File No. CT 2005-006 COMPETITION TRIBUNAL IN THE MATTER OF the Competition Act, R.S.C. 1985, c. C-34, as amended;
lN THE MAT IER. OF an application by B-Filer Inc, B. Filer Inc. doing business as GPAY GuaranteedPayment and NPay Jnc. for an order pursuant to section l 03.1 granting leave to make application underr---CO_M_P_E_i_ITI_O_N-TR_IB_U_N_A_l_ _- .
sections 75 and i7 of the Competition Act; AND IN THE MAT IER OF an application by B-Filer Inc., B-Fi ~Inc. doing business as GPA Y Gua.ranteedPayment and NPay Inc. for interim order pursuant to section I 04 of the Competition Att. BETWEEN: B-FILERINC., B-FILER INC. doing business as .. GPAY GUARANTEEDPAYMENT and NPAY INC.
Applicants -and-THE BANK OF NOVA SCOTIA Respondent REPLY AFFIDAVIT OF FRANK MATHEWSON I, FRANK MATHEWSON, oft he City oft he Toronto, in the Province ofO ntario, MAKE OATH AND SAY: 1. I am a Professor of Economics and Director of the Institute for Policy Analysis at the University of Toronto. 2. I have been asked by counsel to the Respondent, The Bank of Nova Scotia, to provide an expert opinion replying to the validity oft he economic assumptions relied upon the Applicants and by their economist expert. Larry Schwartz ·with respect tot among other things, the alleged effect on the Applicants' business due to the termination by Scotiabank.
TR!BUNAL DE LA CONCURRENa P ~ L D E U D I T
-2-3. I attach my report setting out my opinion on these issues as Exhibit ''A'~ to this my Atl'idavit FRANK MATHEWSON
Schedule "A" File No. CT 2005-006 COMPETITION TRIBUNAL
IN THE MATTER OF the Competition Act, R.S.C. 1985, c. C-34, as amended; IN THE MATTER OF an application by B-Filer Inc., B-Filer Inc. doing business as GPAY GuaranteedPayment and Npay Inc. for an order pursuant to section 75;
BETWEEN: B·FILER INC., B·FILER INC. doing business as GPAY GUARANTEEDPAYMENT and NPAY INC.
Applicants -and-THE BANK OF NOVA SCOTIA Respondent REPLY REPORT OF FRANK MATHEWSON A. INTRODUCTION AND SUMMARY 1. This Report provides my comments on the statistical analysis submitted by the Applicants in this proceeding.1 Section B describes the data contained in the documents submitted by Applicants. Section C provides my comments on Applicants' projections, which are summarized as follows:
a) Applicants' estimates of the value of GPay payments from Scotiabank but for the termination of GPay's banking services by Scotiabank are unreliable and uninformative. Applicants assume that growth rates in GPay payments from Bank of Montreal ("BMO") and Royal Bank of Canada ("RBC") accounts during the post-termination period are appropriate proxies· for the growth rate in Scotiabank payments. This assumption and these proxies are faulty;
1 The Applicants' data and, analysis are contained in the documents BFI001644-BFI001670.
b) There is not enough information available in these documents submitted by Applicants to estimate properly any potential losses in GPay payments because of the termination. To estimate properly any potential losses in payments, at a minimum, data on the characteristics of the users of the GPay service, including whether they made GPay payments from accounts at two or more banks, would be required. Data on the characteristics of account holders at each bank, including whether they hold accounts at more than one bank would be valuable. On the supply side, information on GPay's costs of facilitating payments from accounts at banks where GPay has bill payee status and at banks where it does not have such status would be required. Information would also be needed on the characteristics, including prices, of the financial services products offered at each of the financial institutions;
c) Applicants ignore the growth of payments from accounts held at CIBC and TD Bank ("TD"), the two other major banks where GPay does not have bill payee status, when making its projections. Total GPay payments from these banks grew at a faster rate than total payments at 'bill payee' banks, indicating t~at lack of bill payee status alone is not an impediment to growth by GPay, and;
d) Applicants assume that reductions in GPay payments from Scotiabank accounts (relative to what GPay payments would have been but for termination) are net losses to G Pay with no substitution by customers to payments from accounts at other banks. There is evidence that this is not the case. Some customers with an account at both a 'biller services bank' and an 'EMT bank' make GPay payments from both accounts, suggesting that the EMT limits on GPay payments at EMT banks need not have a large negative effect on the total value of GPay payments.
B. DATA PROVIDED BY APPLICANTS 2. The Applicants' analysis is based on the following data:
• The dollar value of payments processed by GPay from accounts at a number of banks, during each month between September 2003 and June 2006, inclusive.2 These data are reported for Scotiabank, BMO, CIBC, RBC, TD, and Caisse Populaire; and
• The 'value' of payments for three categories: "Bank Biller'', "EMT", and "BNS". 3 Applicants do not explain which banks are included in each of these categories, nor do they explain how the reported values were calculated, but the "Bank Biller" values appear to be simple averages of the value of GPay payments from BMO and RBC accounts; the "EMT" values appear to be simple averages of the value of GPay payments from accounts at CIBC, TD, and BNS; and "BNS" values are the payments from Scotiabank accounts.4
3. A number of other charts and tables are provi<;led. Some charts are graphical representations of the data discussed above.
