FEATURES:
Age of the Empirical

By John O. McGinnis

Computers and the question of what works

Pundits and politicians alike complain that virulent partisanship and the excessive power of special interests distort modern democracy. As result, it is difficult to elicit the consensus for policies that will promote the public interest. These are not new problems. In the early 1800s, for instance, Federalists and Democratic Republicans clashed sharply and vituperatively, disagreeing on such fundamental issues as whether creating a Bank of the United States was wise. Consensus periods of politics in American history have been few and far between.

But our future politics is more likely to forge consensus than that of the past, because we are on the cusp of a golden age of social science empiricism that will help bring a greater measure of agreement on the consequences of public policy. The richer stream of information generated by empirical discoveries will provide an anchor for good public policy against partisan storms and special-interest disturbances, making it harder for the political process to be manipulated by narrow interests.

Many great social scientists have understood that people are ultimately persuaded more by facts than by abstract theories. That is the reason that Adam Smith filled The Wealth of Nations with a wealth of factual observations to demonstrate the power of his ideas. It remains the case that if supporters of a policy can demonstrate that it leads to greater prosperity, the political battle is often half-won. It is true that facts alone cannot generate values, and thus no empirical evidence by itself can logically mandate support for a specific social policy. Smith’s contemporary, the philosopher David Hume, himself made this clear with his famous “is-ought distinction.” But politically, most people within modern industrial society adhere to a rather narrow range of values, at least in the economic realm. They favor more prosperity, better education and health care, and other such goods that make for a flourishing life. As to these issues, what is debated is which political program will in fact broadly deliver these goods.

Empiricism has particular power in the United States, where a spirit of pragmatism limits the plausible boundaries of political debate. Republicans try to show that tax cuts will stimulate economic growth, while Democrats argue that the resulting deficits will impede it. Republicans argue that, in the long run, such tax cuts will raise the incomes of all. Democrats tend to disagree. Such consequential arguments are key to persuading the vast middle of American politics. For instance, if the facts show that school choice improves test scores overall, it is unlikely that vague moral claims, like unfairness to teachers, will stop advocates of school choice from making substantial political gains.

Fortunately, we are at the dawn of the greatest age of empiricism the world has ever known. The driving force in the rise of empiricism is the accelerating power of information technology, often referred to as Moore’s law. Moore’s law — originated by Gordon Moore, one of the founders of Intel — is the now well-established rule that the number of transistors packed onto an integrated circuit doubles every 18 months. As a result, computer speed and memory have been doubling at approximately the same rates. Such exponential growth will persist for at least another 15 years. Many observers believe that new paradigms will continue the acceleration of computer power in the decades after silicon chip technology is exhausted.

The fruits of Moore’s law will be not only ever fancier gadgets, but also an ever more informed policy calculus. The accelerating power of computers addresses what has always been the Achilles’ heel of empiricism — its need for enormous amounts of data and huge calculating capacity. Pythagoras famously said “the world is built on the power of numbers.” That is the slogan of empiricists as well, but processing these numbers requires huge computer power. First, the social world must be broken down into numbers that can be calculated, and to deal with matters of any social complexity, that means a lot of numbers. To draw any conclusions, these numbers must then be sliced and diced to test hypotheses about particular social claims, such as the assertion that school choice improves test scores.

But now computers allow more and more facts to be collected and recorded in systematic form, making possible more precise measurements of worldly events. In fact, we can imagine that soon electronic agents will sweep the web to collect data for researchers to use. Greater computer power also permits the construction and implementation of ever more complex equations by which investigators try to exclude the confounding factors always present in the messy social world and thus to reveal the true causes of social phenomena. It also permits methods, like repeated sampling, to produce better error estimates, giving researchers greater confidence in their results. One University of Chicago social scientist is said to have taken the entire summer to run a regression on a mainframe computer 40 years ago. Now researchers can run scores of regressions on their laptops in a few hours. As a result, more empirical papers about social science are written every year, and that trend appears to be accelerating as well.

The resulting rich vein of data and stream of studies analyzing those data will, over time, transform our politics. No matter what the machinations in Washington in a particular week, a new empirical study will likely offer the fruits of some investigation, calculating the effects of a flat tax on economic growth, for example, or of extra school spending on school achievement. The profusion of such work can make a very substantial political difference by changing the information mix in which politics plays out.

Political discussions all have a policy landscape that is shaped by our common knowledge, and it is this common knowledge that empiricists are changing through their discoveries. The effects of any one empirical discovery, to be sure, will be incremental, but the long-run effects of diffusing cumulative empirical knowledge will be enormous. As in science, there will often be good-faith disagreements among different researchers, but, as with other sciences, in the long run empiricists will develop a consensus about the effects of a policy, and that consensus will influence the political world. Greater common knowledge no less than technological innovation is a source of long-term prosperity.

