The Politics of Precision Journalism
Technological advances and the social systems for dealing with them do not develop at the same pace. When old social and cultural systems are applied to new ways of doing things, the fit is sometimes awkward and even painful. The reaction to that pain is a common theme in history. In England, early in the nineteenth century, the Luddites tried to stop the economic dislocations resulting from labor-saving machinery by destroying the machinery. At about the same time, the Erie Canal Commission passed up a chance to build a railroad, for which the technology was then available, to construct a waterway, whose concept was more comfortably familiar.1 It should be no surprise to find that the same cultural lag that afflicted the industrial revolution is also present in the information age. Precision journalism will take some getting used to -- both by the practitioners of journalism and by its customers.
In its early applications to coverage of the political and social movements of the 1960s and 1970s -- civil rights, the anti-Vietnam war movement, and the youth counterculture -- no paradigm shift seemed called for. The application of social and behavioral science research methods was simply the extension of journalism by other means. "The ground rules are no different from those on which we've always operated," I admonished my fellow journalists in 1973. "Find the facts, tell what they mean, and do it without wasting time. If there are new tools to enable us to perform this task with greater power, accuracy, and insight, then we should make the most of them."2
description suggests that precision journalism involves little more than
the maintenance of the journalist's traditional role with only a modest
quantitative improvement in speed and accuracy. When a quantitative change
reaches a certain magnitude, however, it becomes a qualitative change.
Some current objections to precision journalism are based on the assumption
that such a change in magnitude is taking place or is likely to take place
in the future. These objections are beginning to lead to proposals to regulate
the practice of precision journalism ñ either
directly by law, or indirectly by pressuring the media to adopt voluntary
self-restraint. The efforts by some state legislatures to ban exit polls
are one example. Another is found in the voluntary withholding of exit-poll
information by the networks until the voting is finished in the jurisdiction
in which the poll was taken. Pressure for other forms of restraint may
The notion that a journalist should be concerned with a possible invasion of privacy may seem surprising and self-contradictory. Invasion of privacy is almost part of the journalist's job description. However, there are precedents for observing self-restraint in some circumstances. The codes of some of the major professional associations in journalism recognize a duty to provide protection of privacy. In the utilitarian ethical systems used, consciously or not, by most journalists, the right to privacy is easily overridden by a more pressing concern for the public's right to know. The question for precision journalism is whether the power of its methods adds a moral burden that did not exist for less powerful methods.
Elliot Jaspin and Maria Miro Johnson's exposure of the criminal pasts of Rhode Island school bus drivers is an example.3 That the named individuals were school bus drivers and that they had criminal records were both matters of public information. At the Milwaukee Journal, James Rowen linked public records of drunk driving convictions with pilots' licenses for a story on aviation safety.4 Without the computer, neither reporter would ever have made the connection. The quantitative change in the amount of time and effort to search and link such records has led to a qualitative shift in the things that journalists can discover.
Is this extension of journalism by other means still journalism, or is it something new and different that ought to be regulated? The regulators have their eye on us. While the exercise of freedom of the press cannot be regulated directly, barriers can be placed between the journalist and the sources of information. A current theory of regulation that has yet to be fully tested holds that a public document is public only if it is on a piece of paper that can be read by a person. If the same information is in a medium that can be read by a machine, it is not, this theory holds, a public document -- or at least not the same kind of public document and subject to the same kinds of rules.
The theory is novel and not generally accepted. Lawyers and judges like to reason by analogy, and the analogy between a paper record and an electronic database is easy to understand. But some government agencies have begun to claim a substantive difference between the two kinds of records. When the Boston Globe asked for computerized Treasury Department records for a story on money laundering, the department coughed up the records, but laid down some ground rules. It asked for a detailed statement of the information sought and how it would be used. It demanded an agreement that the computer programs for analyzing the information would not produce the identities of individuals or businesses.
