We’ve long argued that PBM gross margins should decline[1] as a byproduct of several pressures, the most prominent being a shift away from the Average Wholesale Price (AWP) drug pricing benchmark to either of two alternative indices, each of which are far more reflective of true acquisition costs for generic drugs[2]

Since March 2010, when we initiated our negative view of the subsector, PBMs have generally underperformed (Exhibit 1); much of the underperformance is relatively recent, and presumably reflects concrete steps having been made toward the replacement of AWP. This raises the obvious question of whether valuations now account for the full effect of the various pressures building on PBMs

Despite the recent underperformance, we still believe share prices overestimate PBMs’ longer-term earnings prospects, for three general reasons: 1) the contribution of generic margins to overall profitability is far larger than recent changes to valuation would imply; 2) irrespective of the shift away from the AWP index, we see increasing price competition among the ‘a-list’ PBMs as they exit an oligopoly period; and 3) we anticipate disintermediation of the ‘a-list’ primarily by HMOs, and secondarily by smaller PBMs

Generic buying margins

We use MHS as a proxy for the size and composition of PBM gross margins; MHS’ disclosures provide greater detail than can be found elsewhere, and the time series of MHS disclosures is less complicated by acquisitions

MHS’ gross margin consists of four major components: retained rebates, fees from manufacturers, fees from ‘clients and other’, and buying margin[3]. Exhibit 2 shows the importance of each component by year since 2002. MHS, and we believe other large PBMs also, plainly has transitioned from a business model in which gross margins were driven by retained rebates, to a model in which gross margins are driven by buying margin; by 2010, we estimate buying margins were $4.40 / rx, or 74% of total gross profit. Note that we show gross profit / rx in both adjusted and non-adjusted forms. Because MHS includes costs of call center operations in COGS (non-adjusted), we back this amount out using our estimates of relevant costs (Appendix), to reach what we feel is a far more meaningful adjusted set of numbers – especially for the purpose of analyzing buying margins – and we use these adjusted figures throughout this note

This leads to the question of where buying margin comes from – specifically the question of whether it comes more from brands, or more from generics. Unfortunately it’s impossible to directly estimate generic buying margin. We can see AWP, and can estimate the AWP discounts applied to generic dispensing to calculate reimbursement per prescription – however because AWP bears no relation to true acquisition costs, and because we cannot (yet) see true acquisition costs, we cannot directly estimate generic buying margin. But we can back into it

In contrast to generics, brand buying margins are at least reasonably estimable. Also unlike generics, true brand acquisition costs bear a very consistent relationship to WAC[4], and thus to AWP (which is simply 1.2x WAC). Having useful estimates of brand payment terms (expressed as a discount to AWP plus a dispensing fee), we can easily calculate estimated brand buying margins. PBMI’s[5] annual survey of PBM plan sponsors provides a time series history of estimated AWP-based payment terms for brands dispensed at mail, and at retail (Exhibits 3 and 4). If we use the PBMI figures as they are – and we believe they are reasonably accurate – they imply that PBMs’ buying margins for mail brands are negative, and we believe this is correct[6]. The obvious interpretation is that PBMs have given away – or at least very sharply discounted – dispensing of brands at mail in order to capture (very profitable) generic dispensing at mail

To estimate PBMs’ spreads on brands dispensed at retail, we estimate retailers’ brand margins using the PBMI history of retail brand payment terms (Exhibit 4), then assume that PBMs capture a percentage of this available margin by standing between the plan sponsor and the retailer. We don’t know what specific percentage of the retail brand spread the PBM captures, but we can place an upper bound on the percentage. The 2005 FTC report on PBMs (which captured actual 2002 and 2003 PBM transaction data) measured all-in PBM “spreads” on retail brand prescriptions of $5.22 and $6.34 in 2002 and 2003, respectively. Importantly, the FTC’s definition of spreads included all payments made by manufacturers to PBMs, including payments (specifically retained rebates, and retained fees) that are not in our definition of PBM buying margin[7] – thus the FTC figure is by definition larger than the PBMs’ ‘true’ buying margin on retail brands, and so can serve as an upper limit for our estimates. The FTC estimate of PBM spread is equivalent to roughly 30% of the retailer’s dispensing spread for brands; our working estimate is that PBMs’ ‘true’ retail brand buying margin is 10% of the retailer’s brand margin.

