Deconstructing Pop Culture

Deconstructing Pop Culture header image

Are Newspapers Dead?

August 28th, 2006 by David Kronemyer · No Comments

I. INTRODUCTION
One of the most vexing questions facing the media industry is how “old school” newspapers and television stations will adopt to the new digital era. Are they enmeshed in their historic past, like the woolly mammoths in the La Brea tar pits, or will they be able to transition successfully into a new paradigm? Incisive commentators such as Katharine Q. Seelye of the New York Times and Diane Mermigas of the Hollywood Reporter observe, for example, that “The jury is still out on whether any of the publishing scions, including Tribune Co., Dow Jones, the New York Times, E.W. Scripps and Gannett, can make a successful transition into an electronic information age with their editorial legacies intact,” Mermigas, D., “From papers, local TV to new-media morass,” Hollywood Reporter (Aug. 22-28, 2006). And: “The migration of readers and advertisers to the Internet, as well as rising costs and falling revenue, are threatening the financial well-being — even the very existence — of some of the industry’s most storied brand names,” Seelye, K., “What-Ifs of a Media Eclipse,” New York Times (Aug. 27, 2006).

In the meanwhile, gargantuan transactions continue to reshape the industry. Only a few months ago, the shareholders of Knight Ridder, Inc. voted to sell the company to McClatchy Co., which immediately turned around and announced plans to spin-off a dozen or so of its acquisitions to separate investor groups. Tribune Co. is roiled with dissension as former shareholders of the Times Mirror Co., which the Tribune Co. bought in 2000, question the company’s direction and future prospects.

All of this activity suggests at least somebody perceives old-media assets have continuing value. To address this issue – partially, at least – that is, if there is any value – I analyzed historical financial results from companies owning newspapers with current revenues in excess of $1 billion/year. These companies are: Gannett Co., Inc. (“GCI”); Tribune Co. (“TRB”); Washington Post Co. (“WPO”); New York Times Co. (“NYT”); E. W. Scripps Co. (“SSP”); Dow Jones Co., Inc. (“DJ”); Belo Corp. (“BLC”); McClatchy Co. (“MNI”); and Lee Enterprises, Inc. (“LEE”). For good measure, I also threw in both Knight Ridder, Inc. (“KRI”) and Times Mirror Co. (“TMC”), since they certainly met and would continue to meet this criterion, if they still were around. This exercise resulted in some interesting and possibly surprising conclusions.

II. METHODOLOGY
I examined each company’s 10-K for each year from 1980 to 2006. 10-K’s from 1992 on are on-line at the SEC’s Edgar website. 10-K’s for prior years came from my personal library. For several companies, a complete data set was not available. For example, both SSP and MNI went public in 1988, and, as a result, there is no publicly-available information before 1985 (for SSP) or 1983 (for MNI). TRB went public in 1983, however, its prospectus sets forth previous-period financial information. And, as I noted previously, TMC drops out following 1999.

As you might imagine, this approach presents some significant problems. These include:

1. Inter-company and intra-company differences in segment reporting. Statement of Financial Accounting Standards No. 131, “Disclosures about Segments of an Enterprise and Related Information,” requires companies to report certain information about their operating segments. Unfortunately, it still leaves wide room for interpretation, and is not consistently implemented; and, companies exercise wide latitude to reorganize. For example, beginning with the third quarter of 2004, NYT combined its digital operations with its print businesses to create a “News Media Group.” The company stated: “The aggregation of the Company’s print and digital businesses in this manner reflect the Company’s organizational structure and its business strategy, which emphasizes a multiple-media platform approach pursuing both audiences and advertisers within the markets in which the Company competes.” WPO did the same thing, in 1998.

2. Restatements of prior accounting periods. One of the consequences of segment reorganization is that companies like to go back and restate their operating results. This, of course, reduces the accuracy of period-to-period comparisons. For example, in 1995, GCI changed from three segments, to four segments. Divisional reporting was reorganized, and “certain businesses previously reported in the newspaper segment are now reflected in the other businesses segment. Prior-year segment data has been restated to reflect this reporting change.” GCI slimmed down to two divisions in 1999, with the same consequence.

