AXP: Structurally Improving Business Model
SEE LAST PAGE OF THIS REPORT Howard Mason
FOR IMPORTANT DISCLOSURES 203.901.1635
March 26, 2014
AXP: Structurally Improving Business Model
- The economics of AXP’s business model are improving meaningfully as it overcomes the anti-competitive constraints imposed by V and MA through the exclusionary rules (preventing Amex from distributing its network services through the bank channel) which were struck down by the Courts in 2004. Amex is achieving this in both credit and debit:
- Credit: Amex’s GNS business (where a bank, rather than AXP itself as in the “proprietary” business, issues an Amex-branded card) has grown at a 20%+ CAGR since 2004 and now accounts for $140bn of annual purchase volume representing 15% of Amex’s total. AXP recently announced that WFC and USB will be issuing Amex-branded cards and, unlike partnerships with C and BAC which have tended to focus on the T&E segment, we expect these new programs to include everyday spend cards.
- Debit: Through the Serve brand (with annualized transaction volume of ~$5bn at the end of its second full year), Amex offers consumers a checking-account alternative through retail partners, such as WMT, and online/mobile channels at substantially lower cost (all-in $1/month which is waived with direct payroll deposit) to typical low-balance checking accounts. Serve (which offers card, ATM, and check-access, online bill-pay, and FDIC insurance, check-access) has a structural cost advantage because Amex does not support a branch infrastructure and is not subject to the Durbin cap on merchant debit fees. In the US, the addressable market for Serve (at $5tn including debit, cash, and checks) is twice that for Amex’ credit products (see Exhibit below)
- By 2020, we expect over one-third of Amex’s volumes to be from its GNS and Serve businesses. These are balance-sheet efficient businesses (because, unlike in the “proprietary” credit/charge card business, Amex does not have to support cardholder balances) and generate an estimated ROE in excess of 60% versus 15-20% for the proprietary credit business.
- As a result of this mix-shift, we expect the firm-wide ROE to increase from the current 25% to over 30% creating a positive re–rating of the stock of at least 20%. We are not concerned at the prospect of “brand dilution” as Amex expands its customer and merchant footprint nor the possible acceleration in the decline of the discount rate (which AXP will mitigate, if not overwhelm, by adding value to merchants through end-to-end management of loyalty programs including coalition programs):
- Brand dilution: The days of aspirational card branding to the affluent are waning as consumers look for service and rewards; among the unbanked, Amex remains an aspirational brand and this is part of the appeal of Serve.
- Discount rate: Amex’s discount rate (representing the ratio of revenue, before payments to third parties, to billings on merchants where Amex is the acquirer) has declined by about 1 basis point per year over the last decade and presently stands at 2.51%. This is largely as Amex has diversified from the T&E segment accounting for ~30% of billings versus 34% in 2007. The offset is improving network economics as AXP builds “everyday” spend billings particularly on products where it is not supporting cardholder balances.
Exhibit: The Addressable Market for Serve Globally and in the US
The Distribution Constraint Faced by AXP
American Express is a payments network company that has historically been subject to a vicious distribution constraint.
Until 2004, Visa and MasterCard prohibited distribution partners (i.e. bank issuers of credit and debit cards) from distributing the products of competing networks such as AXP and DFS. These “exclusionary rules” were successfully challenged by the Justice Department as anti-competitive violations of the Sherman Act; in October 2004, the Supreme Court refused to hear the case leaving in place an appellate (and lower) court decision barring Visa and MasterCard from enforcing them.
The legacy of the exclusionary rules casts a shadow even today. Amex has 53 million credit cards in circulation in the US versus 254mm Visa credit cards and 178mm MasterCard credit cards. Furthermore, AXP has no meaningful debit business (versus 431mm debit/prepaid cards from Visa and 158mm from MasterCard) because, by definition, debit is a DDA-access product and AXP was not able to offer network services to the banks administering these accounts; effectively, V and MA locked up the debit distribution channel entirely.