Combining our ‘best estimates’ of PBM mail and retail brand margin, MHS’ total brand buying margin appears to be negative – which implies that generics account for more than 100% of net buying margin, and nearly all (97%) of total company gross margin (Exhibit 5)

Critically, even if we move our assumption of PBM brand margins to overly conservative levels, generics still represent the majority of MHS buying margin. Assuming PBMs take 30% of retailers’ buying margins for brands dispensed at retail (the upper limit described immediately above), and earn brand spreads at mail equivalent to those earned by retailers[8], brands still would represent only 26% of total buying margin, in which case generics would of course be the entire balance (74%), and would represent more than half (53%) of total MHS gross margin (Exhibit 6)

We conclude that generic buying margin is at the very least 53% of MHS gross profit, and we expect the true figure is much higher – about 97%. Being convinced generic buying margins will fall as AWP is replaced, we’re then also convinced the current pricing structure of the large PBM model (using MHS as an analogue for all of the larger PBMs) cannot be sustained. This forces the questions of how PBM pricing may change, what margins may result, and whether such a new ‘post-AWP’ pricing / margin equilibrium is consistent with current valuations

Our straw-man model of PBMs’ post-AWP margin structure

We estimate sustainable PBM pricing in terms of PBM clients’ best alternative to their ‘legacy’ PBM. Beyond the usual logic for such a framework, in the case of PBMs we have an industry-specific reason to think in terms of customers’ alternatives – we believe the large PBMs are exiting an oligopoly period, and so believe customers’ alternatives are expanding. For most of the industry’s history, the heart of PBM value creation was the accumulation and exercise of negotiating power over retail brands. This was made possible by a prescription market heavy with expensive but also highly interchangeable brands – an obvious and essential precondition to negotiating rebates. The leverage necessary to extract such rebates consists of several components, including in particular the requisite infrastructure – call centers and mail-order operations – for the heavy lifting of actually switching prescriptions for the “wrong” brand to prescriptions for the preferred brand. However now, as the dollar value of highly interchangeable brands is a falling percentage of the prescription market, the dollar value of available ‘traditional brand’ rebates similarly falls as a percentage of a sponsor’s total drug benefit cost – thus the economic value of the call center and mail-order infrastructure falls as well, as do the barriers to entry represented by these heretofore essential pieces of PBM infrastructure. Simultaneously, at least two alternative ways for formulary managers to create value for sponsors are emerging, namely the rationalization (i.e. lowering) of generic dispensing margins, and more active management of the specialty drug portion of the overall drug benefit. Because reducing generic dispensing margins requires very little infrastructure, more drug benefit managers can emerge as viable alternatives; and, because optimally managing the specialty drug benefit requires something only HMOs have (real-time line sight into the medical record), we believe HMOs emerge as the preferred alternative. It naturally follows that we see the traditional large-PBM oligopoly fading, with attendant pressures on pricing and market share

We’ll take the perspective of an HMO who has contracted with a plan sponsor for an integrated medical – including drug – benefit, and who now must decide how best to administer the drug benefit. Options obviously include using an existing PBM; however we also believe the HMO has the viable option of a more ‘virtual’ drug benefit network – the post-AWP cost of which informs the emerging competitive constraints on PBM margins

