Russia. Finally, global retail payments markets can be added to the analysis to assess the differences in the extent to which network externalities affect developed, developing and underdeveloped countries. Besides, international comparison allows including more insights on the cross-border payments, which may unveil different mechanisms behind network effects because of lower degree of communication between foreign merchants and individuals.



References

Национальное агентство финансовых исследований (НАФИ) (2014), Барьеры -- в голове. Российская Бизнес-газета - №951 (22) [Online].

Национальное агентство финансовых исследований (НАФИ) (2017), Доверие банкам, страховым компаниям и НПФ заметно снизилось. Аналитика от 8 сентября 2017. [Online].

Центральный Банк Российской Федерации (2014), Платежные и расчетные системы. Анализ и статистика», выпуск 42. [Online].

Центральный Банк Российской Федерации (2016), Платежные и расчетные системы. Анализ и статистика», выпуск 50. [Online].

Ali, R., Barrdear, J., Clews, R., & Southgate, J. (2014). Innovations in Payment Technologies and the Emergence of Digital Currencies. Bank of England Quarterly Bulletin 2014 Q3.

Arango, C., Huynh, K. P., & Sabetti, L. (2015), Consumer payment choice: Merchant card acceptance versus pricing incentives. Journal of Banking & Finance, 55, 130-141.

Arango, C. A., Huynh, K. P., & Sabetti, L. (2011). How Do You Pay? The Role of Incentives at the Point-of-Sale. ECB Working Paper No. 1386.

Arango, C., & Taylor, V. (2008). Merchants' Costs of Accepting Means of Payment: Is Cash the Least Costly? Bank of Canada Review, 2008-2009(Winter), pp. 17-25.

Arango-Arango, C. A., Bouhdaoui, Y., Bounie, D., Eschelbach, M., & Hernandez, L. (2018). Cash remains top-of-wallet! International evidence from payment diaries. Economic Modelling, Volume 69, January 2018, pp. 38-48.

Armstrong, M. (2005). Competition in Two-Sided Markets. The RAND Journal of Economics, Volume37, Issue3, September 2006, pp. 668-691.

Au, Y. A., & Kauffman, R. J. (2008). The economics of mobile payments: Understanding stakeholder issues for an emerging financial technology application. Electronic Commerce Research and Applications, 7(2), pp. 141-164.

Bagnall, J., Bounie, D., Huynh, K. P., Kosse, A., Schmidt, T., Schuh, S. D., & Stix, H. (2014), Consumer cash usage: A cross-country comparison with payment diary survey data. International Journal of Central Banking, 2016, vol. 12, issue 4, 1-61.

Baxter, W. F. (1983). Bank Interchange of Transactional Paper: Legal and Economic Perspectives. The Journal of Law & Economics, 26(3), pp. 541-588.

Bayero, M. A. (2015). Effects of Cashless Economy Policy on Financial Inclusion in Nigeria: An Exploratory Study. Procedia - Social and Behavioral Sciences, 172, pp. 49-56.

Bedre-Defolie, Ozlem, & Calvano, E. (2013). Pricing Payment Cards. American Economic Journal: Microeconomics, 5(3), pp. 206-231.

Block, F. L., & Keller, M. R. (2015). State of Innovation: The U.S. Government's Role in Technology Development. Routledge.

Bolt, W. (2012). Retail Payment Systems: Competition, Innovation, and Implications (DNB Working Paper No. 362). Netherlands Central Bank, Research Department. De Nederlandsche Bank Working Paper No. 362.

Bolt, W., & Chakravorti, S. (2008). Economics of payment cards: a status report. Economic Perspectives, (Q4/2008), pp. 15-27.

Bolt, W., & Mester, L. J. (2017). Introduction to Retail Payments: Mapping Out the Road Ahead. Journal of Financial Services Research, 52(1-2), pp. 1-3.

Bolton, R. N., Kannan, P. K., & Bramlett, M. D. (2000). Implications of Loyalty Program Membership and Service Experiences for Customer Retention and Value. Journal of the Academy of Marketing Science, 28(1), pp. 95-108.

