The Predictioneer’s Game: Using the Logic of Brazen Self-Interest to See and Shape the Future

by Bruce Bueno De Mesquita

  It is not hard to see how territorial concessions can be organized on a continuum. Although the scale above is not based on percentages of land or land value, still there is a natural progression in choices ranging from Israel’s annexing contested territory on to no concessions by the Israelis and finishing at the other end of the scale with a fully independent Palestinian state. From these beginnings in 1987, I prepared a forecast that would closely predict the actual territorial concessions agreed to between the Israelis and the Palestinians in 1993 at Oslo. (We will look more closely at this forecast a little later in this chapter.) Let me reemphasize a central assumption behind the scale, an assumption I introduced casually in the discussion of North Korea. The scale tells us that someone advocating a weakly autonomous Palestinian territory (70 on the scale) strongly prefers the status quo (at 85 on the scale) to a territory closely in federation with Jordan (25 on the scale). We know the preference for 85 is stronger than for 25, even though 85 and 25 flank the advocated position at 70, because 70 is much closer numerically to 85 than to 25. That is how each scale works. Numerical values that are closer to the numerical value of the player’s advocated position are liked better by the player than positions reflected by numerical values farther from the advocated position. Knowing that, let’s have a look at a business issue that can help us understand more about how to turn problems into a well-defined numerical scale.

  Here the focus is on what an issue’s scale might look like when the question involves something less obviously numerical in character than land for peace. To see an answer to this, consider the following range of choices from a litigation I worked on some time ago. Naturally I have masked the details to protect my client, but the ideas should be clear enough.

  WHAT CHARGES WILL THE U.S. ATTORNEY BRING AGAINST THE DEFENDANT?

  Scale Position

  Meaning of the Numerical Value on the Scale

  100

  Multiple felony charges including several specific, severe felonies

  90

  Several specific, severe felony counts but no lesser felonies

  80

  One count of the severe felony plus several lesser felonies

  75

  One count of the severe felony but no other felonies

  60

  Multiple felony counts but none of the severe felonies

  40

  Multiple misdemeanor counts plus one lesser felony

  25

  Multiple misdemeanor counts and no felonies

  0

  One misdemeanor count

  This scale makes clear how the client prioritized possible outcomes. The definition of 0 on the scale tells us that the client did not believe that any stakeholder would argue for the dismissal of all charges. That was not a choice included in the range of feasible outcomes. The upper bound on the scale indicates that the client believed at least some in the U.S. Attorney’s Office or others involved in the process would argue for severe criminal penalties. The real action was to unfold in between, and we’ll take a deeper look at this “case” in a later chapter.

  Let’s take a quick look at this issue scale. It can shed a lot of light on how to think about issues. We can see that the two key “distance” differences here, laid out by the client, are the plea from 25 to 40 (no felony vs. one lesser) and from 40 to 75 (no severe vs. one severe). After that, life gets worse, but not nearly as dramatically worse as happens with the move from 40 to 75. This tells us, among other things, that avoiding a severe felony charge altogether was of greater value to the client than adjustments in the number of severe felonies that might have to be pleaded to.

  In framing issues this carefully, the decision makers begin to learn things they had not recognized about their own problem. Instead of an amorphous question, they now have focused issues to address. They have confronted the problem and defined what the meaning is behind the choices they face. Once they go through the interview process to identify the remainder of the data—who the players are, where they currently stand on the issue, how salient it is to them, and how much clout they can each exert—they have, for the first time, a genuinely comprehensive view of how high a mountain they must climb. As I mentioned, we will revisit this case in a later chapter to see how they thought it would turn out and how it actually did turn out. For now, let’s return to the napkins covered with data on how to promote peace in the Middle East.

  PEACE IN THE MIDDLE EAST?

  The information scribbled on a couple of napkins turned out to provide a terrific grounding for figuring out what was going to happen in the Middle East over the several years following 1987. Eisenstadt knew his stuff, and so did Harold Saunders. What did we learn when the data were subjected to the model’s dynamics, allowing all of the players the opportunity to bargain with each other, trading territorial concessions for political credit? The predicted outcome was reported in a 1990 journal article I wrote based on the cookie-hour napkins.2 The article was published three years before the Oslo Accords created the Palestinian Authority. It predicted resolution at position 60 on the issue scale out of a feasible range of 0 to 100. Sixty on the scale was defined at the time as being equivalent to the Palestinians’ gaining some weak localized territorial control over a semi-autonomous entity. Remember, the status quo was 85, and lower values reflected greater and greater concessions to the Palestinians. We now know that the territorial concessions that were actually worked out between the Israeli government and Yasser Arafat were equivalent to about 60 on the scale, as I predicted in print three years beforehand.

