Embargo (film)
Embargo (film) (Photo credit: Wikipedia)

This is a research report, based on a grant from the American Society of Civil Engineers to explore the potential for adverse economic impact on journals from imposed public access embargoes. Basically I am describing a model that has yet to be built, sneaking up on the equations as it were.

I began by examining two years of detailed download data for over thirty ASCE journals. I found a surprising pattern, similar to that recently reported by Phil Davis, namely that most downloads occur well over 12 months from the date of publication. To see a small example of this pattern, ASCE has a webpage that displays the top twenty downloaded articles for the last twelve months, on a running basis. As of this writing well over half of these articles were downloaded more than 12 months after publication.

The question then becomes how to relate this download pattern to subscription revenue in the case where a journal’s content becomes free after an imposed embargo period? This question is explored below, especially with reference to the emerging US public access program.

The US Government is moving to impose an embargo period on all journal articles which are based on federal funding. The initial guidance issued to the federal funding agencies suggests a 12 month embargo period, after which a free, publicly available version of the article will be required. Other funding agency and institutional policies have called for a 6 month embargo period instead.

Presented below is a preliminary analysis of the potential adverse economic impact of a 12 month embargo period. The starting point is that a great many of the downloads of journal articles occur more than 12 months after publication. Making these articles freely available has the potential to reduce the value of journals to a similar degree. Thus revenue losses of 50% or more are possible. Several factors that might affect the magnitude of this devaluation are then discussed. The approach is that of describing an economic model that can be used to explore the various issues, as well as defining the relevant empirical questions.

But the basic point is that embargoes devalue journals. In economic terms embargoes create a free rider problem. A free rider is someone who gets the benefits of economic activity without paying their fair share.

A simple model for estimating loss of revenue due to embargo induced devaluation looks like this:

The simple model: Devaluation revenue equals present revenue times the ratio of affected post embargo downloads to total downloads.

Note that affected post embargo downloads only includes that fraction of articles that would be subject to embargoed release, for example, those based on US funding. In our worst case scenario we assume that all articles outside the embargo period are subject to the embargo and will be made freely available.

This simple model is based on several assumptions, including the following:

1. Downloads accurately reflect total usage.

2. The value of usage does not change over time.

3. Almost all of the value of a journal is in its use.

Each of these assumptions can be questioned and many of the specific debates are related to them. Thus such debates can often be characterized by alternative models, which in turn raise specific empirical questions. This is discussed below. Moreover, other assumptions may appear in the course of discussion, raising new questions. The point of models like this is not to make predictions, but rather to clarify the issues, uncertainties and possibilities. But these possibilities must be taken seriously by those proposing to impose mandated embargoes.

This simple model can be modified and made more complex in order to incorporate the various factors we discuss below, as well as other factors. For example if near term downloads are thought to be more valuable than longer term downloads there are various weighting schemes that might be incorporated, perhaps along the lines of the various depreciation formulas. The decline in value of articles over time might even be nonlinear, logarithmic for example. Of course these modifications would need to be justified. As noted below, articles might get more valuable with time, not less.

Also our analysis is based solely on downloads, not total usage. In this regard downloads are what is called a proxy for total usage. A proxy is a rough measure or indicator. Downloads are probably a good measure of serious usage because they take effort, far more than simple page views, and they suggest that the article is considered to be important. Moreover, once downloaded an article may be read by more than one user, such as a project team or a research group. Thus downloads may actually underestimate usage.

Based on the simple formula a given journal might see revenue losses of 50% or more under a 12 month embargo. Note that this is not a prediction, merely a plausible scenario based on available data and a very simple model. But these are scenarios that should be taken seriously. The over 50% loss scenario is based on the assumption that all major funders follow the US lead and impose a 12 month embargo. A lower loss scenario assumes that only the US imposes an embargo, which seems unlikely. Even if only 20% of the articles are affected and only 50% of the downloads occur after 12 months, this might lead to a 10% revenue loss, which is still a significant loss These are serious potential adverse economic impacts.

Factors that might reduce or increase the adverse economic impact include the following. Note that we need proposed equations for each, with empirical support, not just arm waving conjectures.