Of course, new facts inimical to their causes will not induce interest groups — from teachers unions to the automobile trade associations — to abandon programs that serve their interests at the expense of the public. Nor will the new information change the fact that the public has little incentive to understand the complexities of public policy. Nevertheless, the new information will shift the debate in many areas, little by little, as the results become diffused through elites and the many sources of the modern media. Cumulatively, the information will often force interest groups to give ground.

Two other factors — themselves a product of our ongoing technological revolutions — will amplify the power of empiricism. One is the rise of blogs. Blogs help police and expose false studies with which interest groups and partisans may attempt to counter the empirical work that undermines the factual bases of their positions. Academic experts regularly write for blogs and, unlike reporters, are well suited to subject empirical work to searching scrutiny. Recently, for instance, a group of prominent legal scholars has begun a blog wholly devoted to law and empiricism. Such developments will also force empiricists to be more careful and transparent about the discretionary decisions they make, such as their choices of time periods to include in their investigations, because their colleagues will be able to call them to account for misjudgment or bias more easily.

The second development reinforcing empiricism is the rise of information markets. Information markets are sophisticated betting pools that modern information technology has created by making it much easier to gather bets and keep running tallies of the odds that the bets generate. Already, information markets are getting a lot of attention because of their ability to predict current events. For instance, on the eve of the presidential election of 2004, the information market Tradesports predicted the winner of every state correctly. Companies now routinely use them to anticipate which product line will be successful. In the future, individuals may bet on what the growth rate or tax revenue will be, conditional either on the implementation of a specific tax cut or on its absence. Thus, empirical claims about the effects of tax rates will also be tested by those willing to bet on the predictions that flow from them. As the economist Robin Hanson notes, these markets will provide a more democratic check on the empirical claims of experts, making it harder for expert peer review to insulate claims from contestation.

The declining cost of empiricism also promises to transform our universities, because it provides greater incentives to hire empirical investigators in the social sciences. A hundred years ago, armchair speculation was very cheap compared to empiricism, as the cost of the latter enterprise was high, often prohibitively so, in that the available technology rarely delivered any useful results. Thus, it was rational for universities to hire theorists without much interest in comprehensive and statistical inquiry into the social world. But now that the cost of empiricism has fallen, a cascade of empiricists of all kinds — economists, psychologists, political scientists — is flowing into our universities and think tanks.

This inflow has several immediate effects, all of them beneficial to a free inquiry that will in turn aid the gathering of facts conducive to consensus politics. First, the hiring of empiricists will tend to decrease any ideological discrimination in academia, because empiricists, whatever their personal politics, will tend to be much more open-minded than their theoretically inclined colleagues. They respect facts themselves and professionally must give a hearing to other empiricists, whatever their views and whatever the results of their empirical investigations. They will also be less prone to discriminate, even in hiring theorists in their social science departments, because empiricists are likely to regard almost all theories as contestable.

Second, the large number of empiricists will create a rich market for empirical work and provide greater incentives to get the facts and models absolutely correct. There are now so many investigators and the computer technology is so pervasive that results are constantly checked. Moreover, the multiple data bases collected in such places as the Inter-University Consortium for Political and Social Research provide a ready resource for the quick debunking of outlandish or mistaken claims. Indeed, I have recently attended faculty workshops where members of the audience have begun checking data and doing alternative analysis on their laptops as the speakers are presenting. The empirical culture, no less than computing itself, will be ubiquitous in the coming years.

The exponential growth in the numbers of empiricists helps resolve another debate as old as that of the Greeks. Platonists are sympathetic to rule by experts and elites. Artistotelians are more receptive to democratic rule. As Aristotle himself puts it in the Politics, “For the many . . . may be better than the few good, if not regarded individually but collectively, just as a feast to which many contribute is better than a dinner provided out of the single purse.” Expertise in the form of empiricism straddles these two political poles. It is a more democratic expertise in that it is replicable, transparent, and sharable. Moreover, as discussed above, it is disciplined by the information inputs of blogs and information markets. Thus, expert judgments will no longer be those of the few wise persons who rely on authority to impress their conclusions on society. Instead, they will reflect the collective sentiments of the empirical community. To be sure, this is not the same as the entire democratic society. But the number of empiricists is large enough that it will provide many of the virtues of democratic judgments — the application of many different perspectives to the same problem.