C. COMMENTS ON APPLICANTS' PROJECTIONS I. Projections of Payment Values 4. There is no reason to expect that the growth in GPay payments at other banks with bill payee status would be a good predictor of the growth in GPay payments from Scotia bank accounts had GPay retained bill payee status at Scotiabank. The month-to-month growth in GPay payments from accounts at Scotiabank was historically not very closely related to the growth in payments from either BMO or RBC accounts. This is true even in the pre-termination period (September 2003 to September 2005), when GPay had bill payee status at all three banks.
2 BFI001645 and BFI001646, "Table: Value of Payments by Bank by Month, 2003-2006". 3 These data are reported In two tables. BFI001650, "Table: Value of Payments-Billerv EMT Banks by Month, Oct2005-June 2006" provides monthly data for "Bank Biller", "EMT", and "BNS" for October 2005 to June 2006. BFI001652 and BFI001653, "Table: Value of Payments- Billers v BNS by month, 2003-June 2006" provides monthly data for "Bank Biller" and "BNS" for the period September 2003 to June 2006.
4 BFI001650, "Table: Value of Payments - Biller v EMT Banks by Month, Oct 2005- June 2006".
3
5. For example, the cumulative monthly growth rate5 in GPay payments from Scotiabank accounts between September 2004 and September 2005 was 7.3%. The growth rate for GPay payments from BMO accounts was 8.0% during this period, while the growth rate for payments from RBC accounts was 11.5%. The growth rate for payments from RBC accounts over this period was 57% higher than the growth rate for payments from Scotiabank accounts, an indication that the RBC growth rate is an inappropriate predictor of the Scotiabank growth rate.
6. To further illustrate this point that the growth rates of payments from BMO and RBC accounts are unreliable predictors of the growth in payments from Scotiabank accounts, I applied the method used by GPay to 'predict' Scotiabank revenues for the period August 2004 to August 2005, the year immediately prior to Scotiabank's termination of GPay. That is, I assumed that, starting from the base month August 2004, GPay payments from Scotiabank accounts grew at the same rate as each of 1) the simple average of BMO and RBC payments; and 2) the three month moving average of payments from BMO and RBC accounts.a The payments from Scotiabank accounts in August 2004 totalled $[CONFIDENTIAL); under the Applicants' methodology, the 'predicted' value of GPay payments in August 2005 is $[CONFIDENTIAL] using the monthly growth rates in the simple average of payments from BMO and RBC accounts and $[CONFIDENTIAL] using three month moving averages. This means that, similar to Applicants' approach, using these two average growth rates to predict payments from Scotiabank accounts for August 2005 would lead to an over-prediction of 99% (in the case of simple average monthly growth rates) or 55% (using the growth rate for three month moving averages).
7. It is possible to be more precise: that the growth rates in payments from other 'Bank Billers' are not good proxies for Scotiabank growth rates is evidenced from a simple regression analysis. Consider a regression model where the dependent variable is the monthly growth rate of payments from Scotiabank accounts from January 2004 to
5 The cumulative annual growth rate ("CAGR") is a common method of calculating the average year-to-year growth rate of a variable over time. The cumulative monthly growth rate is analogous and is used to calculate the average month-to-month growth rate over lime.
6 Applicants' appeared to use both of theses averages to project growth rates for GPay payments from Scoliabank accounts. 4
August 2005, and the independent variable is the growth rate in payments from RBC accounts: the estimated coefficient is -0.16 which is statistically insignificant at the 95% confidence level. In the same model where the independent variable is payments from BMO accounts, the estimated coefficient is -0.27 which is also statistically insignificant at the 95% confidence level. Regression coefficients in this simple linear model are estimates of the relationship between two variables. For example, if the estimate of the coefficient is 'a' and the estimate is statistically significant, we would say that for every percentage point increase in the growth rate in payments from BMO accounts (or RBC accounts respectively) we would expect that payments from Scotiabank accounts would change by 'a' percentage points. However, these data reveal that the estimated coefficients are statistically insignificant, which implies that growth in payments from BMO or RBC accounts are associated with zero changes in payments from Scotiabank accounts. From these data there is no convincing evidence that these growth rates are correlated with each other. More complete and well-specified regression models may indicate otherwise but the data to pursue such models is unavailable. The available evidence indicates that the growth rates in payments from BMO and RBC accounts are not likely to be good predictors of the growth rate in payments from Scotiabank accounts.