A. Time sensitivity of information

As alluded to above, it might be argued that early downloads are somehow more valuable than later downloads, thus reducing the devaluation due to making later downloads free. In disciplines where information is highly time sensitive, perhaps patentable biotechnology for example, this might be true. But it is hard to see this as a major factor in general science and engineering, where information needs tend to occur as a result of project conditions, which can happen at any time. Indeed it is this feature which may explain the repeated downloading of articles many years after publication. Downloads are need driven. Thus it is possible that older articles are actually more valuable, not less. This is an empirical question.

B. Discount rate

The discount rate reflects the fact that a given return on an investment is more valuable the sooner it occurs. If subscriptions are regarded as investments then there might be an argument that later downloads are not as valuable as near term downloads, at the time of the investment decision. However, given that discount rates are usually figured at 10% or less per year, sometimes much less, this factor will have little impact on the issue of the economic impact of embargo times of two years or less. Nor is it obvious that subscriptions are regarded as investments.

C. Early loaders, late loaders and waiters

We have been looking at downloads as though the pattern were fixed.  That is, where there are downloads either before the embargo period or after it begins. This ignores the fact that usage is done by users. Let us call these people early loaders and late loaders. What is missing is a third group which we call waiters. These are people who are prepared to wait for some time, such as several months, to download for free. The impact of waiters is to shift downloads from before the embargo begins to afterward. Waiters increase the devaluation effect of the embargo. This effect is likely to be greater the shorter the embargo period.

Note that libraries respond to demand and embargoes should reduce demand from late loaders and waiters. But an important question here is to what degree early loaders and late loaders are the same people. To the extent that they are the same the embargo might not reduce demand for library subscriptions. This is an empirical question.

D. Bundling

It has been argued that bundling of large numbers of journals from different disciplines, for sale to libraries, somehow protects vulnerable journals from cancellation due to imposed embargoes. Many journals do not participate in this kind of bundling activity. Nor is this argument convincing because it suggests that libraries do not act economically.

E. The PubMed Central precedent argument

It has been argued that PubMed Central (PMC) has enforced a 12 month embargo period from which no cancellations have resulted. This PMC argument is unproven and probably not credible. Journal prices have gone up significantly over this time, as has the number of journals on the market. Reportedly library budgets have not kept pace, a factor which is driving the open access movement. So there must have been many cancellations. We just do not know the role embargoes have played. This is an empirical question.

F. The vagueness of articles being based on US Federal funding

It is far from clear what constitutes an article’s being based on federal funding and so falling under the embargo. For example, suppose the research used some computer time on a federal supercomputer, or a grad student involved was on a federal scholarship. Do these cases count? There is even the possibility that researchers will create some minor federal involvement just to gain public access via the embargo. Doing so might greatly increase the adverse impact of the embargo.

G. Different journals may have different download over time profiles

It appears that different journals vary significantly in their relative percentage of post embargo downloads. Thus different journals might experience different levels of adverse economic impact.

Uncertainty versus risk

The fact that there are significant uncertainties in estimating the adverse economic impact of embargo based devaluation does not reduce the risk that this impact is large. While the factors listed above in some cases suggest that the adverse economic impact of imposed embargoes may be less than our simple model suggests, these arguments are merely conjectures at this point. Conjecture does not reduce risk.

Basic economics says that if you reduce the value of your product while keeping the price the same you will lose sales. Making articles freely available after an embargo period reduces the value of a subscription. Conjectures that this reduction is small are not a sound basis for public policy.

The US government should not experiment with an industry where the risk of significant adverse economic impact is unknown but potentially large. This is certainly the case with 12 month embargoes for many disciplines, including civil engineering. It is even more so the case for a 6 month embargo, which is potentially catastrophic. Note too that the US action may well become a global precedent. The US needs to take this possibility seriously in its risk impact assessment, so we have included it as an upper bound on our estimates.


34 Thoughts on "Estimating the Adverse Economic Impact of Imposed Embargoes"

While I understand this is a first pass effort at finding a framework to examine these issues and that you are seeking suggestions for refinement, I think the approach falls short of the mark and needs a major rethink. In particular, some of the very basic premises upon which it is based seem incorrect to me. To wit:

But the basic point is that embargoes devalue journals.

