Chapter 7
Do Editors Matter?
Once while attending a newspaper marketing seminar in Durango, Colo., I took a side trip to visit the nearby Mesa Verde cliff dwellings. They were built by a long-vanished people who were part of the migration from Asia, across the Bering land bridge, down to South America. This migration, historians believe, was so gradual that no person who was a part of it realized there was a migration at all.
For decades, the decline of newspaper readership had the same ethereal quality. While total newspaper circulation as a proportion of households was clearly in decline from the 1920s, most people at the turn of the century still read at least one newspaper on most days. The loss was in readers of more than one paper, and it was caused by the weeding out of duplicate circulation in multinewspaper markets.
Even with that obvious explanation, the decline in household penetration was enough of a public relations problem to make newspaper publishers look for a different measure. The Newspaper Advertising Bureau, an arm of the old American Newspaper Publishers Association, stepped in to find a solution. The result was elegant. It helped advertisers focus on readership rather than circulation with the “read yesterday” measure. Developed by Leo Bogart with endorsement from the Advertising Research Foundation, it was a careful and conservative measure that corrected for the tendency of survey respondents to over-report their reading behavior because of its social desirability.
The trick was to ask a series of questions about specific newspapers that the respondent had read or looked into in the previous week. That introductory question vented the social desirability bias. Then, with the fact that the person was a reader established, he or she was asked about each paper claimed: “When was the last time before today that you read or looked into the (name of paper).” If the person volunteered “yesterday,” he or she was counted as a reader. By spacing the interviews appropriately through the week, researchers could convert “read yesterday” into a valid and reliable measure of average daily readership.
Bogart's pathbreaking survey was released in 1961. It showed that on a typical weekday, 80 percent of adults read a newspaper. This “near-universality of readership” became “the basic theme of every sales presentation delivered to advertisers.”1
The pitch was so effective that “read yesterday” became the gold standard of newspaper readership measurement. The Simmons Market Research Bureau, which did audience measurement for TV and magazines, added the newspaper measure to its annual surveys. But then something unexpected happened. The number began to fall.
For Bogart, the trouble started in 1971 when the Bureau's own survey showed read-yesterday readership three points below the most recent Simmons number of 78 percent. Methodological error was suspected. Respondents had not been asked about distant newspapers with local household penetration of less than 5 percent, leaving out people who might, for example, read The Wall Street Journal and nothing else. After much review and argument, the Bureau cranked in a point-and-a-half correction to represent the national papers.
From there, Bogart recalled much later, “we could easily round the number to 77 percent. . . . [T]his reduced the discrepancy between our survey and SMRB's 78 percent to a single percentage point—a relatively innocuous difference.”
The presentation to advertisers, which used to claim newspaper readership by four out of five Americans, was modified to say “nearly four out of five.” Yet, recalled Bogart, “while I went about the country enthusiastically putting on this presentation . . . I was uneasily conscious that the percentage was really closer to three out of four than to four out of five.”
Not long after that, my then employer, Knight Ridder, started detailing me to an occasional readership study along with my reporting duties in the Washington Bureau. When the American Society of Newspaper Editors was in town, a gathering of Knight Ridder editors was organized, and I was asked to tell what the minuscule decline in readership might mean.
“If it happens once,” I said, “it could be a sampling fluke and nothing to worry about. If it happens twice, we should start to worry. If it happens a third time, it's an earthquake.”
James K. Batten, then editor of the Charlotte Observer, recalled that observation years later when he was Knight Ridder vice president for news and on his way to becoming CEO. “It was an earthquake,” he said.
The story of the newspaper industry's response has been told in detail by Leo Bogart in his 1991 memoir, Preserving the Press. The Newspaper Readership Project, the pooled effort of a number of newspaper trade associations, operated from 1977 to 1983 on several fronts. They included advertising and promotion of newspapers, circulation improvement, and audience research.