"In the event that a search of the database results in the inadvertent disclosure of personal identifiers," the Treasury regulations say, the user must "terminate the search until appropriate security measures can be implemented; relinquish all records of personal identifiers to Treasury officials; and make no further disclosure of the information."5
Democratic theory holds that the public's information ought to be available equally to all members of the public. That is fairly easy to do with paper records that anyone can walk in and inspect. But complicated computer records take special equipment and special skill that are not generally available to the ordinary citizen. The issue of who should pay the cost of access has not been resolved. Some government agencies charge special fees for the use or copying of computerized public records, and these fees could easily be manipulated to create barriers to the freedom of information.
The problem is less severe when a reporter or inquisitive citizen wants a specific piece of information and can describe it in some detail: a real estate transaction, a birth certificate, or a list of campaign contributors, for example. It is usually pretty clear that the government has a duty to provide the information and pay the expenses of providing it or charge a nominal copying fee. But should the government support a fishing expedition? Is merging two databases and browsing through them to see what turns up an activity that should be subject to the same rules of access? The Privacy Act of 1974, one of the last projects of the late Senator Sam Ervin, puts barriers in the way of government agencies that would like to swap and match data. A Reagan administration proposal to move the Bureau of the Census from the Commerce Department to the Treasury Department was defeated on the ground that the agency that controls income tax records should not also keep census records. On the other hand, a resourceful journalist can sometimes get government record keepers to provide active assistance in matching records. The Milwaukee Journal, for example, asked the Wisconsin Department of Transportation to match its computerized driving records against a list of licensed pilots supplied by the Federal Aviation Administration. The state officials obligingly supplied reporter James Rowen with the names of 302 Wisconsin airplane pilots who had been convicted of driving motor vehicles under the influence of alcohol. Seventeen of them were licensed to fly passenger airliners.6
future barriers are erected against such searches may very well depend
on the self-imposed restraint of journalists who know how to deal with
computerized public records. Identifying drunks who fly airliners or drive
school buses serves a clear public purpose, and
few are likely to argue that the right of privacy of the pilots or bus
drivers overrides the welfare of the passengers they serve. But if journalists
use the computer to reveal embarrassing private facts just to show off
their technical virtuosity and without any clear public benefit, a regulatory
backlash could result. As a minimum, journalists should impose on themselves
the same moral standards and the same restraints that would be observed
for information gathered by other means. If we are to argue that an electronic
record has the same legal standing as a paper record, we should treat that
record with at least the same sensitivity.
Defining public opinion
Precision journalism is at its best when it is sorting out the conflicts among special interest groups, measuring their support, estimating their potential for having an effect. When it does this, it departs from the referendum model of public opinion, whose only virtue is that it is easy to understand.
By the referendum model I mean a view of public opinion that holds that policy makers ought to be guided by whatever the majority thinks on any given issue. In the referendum model, every citizen's vote counts exactly the same. In the real world, that is not so, not even for those few issues that are actually decided by referendums. Even in those cases decisions are made by those who pay the cost -- in time and energy consumed -- of voting. Nothing in this country is decided by a representative sample of the public.
Why, then, do we go to so much trouble to acquire and interview a representative sample of the public? Because it is there, perhaps. And because we are intuitively comfortable with the fairness of the referendum, majority-decides model. George Gallup, the pioneer American pollster, is also partly to blame, because he promoted the habit of thinking of a poll as a continuous referendum by which majority will could be made known.7
But observers of American politics since Alexis de Tocqueville have known that pure majority rule is neither possible nor perhaps even desirable. Tocqueville worried that pure popular control would prevent the wisest and best from using their gifts for the public good. He welcomed the formation of interest groups as ways of concentrating the power of minorities to create "a necessary guarantee against the tyranny of the majority."8
As originally conceived during the social protest movements of the 1960s, precision journalism was a way to expand the tool kit of the reporter to make topics that were previously inaccessible, or only crudely accessible, subject to journalistic scrutiny. It was especially useful in giving a hearing to minority and dissident groups that were struggling for representation.
government in the United States has always involved tension among competing
factions, as James Madison, the fourth president and one of the authors
of the Constitution of 1787, had foreseen.9
But the referendum model is much too simple for the complexities of representative
government in a large and conflict-ridden society. The majority wants conflicting
things. The composition of the majority shifts from one issue to the next.10
Some voters feel so strongly about single issues that they will yield on
almost everything else to get their way on that one issue. Such trading
is called logrolling, and it takes place quite visibly among elected representatives,
but you can see its origins in the work of single-issue pressure groups.