This may just be poorly phrased, but isn’t it public access mandates that devalue journals? Embargoes are a mechanism to retain value in journals for a given time period. They prolong value, rather than having a negative impact.

But the key flawed premise here is:

Almost all of the value of a journal is in its use

This is off on many levels.

1) As Phil Davis showed, open access articles see higher usage than their subscription access counterparts (http://www.fasebj.org/content/early/2011/03/29/fj.11-183988.full.pdf+html). If “value” is solely based on usage, shouldn’t any measure that increases usage similarly increase “value”?

2) Is there any correlation between price and usage? Journals like “Brain Research” or “Tetrahedron” are much more expensive than “Nature”, yet articles in Nature are more widely used.

3) Journals utilize a wide variety of revenue streams to provide financial support, reaching far beyond subscription. Subscription decisions are often usage based, but any approach that fails to account for these other revenue streams ignores significant portions of the economic value of journals to their owners. What about page charges, color charges, advertising, digital archive sales, translation rights? See tomorrow’s post on secondary rights revenue as an important example that can account for a significant percentage of a journal’s revenue.

4) Even looking at subscription, there is not a one-to-one ratio that can be found between article usage and subscription decisions. If a journal is used 10% less by an institution, the price paid, and the “value” of the journal, remains the same. If 99% of a journal’s articles become freely available, yet the library continues to subscribe at the full price it once paid for access to 100%, can it be said to have lost “value”? One likely needs to think in terms of thresholds and tipping points. Is there a percentage usage decrease where a journal is likely to be dropped, hence losing all its subscription value? Why not look at usage patterns of journals that have been dropped by institutions, and combine that with surveys of decision makers to better understand how such decisions were made?

5) Journals can also be looked at as investments, rather than annual revenue generators. On occasion, a journal or set of journals will change hands and be sold off to a new owner. How does the price paid in such circumstances relate to journal usage? Is there value in the brand of a journal beyond usage numbers? What about a journal that is purchased by a publisher looking to add prestige to their program or increase their program’s overall Impact Factor numbers?

I do think modeling like this is important, but the economics of journal publishing is far too complex a subject to be simplified down into one or two bullet point factors. More on this in tomorrow’s post.

This comment is rather long to respond to all at once, so just two quick points. First I agree that embargo is the wrong word for what is actually mandated access, but it is the word in common use so I have used it.

Second, that the value of a journal is in its use is not an assumption of the model, as you seem to think. It is merely an assumption of the initial simple model equation. I am looking for additional factors (and equations), some of which I listed to kick things off. You seem have listed some other important factors, or we may have an actual disagreement; it is hard to say which at this point.

It is the case that some publishers have large secondary revenue streams. The adverse economic impact of embargo mandates should be larger for them. If one assumes that everything past the release date is lost then the equation is pretty simple. I will put this on the list, while I think about your other points.

If one assumes that everything past the release date is lost then the equation is pretty simple.

I don’t think one can assume this. An article that is freely available still draws traffic and thus generates advertising space and revenue. When we have an important article that’s drawing a lot of attention, we usually make it freely available. This raises the profile of the journal, adding value to the brand. In addition, if a librarian sees an enormous amount of usage of a journal’s free articles on top of a reasonable amount of usage of subscription articles, it is likely a signal of that journal’s importance to researchers on campus, and thus is potentially a valuable driver of retaining subscription revenue.

One can always refine the refinements, but for regulatory impact analysis we are just trying to get a rough idea of what adverse impacts an uncertain future might hold. There have been claims that twelve month mandates are harmless, so I am exploring the alternative hypothesis. For publishers with large secondary sales it is hard to see the prestige of free articles offsetting the loss of those sales on a regular basis. So I am inclined not to include this refinement.

Advertising revenue might well grow with mandated access, thus offsetting some potential loses. But my understanding is that many journals sell little advertising and I wanted to start with the core process of subscription sales. If a publisher wanted to adapt the model to their own case it would be a different story.

I nowhere suggest that the economics of journal publishing can be simplified down into one or two factors. In fact I have presented a starting point plus seven additional factors that need to be considered. You have added several other factors, for which I thank you. Regulatory impact models can be quite elaborate and journal publishing is no exception, but they are a well developed technology. If anything, you are over simplifying what I am trying to do.
(This comment went into the wrong place, below, the first time I posted it.)