High growth rates at other banks where GPay did not have bill payee status 8. GPay payments from accounts at CIBC and TD, the other two banks where GPay did not have bill payee status after September 2005, also grew significantly more than payments from Scotiabank accounts between September 2005 and June 2006. The simple average of monthly CIBC and TD payments grew at an average rate of 8% over this period, which was not only higher than the rate of growth of payments from Scotia bank accounts, but it was also higher than the growth in payments from accounts at banks where GPay had bill payee status, which was about 6%. According to GPay's analysis, the sole cause of the reduction in payments from Scotiabank accounts from September 2005 was GPay's loss of bill payee status at BNS. But if the demand for GPay payments depends on whether GPay has bill payee status atthe account holder's bank, then GPay's analysis would imply that we would see lower growth rates for payments from CIBC and TD accounts. However, we do not.
5
Reductions in payments from Scotiabank accounts are not net reductions in GPay payments
9. Some people who hold accounts at 'EMT' banks also hold accounts at 'Biller Services' banks, and could switch to making payments from accounts at biller services banks in response to a restriction on payments from EMT banks. 1o . We know from Applicants' data that some GPay customers can, and do, have accounts at two or more banks from which they make GPay payments. A GPay document7 reports the top twenty GPay customers in May 2006. There are at least two customers who have accounts at an EMT bank and at a Biller Services bank, and both of these customers made GPay payments from both accounts in May 2006. The customer given the nickname "B" made payments from a Scotiabank account (an EMT bank at that time), and also made twenty two payments from an RBC account, for a total value of payments from his/her RBC account of $65,815.00. The customer given the nickname "F" made sixty eight payments for a total value of $31,550.47 from an RSC account, and also made GPay payments from an account at TD. These two GPay customers accounted for over 6% of all RBC payments in May 2006. This is a lower bound on the percentage of the value of GPay payments made from RBC accounts by customers with accounts at two or more banks, because some other, unknown to us, RBC account holders may hold multiple accounts but make payments during the sample period only from their RBC account. This suggests that some GPay customers have the capacity to switch between accounts at different banks when limits are imposed on the value of . payments from one of their accounts.
11. Data provided by Scotiabank reports GPay payments from Scotiabank accounts with value exceeding the EMT limit of $1,000 for September 2006. These data show that just one customer accounted for [CONFIDENTIAL]% of GPay payments from Scotiabank accounts in September 2005 (this person made payments of $[CONFIDENTIAL], while the total value of GPay payments from Scotiabank accounts was $[CONFIDENTIAL]). It seems possible that this customer, and others like him/her,
7 Applicants' document BFI001103, "Top 20 Customers by Total Paid, May 2006." 6
would have used existing accounts at other Biller Services banks or opened an account at a Biller Services bank rather than foregoing the use of GPay altogether, especially given that the cost of opening a new bank account at a bank is low relative to the payment rriade by this one customer.
12. This information suggests that at least some of the reduction in GPay payments from Scotiabank accounts did not represent a net reduction in total GPay payments. Rather, some GPay payments formerly from Scotiabank accounts may post termination have been switched to accounts at banks where GPay still has bill payee status, such as BMOand RBC.
Additional data from GPay is required to project 'but for' payments properly 13. A more general point is that proper estimation of GPay payments from Scotiabank accounts but for the termination requires a more complete model with additional data and specification, for example the characteristics of the clients of all the banks and the variations in accounts offerings and their respective prices. This information would have to include data relating to bank client holdings of accounts in multiple banks, and also data relating to the extent to which payer clients of GPay make payments from accounts at two or more banks, as discussed above. Supply side information about GPay's costs of providing its service from accounts at banks where it has bill payee status and at banks where it does not have bill payee status would also be required. These data are not currently available and a more complete model is not specified. Implicit in any simple growth rate analysis, such as that conducted by Applicants, is the assumption that in all other respects all relevant variables (including characteristics of bank customers) are alike. Suppose, for example, that clients are differe.nt across banks, and in particular that they are different with respect to their propensity to make payments to GPay merchants, who primarily operate online, offshore gambling sites. Then, given the relatively small number of GPay clients, a limited number of 'high rollers' can skew the results: differences in the incidence of high rollers across the banks would therefore affect differences in GPay payments among banks. For example, these consumer differences could flow from differences in regional demands for gambling, if banks have different market shares in different part of Canada.
7
BETWEEN: B-FILER lNC. - and- THE BANK OF NOVA SCOTIA Applicants
McCarthy Tetrault IJI. . P Barristers & Solicitors Box 48, Suite 4700 Toronto Dominion Bank Tower Toronto, ON MSK 1E 6
F. Paul Morrison LSUC #: 17000P Tel: (416) 601-7887 Fax (416) 868-0673 Lisa M. Constantine LSUC#: 35064B Tel: (416) 601-7652 Fax: (416) 868-0673 Solicitors for the Respondent 4/()6611
Respondent Court File No. CT 2005-006 COMPETITION TRIBUNAL REPLY AFFIDAVIT OF FRANK MATHEWSON