Like most modern democracies, we are governed by temporary coalitions.
The process of forming those coalitions, always less formal and more difficult
to follow than in European parliamentary democracies, is worth following
but it demands a special kind of public opinion polling, including a recognition
that not all opinions are equal.
People who vote are an interesting and deviant subset of the general population. Except in national elections, they are usually a minority. Measuring their attitudes and behavior is especially relevant and especially challenging. This effort has been one of the most popular applications of precision journalism.
For many journalists, the effort to measure and even predict electoral behavior has been motivated by simple competitiveness. The most interesting fact about an election is who wins. If you can find out ahead of time, it is news by definition. The Literary Digest demonstrated the news value of election predictions in a series of surveys from 1916 through 1936. At just the moment that their technology failed, George Gallup demonstrated that it could be perfected. He performed a polling hat trick. Not only did he predict that Franklin Roosevelt would win the 1936 election in a landslide, he also predicted that the Literary Digest poll would show the opposite. He even had an accurate estimate of the percentage that the Digest poll would give Alf Landon. (He did it by checking a small sample from the same lists of telephone and automobile users that the Digest used to recruit its two million respondents.) That established Gallup's reputation and made polling credible. Since then the technology has improved to the point where a national poll is considered a failure if it misses the election outcome by more than two percentage points. In the 1988 presidential election, none did.11
When I was a Washington correspondent in the 1960s, poll watchers focused chiefly on the two nationally syndicated newspaper polls, those of George Gallup and of Louis Harris. Interviewers talked to citizens in person in their homes, mailed the results to the home office, and the results of each poll were analyzed and reported in leisurely fashion over the course of several weeks.
technology, and a faster-paced presidential selection process changed that.
In 1988 there were ten national polling organizations in the field and,
on the average, a new poll was reported every second day during the final
months of the campaign. Neither the public nor the press corps was accustomed
to this density of polling data. Because the presidential selection process
was itself in an era of change, some associated the undesirable and worrisome
aspects of change with the polls and began to blame the polls.
A list of complaints
1. There are too many polls. "Every newspaper and television station thinks it has to have its own poll," complained columnist Jack Germond. "So we keep quantifying the obvious."12 That, at least, is a self-limiting problem. Polls are produced in response to a free-market demand for information. When they merely restate the obvious, demand will decline.
2. The polls are not accurate enough. This complaint is typically heard in the early part of a presidential campaign when different polls published within a few days of one another give seemingly different results. Surely, it is assumed, they could not all be accurate.
Much of this problem, if not all, is due to the fact that polling organizations reserve their best and most expensive methodology for their later measurements, the ones that will be compared to the election results to provide an evaluation of the poll. Early polls leave more room for error for a variety of methodological reasons, including smaller samples and less rigorous methods for identifying likely voters. In a nation where only about half the voting-age population participates in a presidential election, identifying those who will vote is both important and difficult.
Differences among polls are also exaggerated by the journalistic convention of reporting the point spread between two candidates rather than the size of the leader's majority. For example, the final Gallup poll in the 1988 presidential election was reported as a twelve-point lead for George Bush while the final Harris survey gave Bush a four-point lead. Surely, it seemed, one poll or both had to be seriously in error. They were not. Gallup gave Bush 56 percent of the two-party vote, and Harris gave him 52. Bush actually got 54 percent, so both polls were accurate to two percentage points, well within normal sampling error. The other national polls that reported in the week of the election were even closer.