Some companies change their segment classifications just about every year. One example is SSP, where “Previously reported 1993 and 1992 segment information has been restated to conform with 1994 segment classifications.” Another offender is TMC, which discontinued its broadcast operations in 1993. As a result, it restated all of its segment information for 1991 and 1992 – but provided different numbers for each year. On the other hand, some companies, such as DJ, have remained remarkably consistent in their segment classifications over time.

The Securities & Exchange Commission recently announced it was going to start focusing on companies that buried restatements in a filing of current results. Instead, the company will be required to make a special filing, in order to highlight the prominence of the restatement to investors. Reilly, D., “No More Stealth Restating,” Wall St. Journal (Sep’t 21, 2006).

Because of this unpredictability of restated results, I determined the best course of action would be to examine each 10-K for each company for each year. This involved looking at a lot of 10-Ks. Unfortunately, it simply is not possible to rely on items such as the “10-Year Summary of Supplemental Financial Information” (or five years, or whatever) that many companies gratuitously supply. With the exception of only one or two companies (e.g., WPO), the summary comprises a basically useless compendium of irrelevant information, seemingly deposited in the annual report with the sole objective of confusing the reader. Similarly, web-sites like Yahoo Finance either do not provide the relevant metrics, or, when they do, the numbers typically are unexplainably different than the ones set forth in the 10-K.

3. What is “Operating Income?” It is hard to believe that the accounting profession would sanction different definitions of what should be a straightforward term. Operating income, of course, differs from “net income,” in that it is before income tax, interest debits or credits, and similar items. It therefore is considered to be a much more useful indicator of how the core business itself is doing, as opposed to later adjustments, courtesy of the corporate group.

This being so, I have little confidence that this basic result, which is buried in most annual reports to begin with, is comparable across-the-board. Companies even have a hard time agreeing on the definition within the context of a single financial statement. For example, the 1980 DJ 10-K sets forth two different numbers for operating income, as does the WPO 10-K for the same year.

Despite these caveats, I had to remain content with the information that each 10-K supplied, because it was the only credible source. I did not attempt to psychoanalyze the numbers, or reparse them into different categories. Perhaps this approach can be rationalized on the assumption that, to the extent there are vagaries in the reported information, they equally are likely to pertain to each company.

III. STATISTICAL TESTS
I then prepared a data-base that, for each company, sets forth the following items: newspaper segment operating revenue (“OR”); newspaper segment pre-tax operating income (“PTOI”); total corporate OR (including the newspaper segment); and total corporate PTOI (including the newspaper segment). It would have been useful to calculate OR and PTOI for all other corporate operating groups sans the newspaper segment. However, total corporate OR typically includes cost centers such as corporate overhead. Therefore, it is not possible to extract this data simply by subtracting newspaper segment OR and PTOI from total corporate OR and PTOI. One necessarily would have to adopt some kind of allocation scheme. In principle, it might be possible to allocate corporate overhead, based on the ratio of each segment’s OR to the whole; but this in turn assumes that each segment consumes corporate overhead equally. Since this seems unlikely, I just left the numbers the way they were.

For each company, I then separated the results into three groups: one for the period 1980 – 1989; another for the period 1990 – 1999; and a third for the period 2000 – 2005. Although any other grouping might have been equally or possibly even more explanatory, this one naturally suggested itself, seeing as how it parsed the data into easily-evaluated chronological decades.

I then undertook various statistical tests on each of these decade groups. These included: calculating return on sales (“ROS”) for both the newspaper division and the entire company; the percentage contribution of newspaper division ROS and OR to company ROS and OR, respectively; the standard deviation (“STDEV”) of ROS; and the R2 (“RSQ”) of PTOI to OR. This measured the following variables:

ROS is the ratio of PTOI to OR. It is a conventional measurement for the profitability of the business segment. I calculated ROS for each year of each decade. I also summed the individual PTOI and OR for each year of each decade, and then calculated ROS for the decade’s total. This evens out “anomalous” results that might pertain for any individual year. Examples: In 1995, TMC’s newspaper segment sustained a large loss due to “restructuring charges”; and, in 1997, DJ sustained a large corporate loss, due to the restructuring (that word, again) of its financial services division.