Bounie D., Franзois A., & Van Hove L. (2015), Consumer payment preferences, network externalities and merchant card acceptance: An empirical investigation. Review of Industrial Organization, 51(3), 257-290.

Bounie D., Francois A., & Van Hove L. (2016), Merchant acceptance of payment cards:`must take' or `wanna take'?. Review of Network Economics, Vol. 15, No. 3, p. 117-146.

Carbу-Valverde, S., & Liсares-Zegarra, J. M. (2011). How effective are rewards programs in promoting payment card usage? Empirical evidence. Journal of Banking & Finance, 35(12), pp. 3275-3291.

Carbo-Valverde, S. и Linares-Zegarra, J. (2012), Payment Card Interchange Fees: Assessing the Effectiveness of Antitrust Investigations и Regulation in Europe. Available at SSRN 2052056.

Carbу- Valverde, S., Liсares Zegarra, J., Rodrнguez Fernбndez, F., (2012), Feedback Loop Effects in Payment Card Markets: Empirical Evidence, Review of Network Economics, vol. 11, Issue 2, article 2.

Chakravorti, S., & Roson, R. (2006). Platform competition in two-sided markets: The case of payment networks. Review of Network Economics, 5(1).

Chernikova, L. I., Faizova, G. R., Egorova, E. N., & Kozhevnikova, N. V. (2015). Functioning and Development of Retail Banking in Russia. Mediterranean Journal of Social Sciences, Vol. 6. No. 6., pp. 274-284. .

Chiang, T. C., & Zheng, D. (2010). An empirical analysis of herd behavior in global stock markets. Journal of Banking & Finance, 34(8), pp. 1911-1921.

Ching, A. T., & Hayashi, F. (2010). Payment card rewards programs and consumer payment choice. Journal of Banking & Finance, 34(8), pp. 1773-1787.

Chizhikova, E. S. (2013). The Current Payment System of the Russian Federation. Middle-East Journal of Scientific Research, 14(2), pp. 244-247.

Darban, M., & Amirkhiz, H. (2015). Herd Behavior in Technology Adoption: The Role of Adopter and Adopted Characteristics. In 2015, 48th Hawaii International Conference on System Sciences (pp. 3591-3600).

Doyle M. A., Fisher, C., & Yadav, A. (2017), How Australians Pay: Evidence from the 2016 Consumer Payments Survey. Reserve Bank of Australia.

Economides N. (1996), Network Externalities, Complementarities, and Invitations to Enter. European Journal of Political Economy, Vol. 12(2), pp. 211-233.

Evans, D., and Schmalensee, R. (1999), Paying with Plastic: The Digital Revolution in Buying and Borrowing. MIT Press: Cambridge, MA.

Guthrie, G., & Wright, J. (2007). Competing Payment Schemes. The Journal of Industrial Economics, 55(1), pp. 37-67.

Hasan, I., Schmiedel, H., & Song, L. (2012). Returns to Retail Banking and Payments. Journal of Financial Services Research, 41(3), pp. 163-195.

Hayashi, F. (2006). A Puzzle of Card Payment Pricing: Why Are Merchants Still Accepting Card Payments??Review of Network Economics, Vol. 5, Issue 1.

Hunt, R. M. (2003). An Introduction to the Economics of Payment Card Networks. Review of Network Economics, Vol. 2(2).

Jonker, N., (2011), Card Acceptance и Surcharging: the Role of Costs и Competition, Review of Network Economics, 10 (2).Katz M. L., Shapiro C. (1985), Network Externalities, Competition, and Compatibility. The American Economic Review, Vol. 75, No. 3, pp. 424-440.

Kabakova, O., Plaksenkov, E., & Korovkin, V. (2016). Strategizing for Financial Technology Platforms: Findings from Four Russian Case Studies. Psychology & Marketing, 33(12), pp. 1106-1111.

Kim, C., Tao, W., Shin, N., & Kim, K.-S. (2010). An empirical study of customers' perceptions of security and trust in e-payment systems. Electronic Commerce Research and Applications, 9(1), pp. 84-95.

Krivosheya, E., & Korolev, A. (2016). Benefits of the retail payments card market: Russian cardholders' evidence. Journal of Business Research, 69(11), pp. 5034-5039.