This comment is rather long to respond to all at once, so just two quick points. First I agree that embargo is the wrong word for what is actually mandated access, but it is the word in common use so I have used it.

David, I think you’re missing an important point here and continuing the confusion that David C. points out. The word “embargo” is indeed commonly used, but it’s not commonly used as a synonym (even a loose one) for mandated access. An embargo is a limitation on the mandate. A mandate says “you are required to make this content freely available,” whereas an embargo says “… but not until some period after initial publication.” In this posting, you seem to be using the term “imposed embargoes” to mean “mandated public access,” which is almost the exact opposite of what “embargo” means.

Rick, I think if you look back at other Kitchen articles and comments you will see that the misnomer “embargo” is the norm.

No, it really isn’t. The term “embargo” is indeed commonly used, but I can’t think of a single example of someone using it as a synonym for “mandate.” (Nor did you use it that way in the posting you cite. In that posting you used “embargo,” accurately, to mean a delay that the mandate allows. The embargo is not the mandate; it’s a qualification of the mandate.)

The term “delayed access” does mean pretty much the same thing as “embargoed access.” But neither term means the same thing as “mandate.” An OA mandate may or may not allow for an embargo. When it does, as David C. pointed out, the purpose of the embargo is to mitigate the expected financial impact of the mandate on publishers. For this reason, it makes no sense to refer to the “adverse economic impact of imposed embargoes.” Embargoes are allowed by mandators, not imposed by them, and the entire purpose of an embargo is to minimize the mandate’s adverse economic impact on publishers and authors.

A few quick comments. I am uneasy about the value of this type of modeling, and I say this as a fan of data. On the one hand, it is useful to look at patterns of usage of journals and articles over time as a measure of the value of the content to the user. This needs to be done at a granular level, looking at differences between fields, type of journal etc. Of course, the value is not just to the user. The value of a journal may be measured in other ways; to the author, the institution, society, funder, and other stakeholders. If you then try to layer embargo periods over this, I fear that the resulting need to simplify the model renders it near useless – this is all despite my sympathy with the concept of knowing more about how the various facets of OA affect the business of journal publishing. To some extent modeling of this sort can be shaped to meet the desired ends. In other words, OA advocates and OA haters will respond according to their faith, without truly serving the interests of our communities who need our content, and for us to provide real, long lasting value.

I nowhere suggest that the economics of journal publishing can be simplified down into one or two factors. In fact I have presented a starting point plus seven additional factors that need to be considered. You have added several other factors, for which I thank you. Regulatory impact models can be quite elaborate and journal publishing is no exception, but they are a well developed technology. If anything, you are over simplifying what I am trying to do.

On the other hand there is something simple that we are seeking, namely the graph of subscription revenue versus mandated embargo period. Other things being equal, if the mandate covers everything and the period is zero, so everything is free, then revenue will also be zero. So that is one end of the curve. Presumably when the period is long enough the revenue will converge with the pre-mandate revenue, so that is the other end.

The empirical questions are (1) when does this convergence occur and (2) what is the form of the curve? Of course these are not simple questions, nor is everything equal, but the concept itself is pretty simple.

You are making a case that can only stand up based on the value inherent in the content. And if the value is in the content, rather than the service provided by the publisher, then there are serious questions about how much a publisher should be make out of a value that they themselves are not providing.

There is no case to be made for government policy to reward publishers for value that the publishers have not added.

Now, I do believe that publishers provide valuable services. That the publisher itself adds value to the product. And that publishers need to compensated accordingly in order to provide those services.

But that’s why prioritization should be given to compensating publishers for providing the services (which they do), not for providing the content (which they don’t). Which is somewhere more than what academics think is fair, but probably somewhere less than what publishers think they should be able to command.

This is clearly untrue, as David points out. The price of a journal has nothing to do with the value of its content. For instance, some of the most important journals sell for less because they have large circulations and large advertising revenue streams (and other revenue streams), so can be priced lower. Also, because site licenses are usually based on FTE counts, journals that have higher usage are lower-priced on a per-usage basis than counterparts.