From an economic point of view, the most effective response was the simplest: compensate for fewer readers by raising prices to advertisers. From 1975 to 1990, publishers pushed advertising rates up by 253 percent, even though newsprint prices were up by only 161 percent and the Consumer Price Index gained 141 percent.2 Charging more for delivering less is essentially a liquidation and harvesting strategy, but this was not the conscious goal. There is no evidence that the owners had given up on the business and were hoping to cash in and get out.3 Instead, publishers and investors congratulated one another on newspapers' pricing flexibility. Like the migrants from Asia to South America, they couldn't see the movement toward the brink because it was so slow.
The editors of newspapers adopted a more alarmist view. They tended to take the readership decline personally and blamed themselves for not providing sufficiently compelling content. Readership studies became a growth industry for a time. Manipulating content costs little or nothing, and so trying to halt the readership decline by tailoring the editorial product to readers' wants seemed particularly attractive. Leo Bogart suspected otherwise.
Bogart was hoping to find the editorial formula for success, when, as I described in Chapter 4, he surveyed editors on their definition of a quality newspaper in 1977. Then he compared the responses from editors of successful newspapers with those of editors whose papers were slipping, expecting to find the secret of success. But the winning editors and the losing editors, to his surprise, gave the same answers!
“[E]ditors of successful and unsuccessful newspapers seemed to be operating by identical editorial philosophies,” Bogart recalled later. “The inevitable conclusion seemed to be that the forces that made a newspaper lose circulation were largely independent of its content. Success or failure had more to do with pricing, distribution, and population changes in the cities where papers published than with the character of the editorial mix or the operating practices or theories of individual editors.”4
The editors were not ready to hear this. Rather than rejoicing that the readership decline was not their fault, they attacked Bogart's survey and his conclusions. Any outcome was better than facing the possibility that they were powerless.
This development had a profound effect on my career. Throughout the 1970s, I had become a roving precision journalist, traveling from my base in the Washington Bureau to help papers in the Knight Ridder group apply social science research methods to their local news stories, mostly dealing with race, poverty, and opposition to the war in Vietnam. These efforts were well received, and top management at the company decided that the same methods could be applied to the readership problem. That's how I found myself detailed to the occasional marketing study. By 1976, these assignments were sufficiently frequent that I moved out of the National Press Building to an office in Reston, Va., near my home and, more importantly, my mainframe computer supplier. Data were still entered on punched cards, and analysis was done with mainframes in those days. I began the transition from Harvard Data-Text, an excellent higher-level computer language for the obsolescent IBM 7090 series, to SPSS, which ran on a great variety of newer machines.
Two years later, I left reporting altogether and was posted to Miami to become Knight Ridder's first director of news and circulation research. My mission was a parochial version of Bogart's. While he was trying to halt or reverse the readership decline for the newspaper industry in general, I was attempting to do that job for Knight Ridder in particular. A number of entrepreneurial research firms jumped into the fray.
The idea of newspapers doing market research at all was controversial. Some journalists regarded it as pandering to lowbrow reader tastes. One of the entrepreneurs, Ruth Clark, helped fuel the controversy with a series of focus groups undertaken for ASNE that provided support for a softer approach to the news. Information from focus groups, of course, is not generalizable, and her report began with the obligatory disclaimer. As soon as that was out of the way, she started to generalize. Newspapers, she told the editors in 1979, should be “more caring, more warmly human, less anonymous.” To compete with television, her argument went, newspapers needed to be more entertaining.5
Five years later, she ran a larger study, funded by United Press International, which included survey research that was generalizable, and she reversed her course. Now reader interests had moved from entertainment to news. If she believed that her original focus groups were wrong, she never acknowledged it. Readers, she implied, had changed in that short time.
We were all failures. The readership decline proceeded in straight-line fashion, with only a hint of leveling off in the early 1980s as the baby boomers passed the age of thirty and began acquiring the community ties that are associated with newspaper readership.6 The Readership Project ran out of funding and shut down. I moved to academe and compiled my work in The Newspa per Survival Book: An Editor's Guide to Marketing Research. Bogart summed his effort in a far more comprehensive volume, Press and Public: Who Reads What, When, Where and Why in American Newspapers.7
After the Readership Project
That pretty much left it up to the academy to develop new measures and ideas. Bogart's work did not go unnoticed, and two good researchers at Michigan State University built on his 1977 survey to try to challenge the dismal notion that what editors do doesn't matter.