Krivosheya, E., & Korolev, A. (2018). Benefits of the retail payments card market: Evidence from Russian merchants. Journal of Business Research, 88, pp. 466-473.

Krivosheya, E., Korolev, A., & Plaksenkov, E. (2015). Measures for stimulating a cashless economy in Russia. Moscow School of Management SKOLKOVO working paper.

Plaksenkov, E., Korovkin, V., & Krivosheya, E. (2015). Cashless economy in Russia: Tendencies, perspectives, opportunities. Moscow School of Management SKOLKOVO Working Paper.

Krivosheya, E., Semerikova, E., Korolev, A. & Tarusova, E. (2017). Cashless economy in Russia 2030: Scenarios for market and industry. Moscow School of Management SKOLKOVO Working Paper.

Loke, Y. J., (2007), Determinants of Merchant Participation in Credit Card Payment Schemes, Review of Network Economics, 6 (4).

Malphrus, S. (2009). Perspectives on Retail Payments Fraud. Economic Perspectives, Vol. XXXIII, No. 1, 2009.

Milne, A. (2006). What is in it for us? Network effects and bank payment innovation. Journal of Banking & Finance, 30(6), pp. 1613-1630.

Promothesh C. and Randall L (2012), Do Payment Mechanisms Change the Way Consumers Perceive Products? Journal of Consumer Research, 38(6).

Rauch, D. E., & Schleicher, D. (2015). Like Uber, But for Local Governmental Policy: The Future of Local Regulation of the “Sharing Economy”. George Mason Law & Economics Research Paper No. 15-01.

Reinartz, W., Dellaert, B., Krafft, M., Kumar, V., & Varadarajan, R. (2011). Retailing Innovations in a Globalizing Retail Market Environment. Journal of Retailing, Vol. 87, Supplement 1, pp. S53-S66.

Rochet, J.-C., & Tirole, J. (2002). Cooperation among Competitors: Some Economics of Payment Card Associations. The RAND Journal of Economics, 33(4), pp. 549-570.

Rochet, J.-C., & Tirole, J. (2003). Platform Competition in Two-sided Markets. Journal of the European Economic Association, 1(4), pp. 990-1029.

Rochet, J.-C., & Tirole, J. (2011). Must-Take Cards: Merchant Discounts and Avoided Costs. Journal of the European Economic Association, 9(3), pp. 462-495.

Rysman, M. (2007), An Empirical Analysis of Payment Card Usage. The Journal of Industrial Economics, 55 (1): 1-36.

Scharfstein, D. S., & Stein, J. C. (1990). Herd Behavior and Investment. The American Economic Review, 80(3), pp. 465-479.

Todd, S., & Lawson, R. (2003). Consumer preferences for payment methods: a segmentation analysis. International Journal of Bank Marketing, 21(2), pp. 72-79.

Trueman, B. (1994). Analyst Forecasts and Herding Behavior. Review of Financial Studies, 7(1), pp. 97-124.

Vickers, J. (2005), Public Policy and the Invisible Price: Competition Law, Regulation and the Interchange Fee, Proceedings of a conference on “Interchange Fees in Credit and Debit Card Industries” (Federal Reserve Bank of Kansas-City, Santa Fe, New Mexico, May 4-6, 2005) pp. 231-247.

Visa USA Research Services (2006), Visa Payment Panel Study: Payment Trends Summary.

Wakamori N., Welte A. (2017), Why Do Shoppers Use Cash? Evidence from Shopping Diary Data. Bank of Canada.

Weiner, S., & Wright, J. (2005). Interchange Fees in Various Countries: Developments and Determinants. Review of Network Economics, 4(4), pp. 1-34.

Weitzel T., Wendt O., Falk G. von Westarp and Konig W. (2003), Network Effects and Diffusion Theory: Network Analysis in Economics. International Journal of IT Standards and Standardization Research 1(2), pp. 1-21

Wright, J. (2004). The Determinants of Optimal Interchange Fees in Payment Systems. The Journal of Industrial Economics, 52(1), pp. 1-26.

Annexes

Graph1. Operations performed in Russia with the use of payment cards.

Source: CBR

Graph 2: Share of the two types of operations performed in Russia with the use of payment cards.