Already, we are beginning to see that the new culture of empiricism is shaping our conversation. We can see both the work of empiricists generally becoming more prominent and the specific empirical work shaping the debate about matters of national importance. On the prominence of empiricists, one has only to note the remarkable celebrity of Steve Levitt of the University of Chicago. His book, Freakonomics, has been on the New York Times best-seller list for more than a year at this writing, and he has been given the prime territory of the New York Times magazine for a weekly column.

The political impact of empiricism is far more pervasive than the efforts of this one prominent and sophisticated scholar. One important point to stress is that many empirical projects do not involve any complex formulas, but rather entail the collecting of data made easier by information technology. An excellent example is the studies that now map the ideologically skewed composition of our universities. From law schools to faculties of arts and sciences, whether measured by party affiliation or campaign contributions, the imbalance between Democrats and Republicans is from five to one to ten to one and occasionally even greater.¹

This kind of work demonstrates how empiricism can have influence without generating its own values. The values being assumed here are those implicit in the operations of the academic community itself, which now defends affirmative action on the grounds that faculty and students of different races enrich the educational experience by leading to “viewpoint diversity.” Accepting that rationale, it would appear that diversity of political viewpoint would be at least as important as diversity of ethnicity to many subjects, including law and most of the social sciences. As a result of the publicity now given to ideological imbalance, administrators have been forced to try to explain how their ideologically monolithic faculties are consistent with their purported educational ideals.

Far more important to the future of America is the battle over substantive issues. Empiricism has already created a consensus in some areas, crime being one of the foremost among them. In the 1960s and 70s, the political community was very divided about the basic approach to crime. Some experts then attacked prison sentences as destructive rather than useful and even contended that certain criminal acts, like writing graffiti, should not be considered crimes at all.

But criminologists have turned to empiricism to show the truth of claims previously contested. It turns out that that prison sentences sharply reduce crime, as does the greater presence of police on the streets. Steve Levitt recognized that prison-overcrowding litigation created a natural experiment for testing the extent to which prison sentences reduced crime because the litigation released prisoners for reasons that had nothing to do with crime rates. By looking at the consequences of the prison litigation, he was able to show that for each criminal released from the prison population, the number of crimes will increase by about 15. Jonathan Klick and Alex Tabarrock used a similar natural experiment by using Homeland Security alerts to demonstrate the effectiveness of police. Homeland Security alerts increase the number of police in a given area but are otherwise unrelated to crime rates. Their paper suggests that a 50 percent increase in the number of police reduces crime by approximately 15 percent.

The consensus on crime and punishment not only has created agreement on sentencing policies and police practices in the United States, but also has radiated outward both to related fields and to other nations. It is frequently remarked that the area of constitutional law in which the conservative and liberal blocs show most agreement is criminal law and procedure. Given the effectiveness of prison, liberals are much less likely than they were in the 60s and 70s to insist on expanding rights that will reduce the convictions that lead to prison. In Great Britain, the cross-party consensus is similar. New Labour’s popular slogan is “tough on crime, tough on the causes of crime.”

What empiricism has done for criminological research in the past decade it will do for educational research in the next decade. Scholars are focusing on the factors that make for better educational outputs. The marquee issue is whether greater choice of school will improve outputs. An important study here is by Paul Peterson and William Howell, who look at the test case of school choice programs provided by private philanthropists. Because more students applied to those programs than could be accommodated, Peterson and Howell were able to compare the samples of pupils who went to private schools with those left behind in public schools. They found statistically significant improvement among African Americans who attended the private schools, although the improvement among other groups generally fell short of statistical significance.

Such a study is only the beginning of an empirical conversation. First, as with any scientific study, confidence in its results will grow if they are replicated elsewhere. Moreover, this study raises other questions, such as whether the results are scalable. If African Americans go to private schools in greater number, will the results be reproduced, or was the effect largely the result of bringing relatively few African Americans into a better-performing school culture? Second, will public schools be worse as a result of losing such students, or will they improve because of the greater competition? Researchers are now beginning to study these specific issues. In the meantime, Peterson’s and Howell’s work certainly provides reasons to create more experimental voucher programs, like the one the federal government is initiating with children displaced by Hurricane Katrina. Only through larger-scale experiments are we likely to be able to answer the key questions about the beneficence of vouchers.