Bogart's sense that editors were off the hook, you will remember, was based on his survey asking them to define quality in news. The basis of his no-effects conclusion was that editors of successful and not-so-successful newspapers expressed the same news values.
But there are some other things to consider. Stephen Lacy and Frederic Fico realized that editors' values might not be a sufficient measure of what they actually do. The capacity of individual editors to act will vary. Two kinds of capacity are involved here. One, placed under intense scrutiny by Rick Edmonds for the Poynter Institute, is based on the resources that an editor is provided by his or her publisher and can be measured by size of staff, newsroom budget, library resources, training effort, and the like.8
Another element of capacity is neither financial nor physical, but intellectual. If two editors have the same news values as measured by Bogart, and the same human and physical capacity as measured by Edmonds, the smarter of the two might be able to convert those resources into better and more effective content than his or her less-talented peer. I call this the “Bellows effect” after Jim Bellows, the editor who made a career of moving from one failing newspaper to another, propping each one up with ephemeral brilliance.9
Lacy and Fico tackled both problems simultaneously with refreshing directness. They assembled a sample of 114 newspapers for their constructed week (seven different days of the week, not sequential, to avoid being overly affected by a single unusual news event). Then they adapted the following measures based on Bogart's editor survey.10
Ratio of staff-written to wire service and feature copy.
Amount of non-advertising content (news sections only)
Ratio of interpretation and backgrounders to spot news reports.
Ratio of illustrations to text.
Number of wire services carried.
Length of stories in news sections
High ratio of non-advertising to advertising content (news sections only)
Three measures considered more important by editors in the survey were given a greater weight than lesser ones.
Fico and Lacy built an additive index from these variables, using standardized scores (meaning that each variable was expressed in terms of its own deviation from the mean). The next thing they needed to proceed with the execution of Bogart's original intent was to find a measure of newspaper success.
They chose circulation as reported by the Editor & Publisher Yearbooks of 1985 and 1986. The Lacy-Fico constructed week for sampling quality was in November of 1984. Circulation reports lag the audits by several months to a year, so the presumed cause and effect were close together in time.
Taking circulation as the sign of success requires, of course, a correction for the size of the market, and this creates another problem, defining the market. Newspapers' own definitions of their markets are highly idiosyncratic, tailored to delivery constraints, commuting patterns, and retail trade zones. The researchers in this case chose city population as an indicator of market size, which could be a reasonable surrogate for actual market size in most cases, especially in the smaller markets. (In larger places, the main newspaper sometimes does better outside the central city than it does inside, owing to the deterioration of downtown, a problem that Bogart detected early on.)
A straightforward way of correcting for market size would have been to work with household penetration, which is circulation divided by households, and is easy to interpret. A newspaper with 50 percent penetration in a given area has circulation equal to half the number of households. (Market researchers prefer households to population because one copy of a newspaper is typically shared by all the members of a household, regardless of the number of people there.)
Instead, they used a quite different method of adjustment, using city population and newspaper quality to predict circulation in a regression model. Multiple regression is a convenient way of estimating the effects of different factors on your ultimate object of interest. Lacy and Fico reported that with this procedure, which tended to level the playing field for circulation size, news quality explained about 22 percent of the remaining variation in circulation. In short, better newspapers sell more copies.
That would be wonderful news except for the problem that haunts us throughout this book. There is no very good way to tell whether greater circulation is the cause or the effect of higher quality. Bigger circulation brings in more revenue, leading to economies of scale that can free up more resources for the newsroom. And, of course, causation could run both ways: as a virtuous cycle if higher circulation leads to still more quality and as a vicious cycle if both are slipping. One way to sort this out is by introducing time as a variable.