Given the bank distribution bottleneck, AXP distributed through corporates (with T&E cards) and ran a vertically-integrated consumer business: it was its own distributor of the network service – in other words, unlike V and MA, AXP was both issuer and network (and recruiting merchants as distribution partners as, for example, in the case of the Costco-Amex product). This distribution strategy had two consequences:
- T&E Tilt: Because of the corporate card business, AXP has a heavy tilt to T&E spending versus “everyday” spending categories (such as supermarkets) so that, even today, T&E represents nearly 30% of AXP purchase volume, albeit down from 34% in 2007.
- Limited Footprint: AXP’s business is more balance-sheet intensive than V and MA (because of the need to finance cardholder balances) so that network growth is limited by balance-sheet capacity. This, in turn, has limited the company’s footprint on the cardholder and hence merchant sides (with AXP accepted at 3.5-4mm merchants in the US versus nearer 6mm for V and MA).
To manage the balance-sheet constraint, Amex has traditionally focused on “premium” customers that are balance-sheet efficient in the sense of using their cards more for spending than borrowing. The cornerstone charge card, of course, does not allow customers to revolve a balance but even when, through its credit card products, Amex does allow consumers to revolve a balance it is highly selective. As a result, spend velocity (i.e. trailing annual purchase volume divided by loan balance) is 3.6x for the US card-lending business (8.7x for the overall US business including charge cards) versus 2.1x for the industry.
AXP made a virtue of necessity by arguing to merchants that its premium customers were more valuable than those of V and MA and merited a higher fee; in 2013, for example, the average annual spend on a card issued by Amex was over $16.3k versus just over $4.1k and $3.3k for V and MA respectively. As a result, Amex products have traditionally had higher acceptance costs than V and MA cards although the fee gap has narrowed over time (see Exhibit 1). AXP’s aggregate fee has declined as the billings mix has shifted slowly from T&E (where merchant fees are typically highest), and the acceptance fees for V increased sharply in 2008 with the introduction of a premium category for cards with qualifying high-spend characteristics.
Exhibit 1: Card Acceptance Fees for AXP and V (premium cards) as % of Purchase Volume
Source: Amex figures are the company-reported discount rate representing the ratio of revenue, before payment to third-parties, to billings for merchants where Amex is the acquirer. Visa fees are estimates at a typical merchant for a $40 transaction from GAO Report to Congressional Attendees, November 2009 and SSR Estimates.
The “EveryDay” Importance of the WFC and USB Partnerships
When the exclusionary rules were first eased in 2004, AXP began to build its Global Network Services (GNS) business where it acts as the network but a third-party “issuer” bank manages the relationship with, and funding of, the cardholder. The GNS business accounted for less than $20bn in purchase volume in 2003 and over $140bn in 2013 (of total AXP purchase volume of ~$950bn) so that it has grown at a CAGR of over 20%. Among the bankcard issuers, early adopters of the Amex network included MBNA (2004) and C (2005). BAC, after acquiring MBNA in 2005, continued to issue Amex-branded cards but, like C, focused on travel-oriented products.
The significance of the credit-card partnership with Amex announced by WFC last August (with cards expected to be launched in the second half of this year) is that it is part of a fundamental rethink of its credit card business (where WFC punches under its weight) and will involve “doing a lot across the full spectrum” of Wells Fargo credit cards including but not limited to a travel-oriented product. We expect the USB partnership with Amex, announced in November, will also involve every-day spend cards. The introduction of Amex-branded cards issued by WFC and USB, along with launch next month of two “EveryDay” credit cards (which will offer cardholders extra rewards for spending at supermarkets and frequency use) issued by AXP itself will break the dam for AXP in everyday spending. In particular, the new Amex-branded everyday-spend will engage network effects: the business case for every-day merchants to accept Amex cards will improve making the card more useful for consumers and hence improving the business case for banks to issue every-day Amex-branded cards.