The HMO needs a retail network and associated payment / adjudication mechanisms, and may also want at least some level of mail network. Beginning with the retail network: the drug benefit card ‘function’, i.e. adjudication of and payment for prescriptions dispensed at retail, looks very much like a closed-loop payment system (e.g. American Express, Discover). Today’s PBM provides cards, payment processing, and some level of working capital to cover the float between payments to retailers (pharmacies) and payments by plan sponsors. Drug benefit cards clearly also carry formulary and payment information for each beneficiary, and this pharmacy adjudication step is an additional layer to the typical closed loop charge card process. Critically, however, the systems used for pharmacy payment processing are open-source, and highly standardized by the NCPDP[9], so any card issuer is in a position to offer cards with appropriate coding, and related back-office processing, that will be recognizable by any retail pharmacy

Customers’ (all customers, not just prescription consumers) preference for using common credit and charge cards is cultivated by card issuers, who offer consumers various incentives to carry and use a given card. Retailers (generally, not just pharmacies) are under pressure to accept the most common cards, as these have become preferred means of payment for large numbers of potential buyers. These dynamics tend to produce a marketplace with dominant card issuers enjoying scale barriers to entry – and associated high margins – as ‘general merchandise’ retailers have little motive (or perhaps even means) to accept upstart cards that are not in common use. Similarly, retail pharmacies also are under pressure to accept common (typically PBM-issued) drug benefit cards, again because these are the (highly) preferred primary means of payment by so many customers – in many cases the only payment form that allows customers to access their prescription benefit. However unlike credit and charge cards, drug benefit cards are sold not to the consumer, but to the plan sponsor, so the market share of any given drug benefit card is a direct function of that issuer’s (i.e. PBM’s) direct relationship to the plan sponsor, and correspondingly of the PBM’s ability to best meet the interests of the sponsor. Thus if emergent drug benefit managers are able to meet plan sponsors’ interests – as we believe they will be – then we would expect more / smaller drug benefit cards flowing into drug retail than is the case for closed loop charge cards at ‘general merchandise’ retail. Correspondingly, we would anticipate smaller closed loop processing margins for the payment processing activity behind drug benefit cards, which suggests observed closed loop payment processing margins might overestimate the true processing costs behind the drug benefit card. However because the drug benefit card requires the additional step of verifying the transaction against the relevant formulary, and controlling for allowable price, we believe present closed loop payment processing margins are a very reasonable benchmark – i.e. we believe an HMO can reasonably expect to establish its own retail network and benefit cards, by allowing an alternative vendor current closed loop charge card gross margins. This immediately begs the question of whether an HMO with smaller purchase share can get the same pricing on retail dispensing – specifically retail dispensing margins – that larger PBMs can. We believe the answer is yes – provided the HMO has at least some reasonable number of covered lives. This is in large part because retail payment terms are very similar across payors – i.e. pricing is at least somewhat standardized – and, because drug retailers almost certainly will see the advantage of having third party payments controlled by more, rather than fewer, providers

Exhibit 7 reviews transaction volumes, sizes, and gross margins for the major open- and closed loop credit and charge-card issuers. American Express and Discover dominate closed loop processing, with American Express being far larger than Discover. On average, closed loop transaction values are just over $100 – very much in line with average retail prescription purchases

Thus believing we can establish a retail network at closed loop processing margins, we can calculate how much it might cost our alternative retail network provider to manage MHS’ (as a proxy for large PBMs) retail claims (Exhibit 8). This closed loop margin, per retail claim, ranges from roughly $1.45 in 2006 to $1.57 in 2009 (back to $1.29 in 2010)

As our HMO goes through the process of pricing an alternative / virtual network, it can keep track of the alternative network’s likely costs using the current (large PBM) network’s apparent costs as a backdrop. Using gross profit/rx as a proxy for cost of services, the goal is to establish a virtual network whose gross profit/rx to alternative vendors is the same, or less, than a large PBM’s. Picking up the question of alternative mail order costs, assume for the moment that our HMO wants the same proportion of its prescriptions dispensed at mail as is reflected in MHS’ share of mail. Having established the estimated cost of an alternative to an MHS retail network, we can subtract these costs from MHS’ total gross profit to get an estimate of how much gross profit we can afford to give an alternative mail pharmacy (or pharmacies) per prescription. Dividing this ‘remainder’ (MHS gross profit less the estimated cost of the alternative retail network) by total prescriptions dispensed at mail (adjusted to 30 day retail equivalent prescriptions), we estimate that in 2010, we would have needed to have our mail prescriptions filled at a gross profit / rx of roughly $13.69 in order for our alternative drug benefit to be as cost-effective as the MHS benefit[10] (Exhibit 9). Remarkably, this figure is extremely close to the NACDS[11] estimate of average gross margin to retail pharmacy for 2010 (Exhibit 9, again); thus it seems reasonable to believe that our alternative network can be comparably efficient – in terms of the total gross profit we would have to offer a retail administrator and mail order pharmacies – to MHS’ current network