Even if the quality measure by Lacy and Fico had clearly been made earlier in time than the circulation measure, there would still be the problem that newspaper circulation does not change much from year to year. Slow change means that circulation at Time 1 will always correlate with circulation at Time 2. One needs measures of both quality and circulation at clearly different times—with enough lag to produce an effect—plus a form of analysis that adjusts for the autocorrelation, i.e. the tendency of a variable measured at different points in time to correlate with itself.
In their recommendations for further research, Lacy and Fico said that an update of Bogart's 1977 work would be useful. Koang-Hyub Kim and I took them up on it and ran a partial replication in 2003. It was a partial replication because we used only members of ASNE while Bogart also sampled members of the American Press Managing Editors. Many editors belong to both organizations, and ASNE was able to provide significant logistic help, so we gratefully concentrated on its membership. Using a mixed mail and Internet methodology, we received returns from just over 50 percent.
In his search for readily applicable yardsticks, Bogart had given editors a list of twenty-three criteria and asked them to rate the importance of each on a scale of +3 to -3. Then he averaged the scores and ranked the items. Focusing on the 15 items that editors had ranked the highest for Bogart (and adding some of our own) we came up with relative scores that were very close to those obtained fifteen years earlier. The basic values of editors had not changed.11
Among those that were used in both 1977 and 2002, the same top ten emerged, although in different order. Here they are:
One of the values promoted by the Hutchins commission in 1947, to go beyond mere reporting of facts and provide interpretation that would yield “the truth about the facts,” appears to have shifted as the importance of interpretations and backgrounders fell from third to eighth place. On the other hand, diversity of political columnists gained somewhat in importance as did number of staff features, and both of these can be construed as bows to the Hutchins recommendation. The values of editors are remarkably stable.
To reduce all the data from our survey to a smaller number of manageable concepts, Kim and I applied factor analysis, the tool developed for psychological testing decades ago. We were seeking clusters of question items that correlate with another, an indication that all are measuring the same underlying factor. There were five such clusters. In order of importance ascribed to them by editors, they were:
1. Localism
2. Interpretation
3. Editorial vigor
4. Quantity of news
5. Ease of use
Bogart had avoided asking about qualities that he considered too subjective to measure. Accuracy and literary quality were among these. (Accuracy was dealt with in Chapter 5, and literary quality, interpreted as readability, was considered in Chapter 6. In this analysis, the Flesch scores from Chapter 6 will be used as the indicator of ease of use.)
Following the example of Lacy and Fico, I then looked for relationships between each of these five factors and business success. Instead of using circulation as the indicator of success, I chose household penetration in the home county as my static measure and the robustness of that penetration—its staying power over time—as the dynamic measure.
This strategy is fine in theory, but shares the same limitation that Lacy and Fico faced: the content measures come from a single brief period in time: November 1984 in their case, and April 2002 in mine. This is fine if we assume that newspaper culture so inhibited change over time that the brief snapshot of content was an adequate indicator of how a paper was edited over a longer period. Their content sample was a full week, including Sunday. I stuck to weekdays. Since I collected circulation penetration data for 1995, 2000, and 2003, a number of possible effects can be sought. But because of the limitations of the content measurement, we should expect some frustration and consider this research more for exploration than confirmation.
Localism
Let's start with localism, measured on a sample of thirty-two newspapers. This is a convenience sample, consisting of papers in communities supported by the Knight Foundation communities plus a few others for which data were available. The sample consisted of four days in April 2002: April 11, 12, 16, and 17.
Coders examined every news story in the front and local sections of each newspaper and classified them according to source: staff written or wire.12 A column of briefs was considered a single story and counted fractionally if its individual components came from both kinds of sources. Staff-written stories filed from remote locations were counted as local on the theory that the purpose of sending staff far afield is to obtain reports of remote events that are tailored for the local audience.
The mean across the thirty-two newspapers was a staff-written rate of 46 percent. But the range was wide: from 64 percent at the Detroit Free Press to 31 percent at The Macon Telegraph. Intuitively, one might think that smaller papers, being more focused on their communities, would have a higher percent local, but they didn't.