Another complaint is related to the one of inaccuracy:
3. The polls are often wrongly interpreted. This, too, seems to be a self-solving problem. Twenty years ago, I wrote an article on misinterpretations of polls by journalists for Columbia Journalism Review, and a rereading shows that the media have made considerable progress in understanding and interpreting election polls.13 Some common assumptions then, that a poll in one city or county is generalizable to the entire nation, or that a poll in August predicts voter decisions in November, are no longer so widely held. Errors are still made today, but they are more subtle: for example, the failure to distinguish between a poll of voting age adults and one of likely voters. The proliferation of polls will, of course, force journalists to become more proficient at interpreting them. The information marketplace will demand it. But that will only aggravate the next complaint on the list:
4. Polls are too accurate. The logic behind this complaint is that polls were relatively harmless when they were wrong much of the time, because nobody took them seriously. But when they are right, the public believes them and responds to them, and this possibly affects voting behavior. The logic parallels that of seventeenth-century English libel law, where it was held that the greater the truth in a defamatory statement, the greater the tort.
This complaint is related to the one that is probably discussed the most:
5. The polls affect the outcome of the election. Until recently, the polling fraternity tended to dismiss polling effects on the election process itself as either nonexistent or negligible.14 As recently as 1980, Albert E. Gollin was able to write that such concerns "have faded away for lack of supporting evidence."15 No more. As research has become more sophisticated, that position has become difficult to sustain, particularly as researchers look at indirect effects through political contributors, campaign volunteers, and endorsers.16 Early polls are a special problem. Such polls are name-recognition tests rather than predictors of who will win. Unsophisticated consumers of polling data do not always recognize this limitation, and that can make it difficult for a little-known challenger to attract the backing needed to mount a serious challenge against a well-known incumbent. Potential backers, mistaking the name-recognition data for an election prediction, are frightened off. That error contributed to the defeat of Hubert H. Humphrey by Richard M. Nixon in 1968. Humphrey lagged in the early polls and failed to get the financial backing that he needed to get an advance commitment of late-campaign television time. Because that election was so close, a late media push could have made the difference.
Campaign contributors still make that kind of mistake. "A poll of 47 percent to 30 percent looks good for the incumbent and makes it hard for the challenger to raise money," says USA Today pollster Gordon S. Black. "In truth, we know from research that an incumbent with less than 50 percent of the vote is in deep trouble. . . ."17 The democratic solution, of course, is not to curtail the polls but to teach those who use them to make decisions to do so rationally.
Where direct effects on voters are concerned, the effects are often so slight as to lack statistical significance, but that does not mean they lack substantive importance. In a close election, there is no such thing as a negligible effect. The presidential elections of 1960 and 1968 could have been tipped the other way by any number of normally inconsequential factors.
The complaint of election effects begs the question of whether such effects are good or bad for democracy. Even the pollsters trying to defend themselves often start with the assumption that such effects are bad. But another argument needs to be considered. In today's media-driven elections, the information provided by polls ñ from the primary polls to the early calls on election night may actually help the process.
That this sounds like a radical notion is a measure of how much we have forgotten the roots of democratic theory.
The sanctity of the ballot has an emotional, almost mystical, quality in the United States, as though democracy were a religion rather than a practical way of letting the people govern. We are a nation of immigrants, and for those who came from nondemocratic or less democratic countries, the right to vote symbolizes the difference between tyranny and popular control. We tend to forget that the secret ballot is an import from Australia and a fairly recent development in our history, and that its original purpose was not to protect the privacy of the citizen but to make it more difficult for corrupt politicians to bribe him. A citizen whose vote is unknown is less likely to be paid for voting a certain way, because he has no way of proving that he carried out his side of the contract. The earliest postcolonial elections in the United States were carried out in full public view. In Virginia, for example, the voter entered the courtroom, stood before the sheriff, and announced his vote. The candidates were even present to thank him personally in some cases. If not, their representatives were there to record the vote.18
in the New England town meeting, the archetypal model of democracy in America,
was done in full public view. And voting on the record is standard practice
in legislative bodies. Tactical voting, where one withholds one's vote
until he or she sees how other members are voting and makes a decision
based on that information, is an accepted practice in legislative bodies,
and it makes perfect sense.