But though such serious empirical research on school structure has been welcomed, less serious studies come under withering fire from the new and old media. The New York Times provided a front-page story about a study by the American Federation of Teachers that argued the Education Department was covering up the failure of public charter schools — a failure the aft claimed was apparent from its own data. The aft noted that these data showed that students at charter schools performed worse than those at ordinary public schools. Unfortunately for the aft, however, these raw discrepancies proved nothing because, unlike the work of Peterson and Howell, their researchers did not control for selection bias — the fact that those who chose charter schools were not a random selection of the student population. Charter schools have often arisen in poor communities with students who are not doing well; thus, even if they improve the outputs of students, these students may still underperform compared to the average student. Blogs and even some of the mainstream media continuously contested the aft analysis for weeks after it was published.

But school choice is not the only matter about educational outputs that is subject to rigorous study. Educational empiricists are looking at the effect of school funding and class size on school performance. As important, they are considering how different kinds of educational programs affect different students so that better programs can be designed at the micro as well as macro levels. Within a decade, we will know a great deal more about what matters in primary and secondary education. These results will influence the shape of education for the next decade until even better data and analysis refine the hypotheses still further.

Some might counter that entrenched forces arrayed against change, like teachers unions or ideologues of some particular stripe, may prevent this work from having transformative influence. This seems unlikely. To be sure, interest groups will resist even reforms with consensus empirical support, but they cannot prevent the empirical results from seeping out from academics through the new media to the large bloc of genuinely independent voters and a not insubstantial number of conscientious politicians. As some jurisdictions adopt better policies and get better results, they will generate more information for change elsewhere.

If empiricism is good for consensus and bad for special-interest and partisan special pleaders, a culture of better-checked empiricism is even better. The most notable victim of fact-checking in recent years was Emory historian Michael Bellesiles, who won the prestigious Bancroft Prize in American history with a book claiming that gun ownership was not widespread in colonial America. That claim obviously pleased partisans on one side of the debate about gun control. But James Lindgren, a leading empiricist at my own law school, found that some of his key statistical claims appeared to be based on fabricated evidence. After an investigation by his own university, Bellesiles resigned from his tenured position, and for the first time in its history the Bancroft Prize was withdrawn.

The growing spontaneous order of empiricism is the answer to those who worry that political movements or interest groups can take advantage of the empirical trend by manipulating facts to their advantage. Marxism styled itself scientific, and Marx and Engels did try to support their theories with data. But in their time, the strong culture of replicating the database and calculations underlying theories did not exist. Nor were there powerful statistical tests designed to shed light on social claims from multiple angles. In today’s culture, the factual claims of Marx and Engels would not have withstood the light of empirical scrutiny.

One of the advantages of the empirical approach is that it is not necessarily allied to the narrow view of rational man that dominates many social sciences today. If individuals act for reasons that seem irrational, like valuing something more simply because they own it, empirical methods can test the effects of that behavior on society. Nor is empiricism blind to the importance of values. Indeed, much interesting work has been done on the way religious organizations provide mutual help and raise the incomes of their members. Thus, empiricism is able to cash out the frequent claim of pundits and politicians that values matter; and fortunately, unlike them, it has a neutral metric for establishing which values have what particular effects.

Similarly, in the long run empiricism may be able to narrow the division between the advocates of liberty and the advocates of equity, because empiricists can investigate the extent to which there is a tradeoff between these two great social imperatives. For instance, empirical work has already shown that redistribution policies may create dependence and entrench an underclass, harming equity as well as growth. Empiricists are also considering the long-term effect of tax cuts on both growth and equity.

Of course, empiricism is not going to forge consensus on all issues. Empricists may show that abortions reduce crime or that abortion waiting periods reduce suicides, but these secondary consequences are not going to persuade the great majority of people that abortion is right or that waiting periods should be encouraged. Views about the intrinsic evil of abortion or the beneficence of abortion rights are likely to swamp all empirical considerations. But abortion is an unusual issue in this respect, because it divides citizens on the metaphysical question of when life begins, a subject on which many have strong views but no mechanism for demonstrating to their opponents that they are right. Most issues of our politics are happily not like abortion, where disagreement is fundamental and compromise difficult to imagine.

One might wonder whether the greater consensus on most issues — particularly economic matters and other areas in which differences revolve around measurable consequences — may actually increase the salience of intractable moral issues like abortion. Thus, more agreement on more policies may actually generate the political equivalent of the narcissism of small differences, in which individuals overlook the agreements they have to focus even more virulently on the disagreements that remain.

But this danger should not be exaggerated. Empricism has been increasing consensus for some time, yet differences on abortion and other such matters do not appear more virulent than in times past. Moreover, as Adam Smith observed, when religions compete, they have a tendency to become more moderate. Religion in the United States sustains the most important organizations that emphasize the role of values in social life; and as long as they must compete, the rise of consensus politics will intensify, not diminish, that competition, as religious sects will have to offer incentives to attract the many who are influenced by the empiricism all around them. Moreover, groups divided over basic issues of economic and social policy often exploit religious fault lines and other divisions over values to pursue their own partisan agendas. A more substantial policy consensus fostered by empiricism should diminish the impulse for such mischief-making and thus temper clashes over values.