  Of course, this prediction was made a long time ago and was intended to be good only once there was a change in Israel’s government. As I wrote at the time, “Given [then Israel’s prime minister Yitzhak] Shamir’s apparent perception of the situation, we must conclude that as long as he is Prime Minister of Israel, there is no reason to expect significant progress.” Shamir, who became Israel’s prime minister in October 1986, was replaced by Shimon Peres in July 1992, paving the way for real progress between the Israelis and the Palestinians.

  The analysis, however, was more detailed and nuanced than just a prediction of territorial concessions. My 1990 study went on to suggest that the then large and influential Popular Front for the Liberation of Palestine (PFLP) would “become politically isolated and irrelevant in negotiations. We can anticipate that they would respond to such a situation by increasing terrorist acts, aimed not only at the Israelis but perhaps also at the PLO leadership. If the analysis is correct and if the PLO adopted a strategy of incremental moderation, the future of the PFLP would be to flicker out of the picture.” Remember, this was out there for anyone to see and read in 1990. That is what I meant earlier when I said we must be willing to risk embarrassment if we want people to have confidence in our predictions. Certainly few in 1990 would have argued that the PFLP was likely to “flicker out of the picture.”

  Today, two decades later, we can look back to see how accurate that prediction was. For instance, what do independent sources now say about what happened to the PFLP following the creation of the Palestinian Authority in 1993? BBC News Online, reporting on January 16, 2002, quotes Abdel Bari Atwan, editor in chief of the London-based Arab newspaper Al-Quds, as saying “The movement [i.e., the Popular Front for the Liberation of Palestine] had become marginalized. … It had gone from being the second most important Palestinian group to forth [sic] or fifth.”3 Anthony Cordesman, a highly regarded Middle East scholar and expert commentator for ABC News, echoes the BBC view. Reflecting back on the PFLP, he writes, “The PFLP opposed the Oslo peace process. As the PA [Palestinian Authority] and Arafat’s Fatah gained strength, the PFLP became increasingly marginalized.”4 Similarly, GlobalSecurity.org, generally seen as a reliable source of information, notes that, “Once a key player in Palestinian politics, the PFLP lost influence in the 1990s and was sidelined as Yasir Arafat established the Palestinian Authority.” The PFLP’s decline is similarly reported by numerous other sources.
Most of these sources attribute the PFLP’s decline to events that happened after my 1990 article was published and long after Shmuel Eisenstadt wrote numbers on a couple of napkins. That is, the modeling exercise undertaken before the 1991 Gulf War, before the first intifada, and before the Oslo Accords foresaw the territorial changes between the Israelis and Palestinians, the necessity of a Labor government coming to power in Israel, and the decline of the PFLP. It foresaw the fundamental developments far enough ahead that even today people have trouble recognizing that the seeds of the agreement in 1993 had already been planted and were growing, unseen by most onlookers, well before.

  What was it that the model identified in the give-and-take between the Palestinian side and the Israeli side based on data collected in 1987 and 1989 that led to a successful prediction? This is really the crux of the matter. After all, working out that the predicted outcome could lead to a stable agreement involved more than just pondering Eisenstadt’s original, vague question: “So, you can predict how to make peace in the Middle East?” Since Arafat and the Israeli prime minister both had vetoes over any deal, it was essential to figure out whether they were likely to end up supporting the same agreement, and if so, why. Here’s what the model indicated:

  Israel’s Labor Party leader, Shimon Peres, was not in power when the study was done, and it was evident in the model’s results that until Labor came to power no agreement could be reached. That meant that the focus of attention in analyzing the possibility of a peace deal—the problem put to me by Shmuel Eisenstadt—had to be on Peres and the Labor Party. The model’s logic produced output that indicated that Peres believed he would face a lot of political pressure at home over his stance regarding the Palestinians. To attain and retain power—the goal of every politician—he believed (according to the model’s logic) that he had to look as though he would be tougher in Palestinian negotiations than was implied by the quite moderate stance he took in the late 1980s. The model showed he needed to be only a bit more moderate than Shamir, leading him to agree to occupy a position in the mid- to upper 60s instead of Shamir’s drift between 70 and 85 on the scale. So Peres was predicted to be responsive to political expediency at home.

  For his part, Arafat concluded, according to the model’s simulations, that he needed to moderate his own stance to keep Labor from taking too hard-nosed a view within Israel. The model suggested that he would choose a course of action based on his personal political welfare rather than the well-being of the Palestinian people per se. As I wrote in the 1990 article, “The model solution suggests that Arafat would stabilize his political position, leaving himself devoid of serious political opposition either among the Palestinians or within Israel. … If Arafat does choose to moderate his stance, this suggests that he is willing to sacrifice both the Palestinian cause and his opponents at the altar of his personal political welfare.” That seems to have been the case. Thus the analysis went from a vague question to precisely structured propositions and a detailed analysis of the negotiating dynamics that made it impossible for Shamir and Arafat to reach agreement and those that made it possible for Arafat and Peres to come to terms once Peres became prime minister.