Financially, the impact will be increasing diversification for AXP from T&E spending and hence downward pressure on the discount rate (although we expect AXP to mitigate, and possibly overwhelm, this by adding value to merchants through end-to-end design of loyalty programs particularly in the mobile channel), and an accelerating mix-shift to from proprietary volumes (where AXP is the issuer) to GNS volumes (where a third-party bank is the issuer). This mix-shift is important since we estimate GNS generates an ROE of over 60% versus ~20% for issuing business. GNS volumes currently account for ~15% of AXP’s total volumes but this will increase to near one-third by 2020.
The Debit Opportunity
Network economics are highly-sensitive to volumes not only because marginal cost is near-zero but also because of network effects: the value of a network to all participants increases as new participants join. The challenge for Amex is that even as it grows, and with the WFC and USB partnerships will likely continue to grow, share of the credit market (see Exhibit 2), the Visa and MasterCard networks generate scale and scope economies, as well as network effects, from their debit businesses. Thus, while AXP has a 26% share of the credit market, the relevant competitive market is purchase volume across credit and debit/prepaid cards where Amex has only a 15% share (see Exhibit 3).
Exhibit 2: Amex Share of US Charge/Credit Purchase Volume
Exhibit 3: Share of Purchase Volumes on US General-Purpose Cards
While we expect the WFC and USB partnerships to evolve over time to Amex-branded debit cards, the initial focus is Amex-branded credit cards. In the meantime, Amex has focused on developing its prepaid business, under the “Serve” brand, leveraging the mobile channel and merchant distribution (with CVS and 7-Eleven as distribution partners). Indeed, WMT is now a distribution partner for Amex’ prepaid products marketed by WMT under the Bluebird brand. The Bluebird-Amex card is reloadable prepaid card with consumers able to access funds through the card (at point-of-sale where Amex is accepted or through 23,000 MoneyPass ATMs), through online bill pay or, subject to pre-approval, checks. In short, Serve including Bluebird provides checking-accounts on a card including offer direct deposit for payroll and FDIC insurance.
Ironically, the impact of Durbin is to give Amex a regulatory advantage in this alternative to a bank checking account. The reason is that, unlike bank debit cards, Serve products are not subject to the Durbin fees caps because Amex, as a three-party network, is not subject to the cap. There is an important subtlety here: in general, prepaid cards are not subject to Durbin provided they offer only card-access to funds. If cardholders have non-card access to funds (such as through online bill pay or checks, for example), regulators deem the product a debit card and subject to Durbin fee caps. Serve is exempt not because it is considered a prepaid product (since non-card access is a key product feature) but because Amex is exempt. If Chase, for example, were to offer a product with Serve features, its economics would be inferior (at least on transactions over $10) because Amex assesses merchants its usual discount fee for accepting Serve while Chase would be limited to 24 cents/transaction.
As a result of this fee advantage on the merchant side (and without a branch or ATM infrastructure to support which probably accounts for ~50% of the cost of a typical checking account), Serve is able to offer its checking-account alternative to the consumer at materially lower-cost than a typical low-balance checking account. The fee is $1/month and that is waived for consumers who use direct payroll deposit or add $500 during a monthly billing cycle; there are no overdraft fees (and the pre-approval process means that checks cannot bounce) and no minimum-balance requirements. The prepaid business at Amex, including Serve, is growing fast and now generating transaction volumes of ~$5 billion annually (versus $1.35bn in 2012 which was the first full year) with over 5 million customers. The expansion of Amex’s cardholder footprint is important since most consumers are new to the brand: Bluebird alone has added 1mm accounts of which 87% are new to Amex. There is significant runway for growth in the prepaid segment as part of the global shift from cash and checks. Amex clearly sees the Serve offering as positioned to compete with cash, checks, and debit and estimates the addressable market as over $5tn in the US and $25tn globally (see Exhibit 4).
Exhibit 4: The Addressable Market for Serve Globally and in the US
Source: AXP Financial Community Presentation, August 2012