The building of consensus
A representative democracy in a large and complex nation faces the problem of forming a working majority from many small and diverse segments of interests. There are never enough resources for everyone to get everything he or she wants. The problem of pulling together enough of the conflicting goals to make the best and most coherent policy is one that has haunted philosophers from the dawn of recorded human thought. Economists call it optimization. Moral philosophers who worry about it are called utilitarians. Political scientists call it consensus formation. In every field where the problem is considered, information is crucial to a solution. This need for information is especially important in politics, where the main tool of consensus formation is coalition building. To build a coalition you have to know what each group wants and how badly it wants it, i.e., what it will give up to get it.
In European parliamentary democracies the process of coalition building comes mainly after the election as members of specialized parties representing minorities of various sizes strike their bargains with one another to form a government. In the United States, with its two-party, winner-take-all system, the coalition building has to take place before the election. This structural difference accounts at least in part for the historically lighter voting participation in the United States than in Europe. If the two major political parties have done their job properly, the process of consensus formation will have brought them so close together by election time that the voters will be relatively indifferent to the final choice.
But the major parties in America have not been doing that job as deftly as in the past. The brokering that took place at national party nominating conventions through a series of ballots edging ever closer to a working majority was a fairly efficient method of coalition building. But the last multiballot convention was in 1952 when the Democrats nominated Adlai Stevenson on the third try. The flow of information through the mass media has become so efficient that all the players know who has what bargaining chips and what the outcome will be before they arrive in the convention city.
The role of the media became even greater after the party reforms of 1970 took power away from the party professionals and gave it to the rank and file. This change was effected mainly through rules that encouraged the states to select their delegates through primary elections. One effect of this so-called reform was that candidate selection fell to nonprofessionals who were more interested in promoting their narrow ideological goals than in finding a candidate who could capture the broad center of the spectrum where the likelihood of winning was greatest. George McGovern (who headed the commission that wrote the new rules for the Democratic party) in 1972 and Ronald Reagan in 1980 were examples of the new noncentrist candidates.
building was never as easy in the American system as it is in parliamentary
systems, and the weakening of the political parties has made it more difficult
still. And it places a heavy burden on the mass media of communication,
which must now provide rank-and-file voters with information that previously
only the party professionals needed. Isn't it reasonable to suppose that
giving the voters accurate information about each other, about the relative
voting strength and the preferences of different groups in the electorate,
might help, not hurt?
Take the case of the early primaries, for example. Assume that you are a Republican and your first choice is Pierre DuPont and your second choice is Bob Dole and your least favored choice is George Bush. You need poll information to make a rational choice. If the polls show that Bush is ahead, that Dole has a chance of beating him, and DuPont is dead last in a field of six, your vote will count the most if you go for your second choice, Dole. Why should you not have the benefit of that information?
Here is another kind of example. In 1980, the candidates were Ronald Reagan, Jimmy Carter, and a dissident moderate, break-away Republican, John Anderson. A centrist voter might prefer Carter as long as he seemed to have a chance of winning. Given information that Carter would lose, this voter might want to cast a ballot for Anderson to protest the capture of the Republican party by the right-wing Reaganites.
Consider yet another possibility. The American system works most efficiently when the president and Congress are in the hands of the same party. A rational voter who was aware of this fact, and considered it important, might reasonably use polling information about the probable winner of the presidency to decide how to vote for a congressional candidate.