STDEV measures how widely the values of a series (here, each year’s ROS over a decade) are dispersed from the average value of the same series (the “mean”). When applied to a series of yearly ROS, it is a good proxy for the consistency or “riskiness” of profitability, i.e., the likelihood of achieving the average level of profitability over the same time-frame.

RSQ is the proportion of the variance in one data element (call it “y”) attributable to the variance of a corresponding data element (call it “x”). When applied to the series of PTOI and the series of OR for each year of each decade, it measures the “predictability” of one, versus the other. In other words, given a number for OR, what is the most-likely PTOI, given the decade’s results, and with what degree of confidence can that be asserted?

IV. FIGURES
The accompanying figures and tables depict some of the key results from this study. I suppose I could have supplied some simple line or bar charts, just depicting each company’s relative performance vis-à-vis the others. However, I wanted to see if we could depict the data differently, and in a way that carried more explanatory power.

Figure 1
Figure 1 might seem intimidating at first, but actually it asks a very simple question: what is the right “mix” between risk and return for each company’s newspaper segment? “Return” is depicted as newspaper segment ROS on the x axis, and “Risk” is depicted as STDEV of newspaper segment ROS on the y axis. Each company has three data points – one for 1980 – 1989 (labeled “80”); one for 1990 – 1999 (labeled “90”); and one for 2000 – 2005 (labeled “00”) (the one exception to this is TMC, which lacks an 00).

The basic intuition underlying Figure 1 is to “think geographically.” If one is looking for high return with low risk, then one ought to migrate towards Florida, in the south-eastern quadrants. The worst place is the north-western quadrants, up there by the State of Washington, which couple low return with high risk. The north-eastern and south-western quadrants present different issues. The further west one moves, the lower one’s return; the further south one moves, the lower one’s risk. Conversely, one encounters higher return as one moves east; but greater risk, as one moves north.

Using these principles, one can discern, for example, that TMC in the 90s was in a less-than-desirable location, with low return and high risk. Possibly this is one of the factors that accounted for its sale at the end of that decade? Companies like GCI and LEE, on the other hand, have consistently been able to realize higher return, at a lower level of risk. SSP from 2000 – 2005 generated exceptionally high return; but at a very high degree of risk.

If one assigns to “risk” the same relative premium one assigns to “return,” then it is easy to establish a hierarchal ranking of each company by decade. One simply takes the ratio of risk to return. The lower the resulting product, the better the company’s over-all “blended” risk-return characteristic (again, assuming we are concerned equally with both risk and return). On the other hand, if one were not so much concerned about risk, say, to half the value of return, then one would have to adjust the numerator of the fraction accordingly. For example, if you only were half as much concerned about risk, as return, then you should divide the numerator by two. Table 1 uses the former approach, i.e., no adjustments.

Besides which companies look as though they are less risky than others, the next thing to note about Figure 1 is how exceptionally UN-risky all of the companies are. Keep in mind that the x axis of Figure 1 is scaled from 0% – 35%; and the y axis is scaled from 0 to 9. In order to graphically depict any correlation between the two (i.e., the RSQ) it only would be fair to scale them identically. We then could draw a diagonal line from the lower left-hand corner of Figure 1, to the upper right-hand corner. Points falling on the line would represent companies that were just as risky as they were remunerative; their RSQ would = 100. Points below the line (such as ALL of the points here) would return more than they risked; RSQ would be <100.>100.

If we were to do this, though, all of the points would be crowded down into the lower right-hand corner of Figure 1. That would make it hard to read. It would, however, emphasize the point made by Table 1, which is: the only company-decades where the relative risk was more than half of the relative return, are DJ00 and TMC90. This is an exceptional result under any analysis.