Perhaps they could not afford it. A fairly large market might be required to support heavy local coverage. The evidence is in the way that localism increases steeply with size of market (defined by home county) up to about 300,000 to 400,000 households. For the markets larger than that, size did not matter. Apparently, critical economies of scale accumulate rapidly at first, then more slowly. They vanish altogether once a market reaches 400,000 households. Below that, the ability of a newspaper to staff up for heavy local coverage was quite dependent on market size.13
The effect was clearest when plotted with market size on the horizontal axis and percent local stories on the vertical. The diminishing effect can be visualized as a straight line by re-expressing market size as its logarithm, which exaggerates effects at the low end of the scale and minimizes them at the high end. (Imagine a plot on a rubber sheet. The log scale stretches it out on the left or low side and compresses it on the right or high side.)
Some of the papers are identified by their cities to help you see the effect of market size.
Before we accept the conclusion that the size effect was simply a matter of papers in larger markets having more resources, let's consider another way to look at it. Papers in smaller communities don't need to go to as much effort to maintain good household penetration. The dynamics, and sometimes the isolation, of a smaller community enhance demand for a newspaper.14 Perhaps smaller communities got by with less local coverage because they did not need it as much to be successful in business.
Whatever the reason, a paradoxical result was presented. As the proportion of local stories in a newspaper increased, household penetration declined. Lest some publisher or investor take this as an excuse to demand thinner local staffing, let me hasten to add that there is no causal relationship here. How do I know that?
We return to the clever statistician's trick used in earlier chapters that lets us take the effect of market size out of the equation. It is partial correlation. When the playing field is leveled in this way, that bothersome negative relationship between household penetration and percent local stories vanishes into the ether. It does not turn positive. It just goes away. Both localism and low penetration are functions of market size. Nothing more.15
For another reading, my student coders classified the same set of stories not by who wrote them, but by the location of the news event. The proportion of all stories from all sources that had a state or local origin was used as another indicator of localism. Nothing interesting turned up. With or without a statistical adjustment for size of the market, the use of local geography instead of staff writing to denote localism was unproductive. The geography-defined localism correlated with neither penetration nor robustness.
But localism is just one indicator of quality. Let's keep looking.
Editorial Vigor
Ralph Thrift, Jr., was a graduate student at Oregon when he invented his index of editorial vigor. This is an additive index, as opposed to a scale, because the presence of one of its factors does not imply the presence of the others. But the more of them that are present, the more vigorous is the editorial. This is obvious on its face, hence it has what social scientists call “face validity.”
In Thrift's conceptualization, the following elements contribute to the vigor of an editorial.
1. Localism. It takes more vigor to discuss local matters that your readers know about.
2. Controversy. Editorials that examine matters on which local citizens disagree are more vigorous than those on which there is consensus.
3. Argumentation. The vigorous editorial writer will pick a side, lay out its case, and let you know where the newspaper stands.
4. Mobilizing information. People who are persuaded by the argument will need to know what they can do about it.
Two teams of graduate students rated a group of newspapers, one in April 2002 and the other in January 2003. They met the usual tests for intercoder reliability, meaning that the definitions were clear enough so that the two judges agreed on the presence or absence of each element almost all of the time.16
Moreover, editorial vigor was stable across the two time periods. Across twenty-nine newspapers measured at both points in time, the vigor ratings were positively and significantly (although not highly) correlated.17
This analysis used the 2002 measures of vigor that covered a larger group of newspapers and the same dates that were used for the localism measure. This updating turned out to be important, because there was a positive and significant correlation between localism—the proportion of stories produced by the newspaper's own staff—and editorial vigor.18 In other words, a newspaper that could assemble a staff with the resources to cover local news could also attract vigorous editorial writers.
A red flag should go up here. Can editorial vigor, too, be just a function of newspaper size?
The paper with the most vigorous editorials was also the largest, The Philadelphia Inquirer. The least vigorous was a smaller paper, the Columbus Ledger-Inquirer. Weighting the five elements of vigor equally and assigning one point to each, Philadelphia editorials averaged two on the four-point scale. The Columbus mean was 0.43. (The mean for all papers measured in 2002 was 1.37).