Finally, there is the matter of turnout. A voter can make rational use of information about whether or not an election will be close to decide whether or not the possibility of affecting the outcome is enough to justify the effort to vote. Not everyone wants to be rational about this. My California cousins express resentment when the networks tell them the election is over before they have had a chance to vote. They say it makes them feel powerless. I, on the other hand, think they should be grateful because they have more information on which to act than do I, who must vote blindly from the east coast. Some day, I tell them, there will be an election so close that Dan Rather will look them in the eye and tell them that the outcome depends on the voters on the western seaboard. My western cousins can then go charging out of their hot tubs and proceed to the polls feeling powerful indeed.19
Can polls really be used to provide tactical guidance to voters in this way, or is this just empty theory? Let's look at the extreme case, early calls by the networks on election night. The best opportunity for studying the results comes in the extreme cases because they leave less room for subtlety. The extreme case is an early call of a landslide in an election that was expected to be close. And the best example is the 1980 election when President Jimmy Carter was defending his office against challenger Ronald Reagan. Using exit poll and sample precinct data, NBC announced shortly after 8 p.m. that Reagan had won. President Carter promptly conceded even though polls in many states, from New York to California, were still open.
Among the most thorough studies of the effects of that early call on voter behavior were investigations by political scientists. John E. Jackson of the University of Michigan20 used data from panel surveys, and Michael X. Delli Carpini of Rutgers University21 looked at aggregate voting data at the congressional district level. Both found evidence that participation declined among voters potentially affected by the knowledge that Carter had conceded. In other words, the response was rational.
In addition, Delli Carpini found intriguing evidence that the voters were using the information rationally along the more subtle lines that I have just mentioned. In districts where polls were open after NBC's call, there was a decline in the Democratic vote and a 2 percent surge in the vote for John Anderson. And there was nearly a 5-percent swing toward the Republican candidates for Congress. Is there a measurable segment of voters really sophisticated enough to cast a protest vote for Anderson when they know that Carter will lose ñ or to vote for a Republican congressman because they expect a Republican to be in the White House? The Republican congressional boost, Delli Carpini found, came from the districts with higher levels of income and education ñ exactly from the people that one would expect to make such calculations. Jackson's individual-level data contained a consistent, though statistically insignificant, relationship: Republicans were more likely to be deterred from voting by hearing about the early call or Carter's concessions than were Democrats. Republicans tend to be higher in socioeconomic status than Democrats.
This rational use of polling data by voters will still be resented by politicians who believe they are damaged by it. Candidates for state and local office hoping for a coattail effect or a high turnout can justly say that the effect is harmful to them. But these effects are situation-dependent. They are not amenable to being turned into ideological issues. And complaints beg the question of whether the harm is unjust. Nothing in democratic theory says that elections should be decided by voters who care so little about the outcome that they need some external motivation to get to the polls.
On the whole, a case can be made that the rational use of information about what other voters are doing or may do helps the democratic process. Voters have always had this kind of information to act on. What is new is that polling makes the information less often wrong. When the information is surprising, as was the case in 1980 because the election was expected to be close, the effect is more dramatic. But drama is not inherently evil. And one surely cannot argue that democracy is served when voters are acting on bad information, which would have been the case without NBC's early call. Over the long run, there will be elections where the situation is the reverse of 1980, elections that everyone expects to be landslides which are in fact close ñ as was the case in the Humphrey-Nixon election of 1968. If exit polls had been in place then, turnout could have gotten a boost from voters motivated by the knowledge that they had an extraordinary opportunity to make a difference.
rational-voter theory does not take care of all the problems. There may
still be voters who will use the information irrationally. But that's not
against the law in a democracy. No doubt there were voters who opposed
Michael Dukakis because of his Greek ancestry or who voted against George
Bush because he came from Texas. I have not heard anyone argue that information
about a candidate's origins should be suppressed because voters might use
it irrationally or unfairly. A democracy that tries to protect its voters
from information that they might use irrationally ceases to be a democracy.
The voting decision cannot be forced into a sterile, information-free environment,
nor should it be.22
Uneasiness in the media
The mass media organizations that sponsor the polls are uneasy in their new role, but they have no choice except to fulfill it, just as the rank-and-file voters must accept new and heavier responsibility. When the Miami Herald reported the details of Gary Hart's sexual escapades, it was responding to this new need. Voters were given a view of a candidate that in a previous era would have been confined to the smoke-filled rooms where the power brokers negotiated. Because the primary election system has shifted the decision out of those rooms and placed it directly in the hands of the voters, the media have a greatly increased duty to give those voters the information with which to exercise that power.