Price is driven by the market. The same goes for OA journals. The higher-prestige or better-targeted OA journals are commanding a higher APC rate. Brands, positioning, and other market forces drive pricing.

If the “value inherent in the content” really drove pricing, then we wouldn’t need the impact factor or anything else. We’d simply buy the most expensive content, which would, by dint of price alone, be the most reliable and most important science around. Luckily, we live in a better market than that, so that often the best content is also the most affordable because better content means greater market demand, which lowers prices to any single purchaser.

Graham, you seem to be using content to mean what that author supplies, in which case I am making no such assumption. Of course there is no value or use without content but publication in itself is also a value. This is an economic model which simply considers the product as such. Analyzing the contributions to the product is a different issue. On the other hand there is the interesting issue of providing access to the accepted manuscript versus the version of record. These are different products. There is also the issue of how access is provided.

I did not say that you were making that assumption. I said that you are analyzing a market that depends on it.

Yes, publication is a value in itself, which should be rewarded (it also has costs, which the rewards need to cover). But it’s that value which the subscription industry gives away for free, and instead makes it’s returns on the value of what the author supplies.

Your model suggests that a 12-month embargo may reduce revenue by 10 to 50%. For an industry that makes 30-40% profits based on selling what it doesn’t do, rather than selling what it does, is anyone outside of the industry going to be bothered by a 10% revenue reduction, which would still equate to making a fairly substantial profit?

There are few publishers that can use very high production costs / spread across a very wide readership to defend the traditional economic model. And, where embargoes exist, they need to be appropriate on a disciplinary basis (not one size fits all). But outside of that, trying to defend revenue based on denying access will only raise anger, which may have longer term consequences.

In the majority of cases, we should be looking for ways for publishers to be paid directly for what they do, not for they don’t. The amount of revenue – and profit – is far more defensible as a function of the actual services provided. And all the contentious issues of embargoes and access go away.

Engineering content has a very long shelf life. Average citation age is 5 years. Civil engineers are a very conservative bunch and they like to hang their ideas out for a while before committing to them. A twelve month embargo on federally funded papers would seriously devalue the subscription product. We already know that some libraries will cancel subscriptions and instead rely on PPV to fill any gaps in newer content.

The landscape is still in flux. If the mandates call for authors to deposit their own accepted manuscripts in random repositories, that is one thing. Once these manuscripts are easily discoverable via one search, that makes it more of a threat. The manuscript version is a different product as David mentioned. ASCE supports CHORUS but we may be forced to allow access to the final publisher version of the paper due to resource and technological constraints.

I know that this is not the popular opinion of researchers, but I find it amazing that the federal government would create policies that will financially harm societies and in some cases, put them out of business completely. We are not talking about putting paycheck loan businesses or shady insurance companies out of business. We are talking about putting non-profit professional societies out of business. The economic impact report is an entry to this discussion.

There is something to be said for holding up one side of an argument. But I would be interested in a good faith effort to model the benefits of open access/embargoes, both for publishers and other parties. Surely it’s not only costs!
Here’s one example: yesterday I was researching a question that led me to 5 old citations. I had subscription-based access to 3 of them, but those didn’t pan out with the information I needed. My Bayesian update of the probability that the other 2 would pan out dropped, but I wasn’t going to pay $30 each for the privilege of finding out. Society lost my research time, some of which went to looking for non-gated versions of the content and some of which resulted in 3/5ths of an answer to a research question.

I have seen at least one attempt to estimate the benefits quantitatively and as an issue analyst I would be happy to look into it. If the US agencies have to do regulatory impact analysis (as they should) then cost benefit analysis is mandatory. The biggest problem is that if you drive a journal out of business, as Angela mentions above, then there is nothing to access.

“The biggest problem is that if you drive a journal out of business, as Angela mentions above, then there is nothing to access.”

I’m afraid I don’t follow that logic. Just because a society ceases to exist, does that mean that people no longer conduct research in the topic? I doubt it. Societies need to deliver value to their members above and beyond what is available for free. That is why they were started, and that is what will keep them alive.

The research is still there but not the article, so there is no article to access.

The peer-reviewed/edited article is not there, but the original submission would be available. Presumably there is some value in that. I think this gets back to the point made by Graham Triggs above

How would it be available if there is no journal? Are you claiming that journals are unnecessary? Some do of course but I do not agree, nor is it relevant to this post.