And the larger markets tended to have the more vigorous editorials, although the effect was not quite linear. As with localism, market size did not matter once it exceeded 400,000 households in the home county. And it mattered most among the smaller markets.
Here's a look at the scatterplot of editorial vigor by market size expressed as the log of number of households in the home county. Some of the newspapers have been identified by city so that you can see the trend. Once again, a log scale is used to keep the papers from clumping together at the low end of the market-size scale.
Another advantage of the scatterplot is that you can see who is doing better or worse than their market size would lead you to expect. Those above the line, such as Myrtle Beach Sun-News and The Philadelphia Inquirer were vigorous for their size. The Columbus Dispatch and The Charlotte Observer, unless we caught them in bad weeks, were relatively vapid. In the competitive Akron-Cleveland market, the Akron Beacon Journal was more vigorous than its larger neighbor, The Cleveland Plain Dealer.
This variance from expectation can make an interesting variable in itself. It's called the residual because it represents left-over variance in editorial vigor that is not explained by the effect of market size. We can think of it as the vigor score adjusted—or “corrected,” if you prefer—for market size.
We can do this correction, but it doesn't help to predict circulation success.
Editorial vigor in its raw form and editorial vigor adjusted for market size yielded somewhat different rankings. Philadelphia fell from first to third place, edged out by Myrtle Beach and Greensboro. Columbus, Ga., remained on the bottom. But in terms of predicting household penetration or penetration robustness, editorial vigor, adjusted or not, was no help.
Illustration and Interpretation
Using a subset of the sample, the same twenty-two newspapers that were in the accuracy survey (Chapter 5), my assistants measured the proportion of stories with illustrations and counted the proportion that were more interpretive than straight reporting. The editors had considered both of these factors important, but neither correlated with anything interesting, including circulation success.
If editors were actively managing ease of use, there ought to have been a correlation between number of illustrations and the readability scores. There wasn't.
However, in this small sample, there was a relationship between illustration and interpretation. Newspapers with lots of interpretive stories had more illustrations.
Quantity of News
Other things being equal, a good newspaper should contain more news than a bad one. This is the most intuitive of the quality indicators. For openers, let's look at the relative size of the news hole, expressed as the percent of the paper's printable area occupied by news.
Percent news correlated positively with household penetration.19 For those of us who want to make the case that news is good business, it would be tempting to stop right there. But we can't.
On further inspection, it becomes apparent that market size correlates negatively with percent news and the effect is strong.20 Bigger markets have lower ratios of news space to ad space. As they take advantage of economies of scale, they use it to increase the amount of advertising more than the amount of news.
As already noted, the larger markets tended to have lower household penetration. Partial out the market-size effect, and the positive relationship between percent news and penetration disappears into the mist. There is some room for argument here. Maybe the smaller relative news hole was the reason large markets had lower household penetration, and the problem could have been fixed with a larger investment in news. But the intuitive path of causation is that market size independently affects both penetration (big markets are harder to penetrate) and relative size of the news hole (big markets have a smaller percent news). The lower penetration in larger markets is due to their greater distribution problems and a less cohesive public sphere caused by greater population diversity. Larger markets also have more distractions in the form of alternate news outlets. I am inclined to give priority to market size in this causal chain because it is a relatively sticky variable, not as subject to change as news hole or penetration. So my tentative call is that the seeming effect of news hole on penetration is only an illusion.21
Want to prove me wrong? Here's something to try. Segment a metropolitan market into tight zones for both news and advertising as Geoff Dougherty has done in Chicago with his online newspaper, Chi-Town Daily News. Team professional editors with citizen journalists to create a series of cohesive public spheres. Create influence in each of these spheres and reinforce it with the metropolitan influence of the mother paper. Do that in enough markets, and we might find some natural experiments with convincing evidence that the disadvantages of size can be made to go away.
On to absolute size of the news hole. As with other quality indicators, this one is strongly related to market size and forms a nice logarithmic curve with a break at about 400,000 households.22
As with relative news hole, we are in danger of being fooled by the data unless we partial out the effect of market size on the absolute news hole. When we do, the result is discouraging. Size of the news hole, adjusted for market size, had no visible effect on penetration, robustness, circulation, or readership. The correlation was positive in most cases, but it never approached statistical significance.