Critically, the preceding numbers make no accommodation for the likely impact on dispensing margins of a shift away from the AWP benchmark – specifically, the preceding numbers (both MHS’ and the NACDS estimate of average retail dispensing margin) all incorporate what we believe are unsustainably high generic dispensing margins

We’ve until now favored brand dispensing margins at retail – about $5.64 in 2010 – as a proxy for where average retail dollar margins might go after AWP is replaced. The logic was that the negotiation for brand payment terms is a competitively efficient and fully-informed (no AWP-associated information asymmetries) process, and accordingly should reflect a natural equilibrium point for retail dollar margins. Having considered the operating economics of a pharmacy with this average gross margin across all prescriptions, we recognize that $5.64 / rx is almost certainly too low – i.e. it’s now clear to us that retail pharmacies have yielded brand dispensing margin on the expectation that larger generic dispensing margins would average the pharmacy’s overall dispensing margins to an acceptable level. Taken to the extreme, this might argue that current dispensing margins are in fact the post-AWP equilibrium, but we don’t believe this is true. AWP creates frankly enormous information asymmetries and risks in the generic dispensing margin negotiation[12], and this must inevitably find its way into (necessarily larger) dispensing margins. Thus when AWP is replaced, we still believe average dispensing margins fall – the question is where to?

Reaching back to a point (1990) at which PBMs had comparatively little effect (third party payments were 26% of retail payments, v. 80% now), generics were a much smaller percent of retail volume (30.2% then v. 71% now), the number of pharmacy outlets per person was similar to today’s level; and, fewer prescriptions were dispensed in the presumably more price-competitive non-traditional (mass merchant, supermarket, mail) outlets (non-traditional outlets had roughly 25% of rx’s then, v. 40% now), average retail pharmacy dollar margins were $6.62[13]

Inflating this $6.62 margin forward at CPI, its real value in 2010 dollars is $11.04 – $3.07 or 21% less than the current $14.10 estimated average dispensing margin at retail. Bearing in mind that changes in supply (outlets / capita) almost certainly cannot have driven real pricing gains, that competition among outlets should have increased as presumably more price-competitive outlets[14] gained share of prescriptions; and, that efficiency gains[15] in retail pharmacy operations over the same time period might reasonably have been expected to reduce real dispensing margins, it’s at least somewhat odd that real dispensing margins actually grew

We recognize it may be a leap to argue that average retail dispensing margins after AWP will fall to the real equivalent of $11.04 (or so) in 2010 because, as suspicious as we are that AWP information asymmetries and risks are the dominant driver of the real gain in dispensing margin, we plainly cannot prove it

However – maintaining the perspective of a reasonably sizable HMO who’s in the process of ‘costing’ an alternative network – we are very confident of being able to have our mail prescriptions filled in the post-AWP world at or below a gross profit of $11.04 / rx in 2010 dollars. First, we in fact do believe that post-AWP average dispensing margins at retail fall to or toward this real level – and we plainly have the option of shifting prescriptions from mail to retail. Notwithstanding this, our alternative network has a mail-processing advantage over the large PBM – we have no need to build a mail order facility (we don’t think we’ll need the infrastructure as leverage to get retail brand rebates, and we certainly don’t anticipate using it to capture out-sized dispensing margins); and, we’re willing to have our mail prescriptions filled by any legitimate provider that meets our network requirements