  Asking the right questions and isolating the key interests for a given problem is too often a step that’s not taken from the beginning. Instead, we settle for conventional wisdom about the reasons behind actions that seem to fit the puzzle. The costs of this laziness can be grave, particularly when the problems are the kind for which society seeks remedies in order to prevent their recurrence. With an initial misdiagnosis, the wrong treatment will most assuredly follow. Had policy makers paid attention sooner to the pulls and tugs likely to face Arafat as well as the Israelis subsequent to Oslo, perhaps they could have managed circumstances better and might have avoided many of the setbacks between the Israelis and Palestinians since 1993.

  DON’T JUST LOOK WHERE THE LAMPPOST SHINES

  The failure to zoom in on what the issues are, who the players are, what their incentives are, and how to fix those incentives is not limited to problems in foreign affairs. The business world suffers at least as much from the same difficulties. To take an example of this from recent years, let’s look at a model that colleagues and I developed to identify the causes of and solutions to corporate fraud.

  Since the Enron debacle, Congress has made a real effort to strengthen corporate incentives to report honestly by introducing a massive amount of new regulation through the Sarbanes-Oxley bill, passed in 2002. However, I’m afraid Sarbanes-Oxley touches on but does not nail the root causes of fraud, at least not as those causes are seen by the model my colleagues and I developed. As we can see in figure 5.1, fraud litigation is once again on the rise, despite the passage of Sarbanes-Oxley. Since the recession began in 2007, fraud has skyrocketed ahead of its pre-downturn 2006 level. This is just a bit of evidence—there is more to come—for my model-based belief that the premises that guided Congress in passing Sarbanes-Oxley were misguided. President Obama promises a new wave of regulatory controls to rein in the risk of business fraud and failure. Let’s hope that his administration and the Congress are more attentive to incentives so that they get the regulations right.

  FIG. 5.1. Federal Securities Class Action Litigation, 2002-2008

  To regulate the risk of fraud it is fundamental that we first understand the motives for committing fraud. How does the model that my colleagues and I worked on determine which companies have an incentive to commit fraud and which do not?

  Our game-theory approach is a variant on the study we did to understand how nations are governed (recall dear old Leopold). But pay attention, as the context of the corporate setting provides a most interesting twist. As we saw in our study of Leopold and other heads of state, loyalty to leaders is much weaker in democracies because the competition is over policy ideas rather than the personal enrichment of a few supporters. Much the same might be said for corporations and the survival of corporate executives. From this starting point, we might assume that the outcome in the business world would be the same as that in the political: “autocratic” leadership leads to corruption (fraud), while more “democratic” leadership does not. This is not the case. Perversely, as we will see, the strong loyalty engendered by relatively autocratic corporate styles helps reduce the risk of fraud. To understand why, we have to leave the light of the lamppost and really look behind the scenes at the logic that governs corporate behavior.

  Big firms have millions of shareholders. Yet few of them attend the annual shareholders meeting, and they have little idea of how the business is run or how it might have done if it had been run differently. They dutifully send in their proxy, voting the way the board suggests, or they toss their proxy statement in the recycling bin and do not vote at all. Big blocs of votes are controlled by a small number of institutional investors, senior executives, and directors. They, not the shareholders, decide how to run the company. The more institutional investors and powerful shareholders there are, the more parties to whom the management is accountable. Ring a bell? It’s the challenge presented to any leader: the more “democratic” the system (think of democracy not as an absolute concept but rather as a continuum), the more people to please.

  When things are going well, the incentives for the executives running a company are not incompatible with the shareholders’ interests. Growth in profits is good for the executives and it is good for the shareholders. But sometimes things do not go well. Then the interests of management and shareholders might part ways. Let’s see why.

  In developed equity markets, the fraud model indicates that accounting fraud typically results because management is trying to preserve shareholder value. Don’t get me wrong. The fraud model does not think of executives as altruists who lose sleep trying to think up ways to make shareholders better off. They commit fraud to protect their jobs in the face of poor performance rather than as a result of a desire to defraud investors per se. That means we can use public records to link the likelihood
of fraud to any publicly traded corporation’s reported performance, ownership oversight, and governance-induced incentives to manage the firm truthfully.

  Examining publicly available information taken from SEC filings, my colleagues and I found that the amount paid in dividends to shareholders and salaries to senior management during years of honest reporting and during years immediately preceding fraudulent reporting differ in significant ways. In the fraud years, senior management receive less compensation—you read that right, less compensation, not more—than expected given their corporate governance structure and reported corporate performance. Dividends typically also fall short of expectations. All the while, reported performance and therefore the firm’s growth in market capitalization look healthy, just as they do in honest years. It is important to note that the compensation for senior management may still have increased in such fraud years, and in an absolute sense may still be quite grand, but the critical point is that if things are really going so swimmingly for the company, then compensation should be even higher than it is.