Before giving up on content, let's try one more thing. Following the good example of Lacy and Fico, I combined several individual measures of quality, including localism and editorial vigor, into a simple additive index. Maybe their collective power will tell us something that individual measures can't convey.
Lacy and Fico corrected for market size by including city population in a regression equation with their index. I used a different strategy, correcting each item individually and using number of households in the home county as the indicator of market size.23 It shouldn't matter very much if the underlying theory—that quality causes circulation success—is strong.
Like Lacy and Fico, I used standardized scores for my index.24
So what is the effect of this composite measure on circulation, penetration, and robustness?
None. No matter how you slice it, the effect just isn't there.
Most of the correlations were negative, and all were so close to zero that it didn't matter. None came close to statistical significance.
Why this failure to confirm Lacy and Fico's perfectly reasonable and intuitively satisfying finding that quality is associated with better circulation performance?
A major difference between their larger and more carefully selected sample and mine is in the initial correlation between quality and market size. For Lacy and Fico, market size explained less than half the variance in quality, which left some room for other factors to show an effect. In my case, using the log of households in the home county as the size indicator, and home county circulation penetration as the test of business success, size explained most of the variance in the circulation indicators. When I adjusted for market size, there wasn't much variation left for the quality indicators to explain. Putting it another way, the newspapers all looked pretty much alike once the variation for market size was taken out, which makes it more difficult for a correlation procedure to predict anything.
What happened? By focusing solely on home counties, I might have missed the places where quality has the most measurable effect. Perhaps it matters most in the more marginal areas. By focusing on weekdays, I might have missed an important quality effect in the Sunday papers. This is not welcome news. It would have been so much better to find that quality has a robust, across-the-board effect that shows up wherever and however you look. We are less interested in effects that need a lot of excuses for not showing up.
Let's consider another possibility. Maybe the world changed.
Lacy and Fico were working with content measures from November 1984 and circulation figures from about that same time.25 Their sample was carefully drawn to represent a variety of competitive situations: monopoly, including dual ownership; joint agency; and competitive newspapers. My convenience sample, with the Knight Foundation communities at its core, included some joint agencies (Detroit) and dual ownerships (Philadelphia) but no newspapers with intra-city competition. There were hardly any left.
It's true that in some metropolitan areas papers in different counties competed at the margins: Miami-Fort Lauderdale, Raleigh-Durham, Akron-Cleveland, to name just a few. And this competition was worth a study in itself, because it was in those border-warfare areas that the effects of quality were most likely to appear in a way that could be measured.
But that kind of competition is sissy stuff compared to what was happening back in the early 1980s. Then there were a number of cases of newspapers struggling for dominance in their market, and using news-editorial quality as a weapon of destruction. In The Philadelphia Inquirer newsroom, war metaphors were common as editors plotted the demise of the Bulletin, which finally fell in 1982. Where such battles went on, advertisers and readers alike gravitated toward the winning paper, as the loser spiraled down in quality and circulation.
So it's possible that there were more inferior, failing newspapers in November 1984, enabling Lacy and Fico to tap a more diverse and interesting sample than was available in 2002. Add the fact that the Knight Foundation communities were heavily weighted toward ownership by a pretty good, centrally-managed group, Knight Ridder, and we had a situation where the quality differences across newspapers were minimized. To show that quality has an effect, we need papers that vary in quality. Those in my sample did vary, but perhaps not enough. Perhaps quality has to deviate a lot—for better or for worse—to make a measurable difference.
If true, this requirement would be bad news for editors hoping to make a difference with small, low-cost changes in quality.
Or, and this is an important possibility to consider, perhaps Bogart was right in 1977 when he tried to tell ASNE members that they were off the hook. Maybe the readership decline was not then, and never has been, the fault of the editors.