Figure 2
Figure 2 attempts to discern if there is any kind of a relationship between the RSQ of the newspaper segment, and corporate RSQ. In turn, RSQ of the newspaper segment evaluates whether newspaper PTOI is correlated with newspaper revenues; corporate’s RSQ does the same thing, only for the entire company. As with Figure 1, each point on Figure 2 is a company-decade. Missing are: MNI, which always has had only one operating segment, meaning there is nothing to compare; LEE00, at which time the company consolidated itself into one operating segment; TMC00, because it no longer was there; and KRI before 1985 and after 1997, when it too only had one operating segment.

To get a better sense of how Figure 2 works, let’s look at a couple of examples. To start in the lower right-hand corner, KRI90’s newspaper segment RSQ was 92, whereas its corporate RSQ was 13. This means that KRI90 newspaper segment PTOI was highly correlated with KRI90 newspaper segment revenues; whereas there was very little correlation at all between KRI90 over-all corporate PTOI, and KRI90 over-all corporate revenue. From this, it is possible to conclude that KRI90’s newspaper segment exhibited far more predictable results, than did KRI90’s corporate group. A rough-and-ready way to consider this proposition is that, if you knew KRI90 newspaper segment revenue, you would be able to estimate its PTOI, with 92% accuracy. To the extent that this kind of a relationship is a desirable attribute for a business to have – and it probably is – then the KRI90 newspaper segment had an attribute that the company, considered as a whole, lacked.

Another example, the reciprocal of the last one, is DJ80. There was a very low statistical relationship between newspaper segment PTOI and OR (an RSQ of 21); whereas there was a tight statistical relationship between over-all corporate PTOI and OR (an RSQ of 96). One way of interpreting this is that the newspaper segment’s performance during this period was unpredictable, in the sense that there was little relationship between revenue, on the one hand, and profitability, on the other.

Points close to the diagonal line depict company-decades where both the newspaper segment and the corporate segment were in synch. The relationship between newspaper segment PTOI and revenue “matched” the relationship between over-all corporate PTOI and revenue.

There is nothing inherently wrong with a low RSQ. For example, it could signal steep PTOI (or revenue) growth; or, a precipitous decline. To illustrate, NYT00; TRB00; SSP00; and WPO00 all show lower newspaper segment RSQ’s, but higher over-all corporate RSQ’s. And, the 00 newspaper segment RSQ for NYT, TRB, WPO and BLC is lower than the corresponding 90 newspaper segment RSQ.

Let’s imagine that we can fix the start of the “digital age” as 2000 (and, given the way we’ve organized the data, there really isn’t another choice!). One way of interpreting the data, on this assumption, is that the newspaper segment RSQ of these companies has become less predictable as each company attempts to meet the management challenges presented by the “new media” paradigm. A counter-example is DJ, where 00 newspaper segment RSQ was significantly greater than 90 newspaper segment RSQ (though both were > their counterpart corporate RSQ’s!); perhaps it could be argued that DJ already anticipated these changes, and “had its act together.”

Figure 3
Figure 3 also is a little bit different. I wanted to evaluate whether newspaper segments are “over” or “under” performers in relationship to the company of which they comprise a part. Rather than simply comparing newspaper segment revenue with overall corporate revenue, and the same for PTOI, I wanted in effect to do both at once. I decided the best way to accomplish this would be to make the former the x axis, and the latter the y axis. Once again, each point is labeled with the applicable company, and the applicable decade.

For purposes of Figure 3, it doesn’t matter what the respective values of the points are. The key is whether they fall above or below the diagonal line that bisects the figure. If the point falls above the diagonal line, then the newspaper segment is disproportionately contributing to overall corporate profitability, that is, in relationship to its revenue. If the point falls underneath the diagonal line, then the opposite is true – the newspaper segment is a laggard, against the company as a whole.