Importantly, we don’t believe that mail order operations are wildly more efficient than retail operations – the lower pharmacist and real estate costs in the mail setting are nearly entirely offset by the presence of shipping costs (Exhibit 10). Alternative mail-order options plainly already exist (e.g. Drugstore.com); and we believe the entry barriers to de novo mail operations are comparatively low[16]. Consider current order patterns for drug wholesalers, the notion being that wholesalers’ customers order with sufficient frequency to have forced average order sizes down toward levels close to those we would expect for consumers ordering by mail – thus a typical wholesaler is very nearly configured to handle consumer orders. In 2010, the average distribution center order had a retail value of roughly $7,500 – well above the average consumer order. However, we estimate the average wholesale shipment would weigh only about 2.4 pounds, with a shipping volume of just under 700 cubic inches – slightly larger than a typical shoebox, and about the same weight as a hardcover book (Exhibit 11). Thus despite the high average dollar value of average wholesale shipments, the physical characteristics of the order are very much in the range of typical consumer orders. And, because the typical distribution center deals with thousands of customers through both telephone and electronic interfaces, and ships a large percentage of orders using common ground and air carriers (e.g. FedEx, UPS, DHL, USPS), the customer interface (including invoicing and collection) is not wildly different from a consumer-level operation. We recognize that large wholesalers – especially those with large mail order clients – are unlikely to rush into mail-order dispensing. However, we would emphasize that a very large percentage of the nearly 200 wholesale drug distribution centers in US operation could be expanded to mail-order dispensing with a comparatively low-cost build out of order handling (order, invoice, shipment). Accordingly we as builders of an alternative / virtual network plainly have the option of arranging for such build-outs as owners of a facility, or encouraging such build-outs as clients of a facility – or ideally as clients of two or more competitive mail-order fulfillment facilities

Continuing with the assumption that our HMO wants the same share of its prescriptions filled at mail as is the case for an average MHS client, and believing these prescriptions can be filled at or below our ‘constant real’ dispensing margin assumption of $11.04 in 2010 dollars, we might reasonably expect an alternative mail network to fill our prescriptions $2.65 / mail rx more efficiently than we can buy this service through MHS. More meaningfully[17], we might conclude that we can buy the combined retail and mail services from an alternative ‘virtual’ network of providers for roughly $0.91 less per prescription (combined retail and mail) than the current cost of buying these same core services through MHS (Exhibit 12)

For the sake of (at least relative) clarity, we’ve so far ignored the matter of rebates – MHS can plainly get brand rebates for us; but if our virtual network cannot get rebates, MHS remains the more efficient choice. If we thought MHS’ gross profits / rx were the same as our alternative network’s, then we would face zero opportunity cost for using the MHS network, and our net benefit would be simply our share of rebates net of the fees owed to MHS (Exhibit 13). In 2010, this net benefit was roughly 5.6% of the average prescription price

However if an alternative network can operate more efficiently than MHS’, then the excess cost of MHS’ network (relative to our own alternative network) is a true opportunity cost, which reduces our net benefit (Exhibit 13, again). Because we’re only predicting that our alternative network is (in 2010) $0.91 more efficient than MHS, it’s still not worthwhile to build and operate the alternative if we have to forego net (of fees) rebates

Because brand rebates represent a declining percentage of total drug benefit costs, and generic dispensing premia an increasing percentage of drug benefit costs, as time passes our alternative network becomes more and more efficient – meaning MHS’ current pricing structure becomes less and less sustainable. At the very least, we would expect MHS to have to maintain some efficiency advantage relative to alternative / virtual networks in order to remain viable, meaning MHS at some point has to scale its gross margins – i.e. reduce its pricing – to a level much closer to the gross margins of the alternative network. All else equal, if MHS reduced its gross profits to what we believe is feasible in our alternative network, 2010 gross margins would fall by 16.4% (Exhibit 14)