Source: CBR

Table 1. Descriptive statistics

Source: Author's computations.

Table 2 Cross-correlations of the variables.

Source: Author's computations.

Table 3. Probit model results. Baseline model and implementation of Direct Externalities.

Source: Author's computations.



Table 4. Probit model results. Implementation of Indirect Externalities.

Source: Author's computations.



Table 5. Probit model results. Implementation of both Direct and Indirect Externalities.

Source: Author's computations.



Table 6. Marginal effects.

Source: Author's computations.

Appendix

payments market card acceptance

Review of the studies investigating the network effects at merchants' side of the retail payments market

World current non-cash retail payment market analysis

In addition to similar to the above-mentioned studies built on the public surveys and analysis of the open data, there is a new direction of academic research in the world practice dedicated to a deeper qualitative analysis of consumer behavior and trade points in the market of non-cash retail payment services with the help of diary entries (eg., Arango et. al., 2015; Doyle et. al., 2017; Wakamori and Welte, 2017). This method of data collection allows to study the information behind each individual transaction in more detail and look at the choice of payment method based on different situations. Most of these studies can explain the dominance of cash with a small amount of transaction.

The payment card market is a two-sided market and is characterized by the presence of positive externalities between the two groups of consumers - merchants and cardholders. An extensive research literature (Economides N., 1997; Katz M. L. and Shapiro C.; Weitzel T. et. al., 1994) indicates that the dynamics of the network effect is nonlinear, there are two thresholds in it. The first one is the moment of the emergence of the large-scale effect (the gain of "critical mass"), up to which the network is too weak for the participants to enter into an effective interaction, and the second one is the moment of saturation, after which the further development of the network does not bring significant new benefits for the participants. From the point of view of the dynamics of the innovation spread, it is extremely important to quickly overcome the first threshold. The digitalization of payments is important for all spheres of human life. For this reason, the network effect is likely to begin to reveal itself at a time when the level of digitalization is high enough across the wide field of possible transactions of the average person. In order for this to happen, in turn, a combination of two factors is necessary: the availability of opportunities - digital supply from merchants and the implementation of this opportunity, the presence of demand (desire and skills) from the buyers.

The development of digital payments largely takes place due to the general process of penetration of digital technologies in various spheres of daily life. At the same time, The speed of digitalization of society, as a special case of the process of innovation growth, is associated with the disclosure of a positive network effect: with an increase in the share of digital transactions, the relative benefit for the parties to the transaction is increased and, consequently, the probability of further increase in digital density increases. The more outlets accept cards of this payment system, the higher interest of buyers to keep these cards, and the more buyers choose cards of this payment system in balance (with other things being equal, for example, at the same rates). The opposite is true - the more customers hold cards of this payment system and intend to use these cards for payment, the higher the interest of trade enterprises to accept cards of this payment system, and the more outlets accept these cards in balance (Arango, Taylor (2008); Loke (2007); Bounie, Francois, van Hove (2015)) . The described effect is known as indirect network externality on the retail payments market.

A number of empirical works test the hypothesis of indirect network effects and assess how much demand on one side of the market (for example, the demand for cards of this system by the holders) depends on the number of participants in the payment market on the other side (for example, the number of outlets that accept cards of this payment system). One of the first papers on this topic was presented by Rysman (Rysman, 2007). The author confirms the hypothesis about the correlation between the use of cards by consumers and the acceptance of cards by commercial enterprises within the four main payment systems (MasterCard, Visa, American Express and Discover). Regional data from 1998 to 2001 are used to test the hypothesis. Part of this data was obtained from the Payment systems Panel Study conducted regularly by VISA, and containing data on cardholders made during the month of payment transactions, including the amount of the transaction, the name of the outlet, the index, the type of payment instrument, the name of the payment system, etc. Data of VISA system for all trading enterprises on the number and amount of transactions conducted under the VISA system and three other major systems were also used. A statistically significant and positive correlation between the number of outlets accepting cards of the payment system and the number of buyers using cards of this system for payment was found. The obtained estimates can be used to assess the influence of the number of merchants accepting cards of a particular system on the probability of the buyer choosing the card of this system for payment.