The Danish Research Council for Culture and Communication (=humanities) demanded some years ago its supported journals to be OA after 12 months.
Most of the 32 journals were heavily affected by this. Many lost 15-17 % of their subscribers the first year and the losses continued at a lower speed in the following years.
Result: several of the journals don’t exist anymore and most of the remaining journals have not enough income to cover the traditional expenses for the editorial work as they could before = the quality of the handling process of the articles are lowered.
Many of the subscribers (around 93 %!) were private persons interested in a topic and have so far used the journals to be informed about the development in research areas within their interests and they supported the journal/society at the same time. Reading of abstracts was an important part of the reason behind their subscription. The heavy agitation for OA gives those private people the opinion that OA must be a sustainable solution = they can cancel their subscription without doubts. Those cancelations won’t be included if the discussion of impact of embargos is only focused on download of single articles.

Do you know if any formal data or a writeup of this has been published anywhere?

As I see it, librarians are faced with conflicting demands: on the one hand, flat or only slowing increasing library acquisition budgets compel them to make tough decisions about what journal subscriptions to drop; on the other hand, faculty are going to want access to the current issues of journals whether or not they download articles during the first year, just because they think they might miss something important. How librarians reconcile these competing demands will no doubt vary from one institution to another. But i see no easy way of modeling their behavior, which after all is what ultimately determines the economic outcomes for publishers.

Sandy, I think we just need to be more creative with our offerings to libraries. I haven’t figured it out just yet but there has to be a business model that works for all parties involved.

It would be very interesting to do a librarian decision model rather than an economic model, and I have thought about what such a decision model might look like quite a bit. My thought is that librarians respond to pressure from their constituents and if a lot of articles in a journal become freely available then that pressure will slacken. The librarian need not, indeed will typically not, know why pressure is not there, but the lack will support cancellations.

On the other hand you have raised an economic model factor which I tentatively call prestige. This is that faculty want certain journals in the library simply because they are the leading journals in their field. If you have a strong department in field F then it would be an embarrassment not to have the top journals in the library. This may be an important factor, but as Joe Esposito has pointed out, it only protects the top journals.

I am also alluding to this when I talk about early loaders, late loaders and waiters. It makes a difference who they are.

My thought is that librarians respond to pressure from their constituents and if a lot of articles in a journal become freely available then that pressure will slacken. The librarian need not, indeed will typically not, know why pressure is not there, but the lack will support cancellations.

Speaking as a collection development librarian, I can say that this is exactly correct. You’re also right that subscription pressure from faculty arises not only from their desire for access to the subscribed content, but also from their perception that having a particular journal in the collection confers prestige on the institution. This is usualy expressed in terms like “It’s absurd that an institution like ours does not subscribe to Fancy Journal X.”

On an historical note it is my understanding that this data comes from COUNTER, which is a new initiative. It is quite normal for new data to lead to surprises such as Phil’s analysis, and perhaps my model as well. I think Phil’s analysis indicates that scientific communication may work significantly differently from what has been typically assumed. That is, use is driven by need, not by novelty, and needs only arise spread over a long period of time.

There is then in a sense no cutting edge in the sense of a linear, shoulder to shoulder, march forward. It is more of a groping forward in many directions. Need only occurs because a given project happens to find itself there, probably unexpectedly. My issue tree model of scientific progress may play in here: http://scholarlykitchen.sspnet.org/2012/07/17/how-does-science-progress-by-branching-and-leaping-perhaps/.

In the discussion above there is some confusion regarding the concept of “value.” In economics price is often taken to measure value but the case we are looking at is one where the price of a product is changed so that concept is not applicable. Specifically, what an embargo based mandate does is raise the unit price of usage.

Suppose that pre-mandate a journal costs x and sees y downloads. The cost per download is x/y. Now suppose that a mandate makes half of the downloads free. The cost per download of the remaining downloads is now 2x/y, so the unit cost has doubled. This is why I say the journal has been “devalued” by the mandate, because the customer gets less for the same money. It is like making a candy bar smaller without changing the price, thus raising the unit price where the unit is weight.

Comments are closed.