People in newsrooms—and I have been one—tend to overestimate the effects of their work. The notion of media effects as swift and powerful was nurtured by such exceptional events as William Randolph Hearst promoting the Spanish-American War and the famous 1938 Orson Welles broadcast of “War of the Worlds.” But the first systematic research in the 1940s, by Paul Lazarsfeld and others on the effect of media on voting behavior, led to the opposite theory, that media effects are minimal.26
The recent consensus is somewhere between. Media effects exist, but they are subtle. They cultivate cultural change, they set political agendas, but they act slowly and quietly. Like the slow-moving migrants from the Bering land bridge, we're not aware of the creeping consequences.
And so it could be with the effect of quality news and editorial content on the business of newspapers. Tracking it over a long period of time, we could tease out the effects and even put a dollar value to them. But one-shot measures, this one included, are not adequate. For vindication of the belief that editorial talent and social responsibility yield profitability, we must look beyond cross-sectional studies of daily content and turn to process measures.
In the next chapter, we'll look at a key player in the process, the copy editor, the last line of defense in newspaper quality control.
1. Leo Bogart, Preserving the Press: How Daily Newspapers Mobilized to Keep Their Readers (New York: Columbia University Press, 1991), 36.
2. Bogart, Preserving the Press, 53.
3. Michael E. Porter, Competitive Advantage (New York: The Free Press, 1985), 310.
4. Bogart, Preserving the Press, 108.
5. Ibid., 142.
6. The leveling is visible in the General Social Survey. See the chart in Chapter 1.
7. Meyer, The Newspaper Survival Book: An Editor's Guide to Marketing Research (Bloomington: Indiana University Press, 1985); Bogart, Press and Public: Who Reads What, When, Where and Why in American Newspapers, 2nd ed. (Mahwah, N.J.: Lawrence Erlbaum Associates, 1989).
8. Edmonds, “Measuring News Capacity: A First Cut at an Indicator,” www.poynter.org/content (posted May 21, 2002; retrieved June 5, 2003).
9. Bellows, The Last Editor (Kansas City, Mo.: Andrews McMeel, 2002).
10. Lacy and Fico, “The Link Between Newspaper Content Quality and Circulation,” Newspaper Research Journal 12:2 (Spring 1991): 46–57.
11. “Quantifying Newspaper Quality,” presented to AEJMC, Kansas City, Mo., July 30, 2003.
12. My thanks to Owen Covington and David Freeman who performed this work in the summer of 2002.
13. The effect is best illustrated by a log transformation. The correlation between percent local and the log of the number of households is .629 (p = .0001). Using the raw number of households, correlation is significant but less impressive: .427 (p =.015).
14. This, too, is an effect that becomes less pronounced as community size increases. The correlation between household penetration and the log of households is -.656, p = .00005.
15. With log of households as control, r = -.098, p = .60.
16. All measures attained a minimum value of .7 on the Scott's pi test.
17. The correlation was .464, p = .026.
18. Correlation = .396, p = .021, N = 34.
19. Correlation = .369, p = .038, N = 32.
20. Correlation = -.568, p = .001, N = 32.
21. Controlling for log of households, the correlation between percent news and penetration is -.0471, p = .802.
22. Correlation with log of households = .907, p = .0000000000008 (about one in a trillion), N = 32.
23. The correction was made by regressing each quality indicator against market size and saving the residuals.
24. Each value is expressed in standard deviation units above or below the mean for all cases. For example, for a given newspaper, a standardized score of one indicates that its vigor is one standard deviation above the mean for all newspapers in the study.
25. Lacy and Fico cited the 1985 and 1986 Editor and Publisher Yearbooks as their source of circulation data. Because of lags in the ABC reporting process, the published numbers usually represent circulation from the previous year or earlier. The study was updated in Charles St. Cyr, Stephen Lacy, and Susana Guzman, “Circulation Increases Follow Investment in Newsrooms,” Newspaper Research Journal 26 (Fall 2005): 50–60.
26. Lazarsfeld, Bernard Berelson, and Hazel Gaudet, The People's Choice: How the Vot er Makes up His Mind in a Presidential Campaign 3rd ed., (New York: Columbia University Press, 1968).