The "bandwagon effect" is a term that originally was applied to delegates in a national nominating convention where the goal of a delegate was to get on the winning side in time to make a difference and place the winner under some obligation to the delegate. It is a useful phenomenon, helpful in coalition building, and no less useful when the arena becomes the nation at large instead of the convention hall. And yet the media persist in feeling guilty about contributing to a bandwagon effect, even when the contribution is in the form of accurate information.
This guilt is especially intense where quantitative data are involved. The election process demands information, and the media cannot and should not avoid supplying information which has the potential to affect the outcome. But when the information is about the voters themselves, a certain circularity is introduced into the process that makes media people uneasy. The additional fact that polling involves numbers gives it some special mystique, making it seem more like a "pseudo-event" than would be the case if the investigation had been initiated by a news medium using purely qualitative methods.23 Moreover, the use of numbers and scientific method gives polling information more credibility than information from conventional reporting modes. This enhanced credibility, paradoxically, increases the cost to society when the information is wrong. And so, in what amounts to intellectual self-mutilation, some organizations go out of their way to cloak their numbers in ambiguity. They do this by ignoring pollsters' attempts to produce a number that can be compared directly with the election outcome (and thereby sacrificing the one direct test of the poll's validity) and by clinging to undecided and leaning voters as if their existence denoted a major uncertainty over the outcome.
When the New York Times reported the final Gallup poll result in November of 1988, for example, it deleted the Gallup Organization's basic prediction of a Bush win with 56 percent of the two-party vote and reported only the less refined numbers, suppressing Gallup's allocation of the undecided portion.24 The Timesevidently did not think its readers were up to handling that knowledge. USA Today used Gordon Black's straightforward prediction of 55 percent for Bush in a page-one graphic display, but failed to mention it in the accompanying story, which emphasized leaners and the undecided.25 The three network polls attempted no allocation of the undecided. Anyone wanting to compare their polls to the election outcome had to do so after the fact by dropping the undecided from the percentage base. And some newspapers have even become leery of incorporating leaners into their final figures. (Leaners, as you will recall from the previous chapter, are persons who give a choice only when asked a follow-up question, e.g., "As of today, which way are you leaning?") One explanation of this bizarre behavior is that the media want to avoid the obvious validity test and leave the leaners and undecided voters positioned as a cover for their polls' possible mistakes. A more subtle explanation is also plausible: the complaints against polls have made the media managers feel guilty about their own precision, and so they seek to conceal that precision.
This guilt may in turn be based on the violation of one of journalism's norms: that the media should observe and report with detachment and not participate. Publishing polls may seem too dangerously close to participation. This stance bespeaks a certain ignorance of how democracy works in the United States. Every decision on what to print and what not to print is, after all, a form of participation.
Democracy was never meant to operate in a sterile, information-free environment. And yet one media pollster wrote in an article published just after the November 1988 election that consensus formation is "the real danger in media polling."26 This view is widely held, and it shows how poorly some of us Americans appreciate the subtleties of our own system and the theory behind it. Consensus formation is the aim of the election process. If precision journalism, in the form of preelection and exit polls, helps the electorate to communicate with itself and bring about consensus, then there is hope for the brave new world of direct democracy that mass communication technology is trying to bring us. Precision journalism is part of that technology.
precision journalism can wholeheartedly embrace the openness of scientific
method, its potential dangers and abuses will be self-correcting. A journalism
based on scientific method leaves a trail where error can be detected and
truth verified. Nowhere is that as true as in the case of election polls.
The public will trust polls as much as they deserve to be trusted, no more
and no less. The comparison of polls with election results is a wonderful
way for the public to judge, and journalists should not create barriers
to inhibit or cloud that judgment. If the accuracy of other forms of journalism
could be put to such a test, the marketplace of ideas would reward the
purveyors of truth relentlessly and efficiently. Journalists should welcome
the chance to be put to such a test.