In the work of Carbo-Valverde et al. (Carbo-Valverde et al., 2012a) the hypothesis of the presence of externalities between cardholders and merchants who accept cards for payment is also tested. The test is performed on the data from the two largest payment systems in Spain for the period from 1997 to 2007. In contrast to the previous work, which used a regional data, Carbo-Valverde et al. used bank data for testing, which included not only data on the number of cards, the number of POS-terminals, the number of ATMs, but also price characteristics - the size of the trade concessions, the average tariffs for annual service. Thus, the authors of the article were able to assess the demand for payment cards from payers and outlets as a function of tariffs and the number of participants of the system on the other side of the market. The impact of indirect network effects is statistically significant and positive; the impact of tariff rates is also significant but negative. For example, according to the results the increase of 1% in the number of cardholders increases the number of merchants willing to accept cards of this system by 2.7%, and the increase of 1% in the size of the trade concession reduces the demand from outlets by 6.4%.

In recent years, the number of works that note the presence of network effects increases. So, Bounie et al (2017) redefine network effects through the probability that the average shopping cart will be paid by a non-cash method. In another paper by Bounie et al (2017), the authors separate a direct network effect into the strategic decision associated with the fact that the outlets are trying to compete with each other and an obligation to accept cards due to regulatory issues. The work demonstrates that strategic motivation is much more important than the obligation to accept cards due to regulation (approximately 94% to 80% of retailers accept cards of their own free will). Hence, the existing market competition enhances merchants to serve a payment system that has already been adopted by the peers in order not to lose consumers, and market share as a consequence (Jonker (2011); Bounie, Francois, van Hove (2015)). Thus, the second hypothesis that I would like to test is the presence of direct network externalities on the retail cashless payments market in Russia, or the network effect which relates only to the merchant-side of the payments market. The hypothesis is:

Among other card acceptance determinants, previous researchers outline the customers' characteristics, such as age and level of income. The features of the clientele may indirectly influence merchant's propensity to accept cards. Consumer studies such as Bagnall et al. (2014) reveal that card usage increases with income and age. Hence, a merchant who, say, has many high-income clients might want to accommodate their desire to pay by card.

Loke (2007) outlines that the differences in the age of the merchants (in years) may also lead to differences towards the acceptance of cards. As cards acceptance can be considered as a relatively new form of payment technology, older merchants are expected to be more likely to resist accepting cards from credit card payment schemes than the younger ones. This follows from Zmud (1979) and Assael (1981) who found that older people are expected to be more resistant towards new technology.

Another significant factor outlined by Jonker (2011) and Bounie, Francois, van Hove (2015) in their works is the size of merchant discount. It is asserted that merchants accept to pay the resulting from unregulated interchange fees high merchant discounts because they are concerned that turning down cards would impair their ability to attract customers; that is, cards are “must-take cards” (Vickers, 2005).Bounie, Francois, van Hove (2015) find that on the costs side, card fraud, which is measured as an annual value of fraudulent card payments over the total value of card payments, was insignificant to discourage acceptance. The authors also controlled for the tax regime of the merchant, the reason being that of the three main taxation regimes for businesses in France, the “rйgime micro-enterprise” is substantially more evasion prone. The taxation is applied for very small enterprises with sales revenue below certain thresholds. The taxation for such merchants is based on self-reported revenue, thus the merchants who fall under the micro-enterprise regime have more room for tax evasion and will, if they engage in such activities, prefer not to accept cards in order to hide their real sales volume. Thus, the taxation regime of the merchant, which is meant to proxy the room for tax evasion, does have a significant impact on the probability of card acceptance: authors reveal that the microfirm regime exerts a negative effect. Overall, I expect the impact of both direct and indirect network effects to be statistically significant and positive. There are reasonable doubts that the empirical results will be in accordance with the previous researches, as those were conducted for the developed countries (such as Canada, Spain, USA, etc.) with high level of financial literacy and trust to financial institutions and innovations. It can appear that the developing countries have a different approach to the new methods of payment acceptance from either or both sides of the payment market. Furthermore, the impact of the effects can turn out to be not significant or have noticeably lower influence on the probability of the card acceptance than that in the developed countries.

...