Thus, for example, in the 1990s, DJ’s corporate results were so bad that the newspaper group actually made up for an overall corporate loss. The same thing is true with KRI in both the 1980s and the 1990s. On the other hand, BLC’s newspaper segment significantly under-performed the rest of the company in the 1980s; was approximately commensurate (that is, contribution to PTOI matched contribution to revenues) in the 1990s; but started slipping back below the line in the 2000s to date.

The way in which the results for some companies are clustered in the same general region is provocative. For example, NYT’s newspaper segment has consistently over-contributed to PTOI, but not by much; if you were to “connect the dots” between the points for each decade, you can see how the company’s overall results gradually improve over time. The exact opposite is true for WPO, where the newspaper segment over-contributed in the 1980s; was approximately even in the 1990s; but has started to lag the rest of the company for 2000 – 2005.

Another curious result is that the newspaper segment under-contributed in only eight out of 27 company decades. It might be easier to discern this by ordering the results in a table, as we did with Figure 1. This simply is a hierarchal ranking based upon the ratio of PTOI contribution to revenue contribution. If everything were 100, then the contribution percentages would be equal. Results > 100 mean that the newspaper segment over-contributed.

As with Figure 2, it might be possible to use Figure 3 to evaluate which companies are adopting more readily to the digital era. For example, SSP is the highest 00. In its 10-K for FY2005, SSP states it is “a diverse media concern with interests in national television networks (“Scripps Networks”), newspaper publishing, broadcast television, television retailing (“Shop At Home”), online comparison shopping (“Shopzilla”), interactive media and licensing and syndication. All of our media businesses provide content and advertising services via the Internet.” Yes, these activities comprise separate business segments. However, perhaps, cross-pollenization between divisions is one of the factors driving this result. Conversely, TRB is the lowest 00 – and maybe this is yet another reason why there currently is dissension at the company.

V. HYPOTHESIS FOR FUTURE WORK
While these results are interesting in and of themselves, it would be extremely interesting to compare, for example, newspaper segment with ROS with the ROS of non-newspaper-owning companies. I have a working hypothesis that ROS is considerably higher, considerably more consistent, and considerably more predictable for newspaper segments of large media companies, than it is for other types of industrial firms. I even have a reason why I think this hypothesis might be true, and it has to do with product liability.

The annals of recent U.S. industrial history are filled with what we might call “corporate catastrophes.” Examples of this are the Exxon Valdez oil spill; the Bhopal tragedy for Union Carbide; the Johnson & Johnson Tylenol scare; Johns-Manville’s troubles with asbestos; and, more recently, Merck’s problems with Vioxx. Crises like these illuminate the ways that business organizations work, because they juxtapose the corporation’s response to the crisis against its “normal,” day-to-day behavior. The ones cited above each had a massive adverse impact on each company’s activities, operations, and financial results.

One business that doesn’t face this type of problem is the newspaper business. The reason why is because newspapers have a special form of protection that is unique to the industry, which is called the First Amendment. U.S. Supreme Court decisions have made it clear that “public figures” – and we’re only interested in them, anyway – can’t sue a newspaper for libel, except under the most unlikely and extenuating circumstances. Thus, even if there is something terribly wrong with the product the newspaper sells, the newspaper effectively is shielded from liability and financial risk.

I’m certainly not saying that the First Amendment is a bad thing. Rather: (a) if my hypothesis about newspaper segment ROS is true, and (b) if it is possible to identify protection from product liability as one of the reasons why, then this has had a unique and favorable impact on newspaper segment financial performance. Not a single one of the 10-K’s I examined, for instance, had a footnote to a financial statement, identifying libel litigation as a material liability.

An example is NYT’s 10-K for FY2005, which states: “There are various legal actions that have arisen in the ordinary course of business and are now pending against the Company. These actions are generally for amounts greatly in excess of the payments, if any, that may be required to be made. It is the opinion of management after reviewing these actions with legal counsel to the Company that the ultimate liability that might result from these actions would not have a material adverse effect on the Company’s Consolidated Financial Statements.” I’m confident that many manufacturing, industrial, pharmaceutical, and other types of concerns wish they could include a similar note.