March 24, 2009

I have pointed out credit rating agencies got off lightly, even though Madoff's credulous victims lost less money collectively than those who gave credence to their toxic notations. David Segal mischievously reminds us that Warren Buffet "owns a stake of roughtly 20 percent in the [...] parent of [...] Moody's" (*). While the sage of Omaha keeps uncharacteristically silent on the subject, a spokeman for Moody's echoes Dewen Sharma's defense of Standard & Poor's, claiming "potential conflicts of interest with any ratings system, whether issuers, government or investors pay".

I suggested a rule by which rating agencies would be fined for their mistakes may inspire them to work harder. But who wants to be paid on results when one's monopolistic position enables one to be paid upfront? Are weathermen paid on the accuracy of their forecasts? Perhaps we need to check the other business models and see whether they are all riddled, as alleged, with intractable conflicts of interest.

Long time readers of these fillips know that I see recommendation systems as fundamental to our Information Age (1). This is what Google is, for better and for worse. This is what newspapers are and, accelerated by search engines, their impending demise shows how crucial it is to find the right business model. This is again what investment research is, although Adrian Cox confirms how elusive can a sound business model be (**).

Worse still. Crooked models have a head start for insiders are the source of the best recommendations. As John Kay shrewdly analyses, many a victim mistook Madoff for such an insider, bent on enriching both his clever clients and himself by steadily stealing from impotent third parties (***).

The recommender may be interested on each transaction recommended. Whether it takes the form of a salary or a commission, the recommender is but a salesman and buyers beware. Sears auto mechanics famously milked consumers by recommending repairs their cars did not need (2).

When he or she shares the risk of being inaccurate, as AIG did when it insured the mortgage based securities so highly rated by Standard & Poor's and Moody's, the recommender has no apparent conflict of interest. Unfortunately its employees, rogues and stars alike, can find advantageous to saddle their company with excessive long term risks for their short term gains.

Disinterested recommenders must find an independent source of income. If it comes from those who are recommended, willy-nilly they will deliver the best recommendations money can buy. "The big banks retort that they are scrupulous in keeping [their analysts] apart [from their commercial deals]". Scrupulous? Perhaps. Successful? In your dreams! Wolves have never made good shepherds.

More credible is the argument that recommenders paid by those who seek their advice are likely to want economies of scale and their clients reluctant to pay for recommendations "so unreliable" by nature. In a genuine adversarial system though, and recommendations paid by a party imply such a system, both parties to a transaction pay their own lawyers. The truth is adversarial systems do work but are quite expensive.

The best model today is still provided by the recommender acting pro bono in view of an independent source of income. University professors give recommendations on their students, pastors on members of their flock. Local bankers notarize their clients' documents for free. Google itself owes it success to an original and brilliant formula which recommends results to a search for free while auctioning separate ad space keyed to the search.

I admit the best recommending models can still be perverted by greed fed by the ready availability of abundant insider information. "Fourteen trading firms, including subsidiaries of some of Wall Street's top banks [...] settle[d] civil charges that included allegations they "traded ahead" of clients for their own benefit", reports Joanna Chung (****). These market makers, whose useful role is to behave as interested recommenders of share prices, became too greedy. Similarly Google now legally but quite unethically steals consumer private information to target more lucrative ads.

To combat such temptations, I have long advocated decentralization, which cuts risks down to size, and preventing service providers from accessing private data, which eliminates the risk. One way to do so is to insure roles be kept separate. But doesn't this approach rely too much on individual recommenders doing it for fun and no profit? Doesn't it exacerbate the lack of scale mentioned by Adrian Cox?

The key variable to consider is time. Listen to Ian McCallum quoted by Adrian Cox. "All fund management companies will use research. The big difference is whether you are using them as a source of information as opposed to using their forecasts." This sober reminder points to the solution. Recommendations are based on the past. One should not overemphasize their predictive value and, like weather forecasts, update them frequently.

If recommendations are but information, recommenders are but a type of content providers. In today's free information flow, legal or not, recall value is in timely access to the latest version of the facts. Recommenders could give free recommendations to those who ask and, when relevant, make them public for free after a suitable delay, but charge the party to whom a recommendation is submitted for verifying this recommendation confidentially online and bill the recommended party an equal amount for the service.

Notice this model rewards the recommenders who do not hesitate to change their minds to account, like weathermen, for the latest developments. If they had issued a glowing recommendation and learn a fact to the contrary, they are not in the debt of those they recommended nor do they owe anything to those foolish enough not to verify past recommendations. What if recommendations do not change often enough? Recommenders can still make a living. Value will be concentrated within narrow windows but nonetheless updates must be taken regularly lest rare events be missed.

Such a model has no built in conflict of interest and will benefit from whatever economies of scale may exist. Competition will reward those who satisfy their clients the best by their timeliness and their accuracy. Perceptive readers may wonder if this is not like creating and selling insider information. Absolutely. But in the absence of a fiduciary duty, selling such information equally to all who want it is a business, not a crime.

Quoting a US survey by TRUSTe on privacy relayed by Stephanie Clifford, 75 percent of respondents think "The Internet is not well regulated, and naïve users can easily be taken advantage of" (*****). To better fleece them, American companies do love to pull the wool over their eyes by giving tainted recommendations.

To make recommendations more robust, I recommend the government applies appropriate market rules.

Philippe Coueignoux

• (*) ......... The Silence Of the Oracle, by David Segal (New York Times) - March 12, 2009
• (**) ....... Investment research fighting its corner, by Adrian Cox (Financial Times) - March 20, 2009
• (***) ..... How the 'Madoff twist' entices the financially astute, by John Kay (Financial Times) - March 17, 2009
• (****) ... Firms pay $70m to settle trading charges, by Joanna Chung (Financial Times) - March 5, 2009
• (*****) . Many See Privacy on Web As Big Issue, Survey Says, by Stephanie Clifford (New York Times) - March 16, 2009
• (1) see "recommendation mechanisms" in the list of Major Themes of these fillips.
• (2) source: David Streitfeld (The Washington Post), June 1992 as quoted by Highbeam.com
  Note that Google recommendations #9 and #13 on Sears auto repair scandal turned out today to be for term paper mills advertising the same document.
  Welcome to ethics 101, the on line version.