In truth, we believe MHS margins will face more pressure – because we don’t believe our alternative / virtual network has to forego rebates all together. Recall that brand rebates tend to have two components – a ‘base’ rebate for access to the network, and a ‘market share’ rebate for access to the preferred tier(s), and/or for market share actually achieved. Without heavy investment in owned call center and mail-order infrastructure, our alternative network cannot hope to have the same ability to move market share as the large PBMs – so we cannot expect to get the ‘market share’ portion of the rebate

However we can expect to get the ‘base’ portion of the rebate that brands offer for inclusion on non-preferred tiers. Because we know that the incremental market share portion of the rebate is constrained to the difference in the dollar value of co-payments on non-preferred and preferred tiers[18] (≈ $21); and, because we know typical rebates for brands that are paying rebates (≈ $51), we can estimate the current value of ‘base’ rebates being offered by brands as ≈ $30. This implies that roughly 59% ($30/$51) of rebate dollar value being offered by brands is accessible to our alternative network. Adding 59% of MHS rebates to our opportunity costs, we’ve effectively recreated a virtual / alternative PBM network having the same net benefit to our HMO as we would expect as a client of MHS (Exhibit 15). And, because our alternative network’s brand rebates fall (we only get 59% of these) but generics’ share of prescriptions grow (we get all of these savings), we can reasonably expect our alternative to become more efficient as time passes. All in, this implies that MHS (and all large PBMs, for whom we’ve used MHS as a proxy) must adjust their pricing – i.e. lower their gross margins – in order to be more efficient than customer’s next best alternative. Specifically, we believe large PBMs will at the very least need to reduce gross profits by roughly 16.4% (Exhibit 14, again) as a direct response to lower real post-AWP dispensing margin. And, for the large PBM to remain the most efficient option, we believe PBM pricing has to fall at least somewhat further if the large PBMs are to have any reasonable chance of defending their current shares of beneficiaries. All in, we see gross margins falling by at least 20 percent over the near- to mid-term

Importantly, this 20 percent fall in gross margins ignores likely changes in a typical drug benefit’s retail / mail mix. Large PBMs have clearly pushed mail dispensing – in large part by very small to negative brand dispensing margins at mail – presumably because mail dispensing offers greater negotiating leverage over brands, and more recently because dispensing generics at mail has become the mainstay of PBM profitability. As interchangeable retail brands fall in relative importance, and as the AWP-based generic dispensing arbitrage also fades, PBMs are less able to rationalize subsidizing mail order operations at dispensing shares that appear to exceed their clients’ interest levels in mail dispensing. The implications for PBM gross margins are abundantly clear: if mail’s share of total dispensing begins falling, large PBMs look more and more like closed loop payment processing companies – who as we’ve seen earn 2.6% – 2.7% gross margins, falling below the roughly 8% ‘true’ (excluding call center costs) level of large PBMs’ current gross margins

Returning to where we started – the question of whether PBMs’ lower valuations reflect the challenges to PBMs earnings power – we still believe the answer is no. Larger PBMs’ out-year earnings estimates still call for considerable growth; and despite the recent sell off, the larger PBMs’ shares still trade above the market on 1 and 2 year forward price-earnings multiples. Valuations approach market multiples on 3 year forward estimates, implying the market anticipates earnings reaching these levels across this longer timeframe. In contrast, we see at least a 20 percent gross margin contraction – to say nothing of the likelihood that negative operating leverage multiplies this percentage loss at the operating and net income levels

Accordingly we continue to have a very sharply different outlook for PBM earnings than the market. Because timing the availability of reliable alternatives to AWP remains challenging – as does estimating the rate at which PBM contracts shift to the alternative benchmarks – we realize we face the risk of PBM stocks outperforming as generic mix increases into 2012, and as investors potentially anticipate operating synergies (and/or pricing stability) from an ESRX/MHS combination. Despite the risk of missing a rebound in near-term share prices, we continue to favor underweight / short positions on PBMs, as we believe the pressures facing this business model will continue to come into view, and that these pressures ultimately will drive the industry’s earnings power down by a third or more