1. Roger Burlingame, March of the Iron Men (New York: Grosset & Dunlap, 1938), p. 239. return to text
2. Philip Meyer, Precision Journalism (Bloomington: Indiana University Press, 1973), p. 15. return to text
3. Elliot G. Jaspin and Maria Miro Johnson, "R.I., system fails to fully check driving records of bus applicants," Providence Sunday Journal, March 1, 1987, p. 1. return to text
4. James Rowen, "3% of pilots have record of driving while drunk," Milwaukee Journal, February 12, 1989, p. 1. return to text
5. Cory Dean, "Computer Use for News Raises Legal Questions," New York Times, September 29, 1986, p. A12. return to text
6. Rowen, "3% of pilots have record of driving while drunk." return to text
7. George Gallup, The Sophisticated Poll Watcher’s Guide (Princeton: Princeton Opinion Press, 1972). return to text
8. Alexis de Tocqueville, Democracy in America (New York: Schocken Books, 1961) p. 192. (Originally published 1835.) return to text
9. James Madison, Federalist No. 10, The Federalist (New York: Putnam, 1888. return to text
10. For an excellent discussion of the difficulties of the majoritarian model, see James M. Buchanan and Gordon Tullock, The Calculus of Consent: Logical Foundations of Constitutional Democracy (Ann Arbor: University of Michigan Press, 1962). return to text
11. George Bush won the election with 54 percent of the two-party vote. The percentages given him by the various polls (based on decided voters in the network polls plus allocated undecided voters in the rest) were: USA Today (Gordon Black) 55; CBS, 55; ABC, 55; NBC, 53; Harris, 52; Gallup, 56. return to text
12. Quoted in Michael Traugott, "Marketing the Presidency: Is There a Tyranny of Media Polls?" Gannett Center Journal, Fall 1988, p. 60. return to text
13. Philip Meyer, "Truth in Polling," Columbia Journalism Review, Summer 1968. return to text
14. See for example the chapter on bandwagon effects in Gallup’s 1972 book, op. cit. return to text
15. Albert E. Gollin, "Exploring the Liaison Between Polling and the Press," Public Opinion Quarterly, 44:4 (Winter 1980), 445-461. return to text
16. Richard L. Henshel and William Johnston, "The Emergence of Bandwagon Effects: A Theory," The Sociological Quarterly, 28:4 (1987), 493-511. return to text
17. Gordon S. Black, personal communication, January 16, 1989. return to text
18. Neil Spitzer, "The First Election," The Atlantic, November 1988, p. 20. return to text
19. There are such voters. My cousin Marillyn Cozine of Danville, California, advises me that she cast a tactical vote for John Anderson in 1980 on the basis of the network projections. return to text
20. John E. Jackson, "Election Night Reporting and Voter Turnout," American Journal of Political Science, 27 (1983), 615-635. return to text
21. Michael X. Delli Carpini, "Scooping the Voters? The Consequences of the Networks’ Early Call of the 1980 Presidential Race," Journal of Politics, 7 (February/March 1984), 48-50. return to text
22. For an excellent exposition of this argument in a Spanish context, see Jose Ignacio Wert, "Uses and Misuses of Survey Polls in the Media," paper presented at the IAMCR-WAPOR joint session on Mass Media and Public Opinion, Barcelona, July 1988. return to text
23. Identification of the poll as pseudo-event was popularized by Daniel Boorstin in 1961 when both polls and their use by news media were less sophisticated than they are today. For a useful discussion of the concept, see Gollin, op. cit., p. 449. return to text
24. "Bush and Dukakis Travel to West for Final Jousts of Campaign," New York Times, November 7, 1988, p. B14. return to text
25. USA Today, November 7, 1988, p. 1. The editor who supervised production of the graphic and the editor who had final review of the story had different views on what should be reported. return to text
26. Michael Traugott, "Marketing the Presidency," p. 64. return to text