If empiricism is likely to provide a source of consensus and constraint on interest groups in the future, policies that facilitate empirical work will also foster such worthy goals. The first imperative is to make data transparent and accessible, because statistical investigations need data above all. Unfortunately, the government often wants to conceal its operations and hide its failures. But with the rise of empiricism, there are even greater reasons to force what government does in the domestic arena (outside of sensitive national security matters) into the open. The daily activities of government and information about the work of its employees should be posted on the web in machine-readable form. Agencies would do well to follow the lead of Securities and Exchange Commission Chairman Chris Cox, who has begun an initiative to both tag electronically the data sec receives and make sure that its future data are displayed in the most readily usable form. If necessary, special agencies should be created to monitor the dissemination and accuracy of such information.

Second, the rise of empiricism makes appropriate decentralization of government even more useful. One of the most powerful investigative tools in social science is to compare how different laws work in different states and nations. Such careful comparisons make manifest the consequences of good and bad policies. But that kind of investigation can work only if policies differ from place to place, and politicians in governments everywhere have an inherent tendency to centralize power because they can exercise more control. There are, of course, many reasons to resist this trend. Decentralization creates a market for governance by allowing different jurisdictions to compete to attract people and investment. It also permits the formulation of diverse policies that meet peoples’ diverse preferences.

To all these virtues the rise of empiricism adds another: Decentralization facilitates the empirical investigation of the differing consequences of social policy. We can all learn from the effects of the flat taxes sweeping the Baltic states, but only if the European Union resists calls for tax harmonization. Those interested in fostering consensus and restraining interest groups should be even more insistent on the need for appropriate federalism at home and more resistant to international organizations that will force nations to acquiesce in international mandates about domestic matters.

Another policy imperative for advancing empiricism is to sustain the free flow of information that conveys research in useable form to the public. It goes without saying that any regulation of the blogs should be rejected. Unfortunately, some are already suggesting that blog postings should be considered in some circumstances a contribution to candidates and subject to regulation under the McCain-Feingold act. But it is precisely near elections that those with empirical data and expertise are most needed to critique the policies and platforms of candidates.

More generally, campaign finance regulation of all kinds is a threat to the dynamic of increasing the public’s awareness of new empirical conclusions that should help bring consensus. Although information technology has the potential to improve our knowledge of the social world and policies that will improve it, this same technology makes it harder to get the public to pay attention. When thousands of people were willing to attend the Lincoln-Douglas debates, it was the best show in town. But now politicians compete with hundreds of cable television stations featuring people far more attractive than politicians doing things most citizens find far more interesting.

In this new media landscape, we need more, not less money in politics to keep people informed. Of course, the information will be delivered in sound bites that most intellectuals find aesthetically unappealing, but the alternative to 30-second political ads is not a seminar on politics but a beer commercial or sitcom. And since the beginning of democracy, ideas originally generated by intellectuals have always been transmitted to the body public through simplifications and slogans. Such diffusion and general reception has nevertheless taken place over time and overall has been a force for social progress.

If empiricism helps support a consensus politics today, it is likely to be even more effective in the future. We have already reached the threshold where computers can retain billions of pieces of data and make millions of calculations per second. But computer capacity will continue to increase exponentially, making it scores of times more powerful in the next decade. We can expect the amount of data available to grow exponentially as well, as ever more electronic agents sift and collect data. We may even expect there to be computer programs to assess which empirical conclusions are the best when different empirical analyses conflict. Such computer judgments will have the virtue of even greater impartiality. In short, if the fruits of empiricism are now transforming the political information mix, that mix should become only richer in the coming years.

This accelerating power of empiricism does not guarantee enduring agreement on all aspects of the social policy and political structure. I am not here offering a fact-driven version of Francis Fukuyama’s End of History — his famous theory that the dialectic of ideological debate will end in agreement on free markets and liberal democracy. Interest groups, future technological discoveries, and forces no one can now predict may create counter pressures for division of the polity. But the growing importance of factual investigation will generate a continuing tailwind for politics based on consensus and facts rather than special interests and special pleading.

¹ This is the one area in which I have personally contributed to the empirical literature. See John O. McGinnis, Benjamin Tisdell, and Matthew A. Schwartz, "The Patterns and Implications of Political Contributions by Elite Law School Faculty," Georgetown Law Journal 93:4 (April 2005).