A paper, and publishing
Even at my point in life, the moment of publishing an academic paper is a one to celebrate, and a moment to reflect.
The New-Keynesian Liquidity Trap is published in the Journal of Monetary Economics -- online, print will be in December. Elsevier (the publisher) allows free access and free pdf downloads at the above link until November 9, and encourages authors to send links to their social media contacts. You're my social media contacts, so enjoy the link and download freely while you can!
The paper is part of the 2012-2013 conversation on monetary and fiscal policies when interest rates are stuck at zero -- the "zero bound" or "liquidity trap." (Which reprised an earlier 2000-ish conversation about Japan.)
At the time, new-Keynesian models and modelers were turning up all sorts of fascinating results, and taking them seriously enough to recommend policy actions. The Fed can strongly stimulate the economy with promises to hold interest rates low in the future. Curiously, the further in the future the promise, the more stimulative. Fiscal policy, even totally wasted spending, can have huge multipliers. Broken windows and hurricanes are good for the economy. And though price stickiness is the central problem in the economy, lowering price stickiness makes matters worse. (See the paper for citations.)
The paper shows how tenuous all these predictions are. The models have multiple solutions, and the answer they give comes down to an almost arbitrary choice of which solution to pick. The standard choice implies a downward jump in the price level when the recession starts, which requires the government to raise taxes to pay off a windfall to government bondholders. Picking equilibria that don't have this price level jump, and don't require a jump to large fiscal surpluses (which we don't see) I overturn all the predictions. Sorry, no magic. If you want a better economy, you have to work on supply, not demand.
Today's thoughts, though, are about the state of academic publication.
I wrote the paper in the spring and summer of 2013, posted it to the internet, and started giving talks. Here's the story of its publication:
September 2013. Submitted to AER; NBER and SSRN working papers issued. Blog post.
June 2014. Rejected from AER. 3 good referee reports and thoughtful editor report.
October 2014. Submit revision to QJE.
December 2014. Rejected from QJE. 3 more thoughtful referee reports and editor report.
January 2015. Submit revision to JME.
April 2016. Revise and resubmit from JME. 3 detailed referee reports and long and thoughtful editor report.
June 2016. Send revision to JME
July 2017. Accept with minor revisions from JME. Many (good) comments from editor
August 2017. Final revision to JME
September 2017. Proofs, publication online.
December 2017. Published.
This is about typical. Most of my papers are rejected at 2-3 journals before they find a home, and 3-5 years from first submission to publication is also typical. It's typical for academic publishing in general. Parts of this process went much faster than usual. Three months for a full evaluation at QJE is fast. And once accepted, my paper sped through the JME. Another year or two in the pipeline between acceptance and publication is typical.
Note most journals count average time to decision. But what matters is average time to publication of the papers they publish, and what really counts is average time to publication in the journal system as a whole.
Lessons and thoughts?
Anyone doing research on zero bound in new-Keynesian models in the last 4 years, and carrying on this conversation, interacted with the working paper version of my paper (if at all), not the published version. Any work relying only on published research is hopelessly out of date.
Interest rates lifted off the zero bound quite a while ago, so in the policy conversation this publication at best goes into the shelf of ideas to be revisited if the next recession repeats the last one with an extended period of zero interest rates , and if we see repeated invocation of the rather magical predictions of new-Keynesian models to cure it. If the next recession is a stagflation or a sovereign debt crisis, you're on your own.
Rather than means of communication,
the ark, the library, the place where final, and perfected versions of papers are carved in stone for future generations. (Some lucky papers that make it to graduate reading lists more than 5-10 years after their impact will be read in final form, but not most.)
And this paper is perfected. The comments of nine very sharp reviewers and three thoughtful editors have improved it substantially, along with at dozens of drafts. Papers are a conversation, and it does take a village. The paper also benefitted from extensive comments at workshops, and several long email conversations with colleagues.
The passage of time has helped as well. When I go back to a paper after 6 months to a year, I find all sorts of things that can be clearer. Moreover, in the time between first submission and last revision, I wrote four new papers in the same line, and insights from those permeate back to this one.
So, in the end, though the basic points are the same, the exposition is much better. It's a haiku. Every word counts.
But such perfection comes at a big cost, in the time of editors and referees, my time, and most of all the cost that the conversation has now moved on.
The sum length of nine referee reports, four reports by three editors, is much longer than the paper. Each one did a serious job, and clearly spent at least a day or two reading the paper and writing thoughtful comments. Moreover, though the reports were excellent, past the first three they by and large made the same points. Was all this effort really worthwhile? I think below on how to economize on referee time.
Of course, for younger people
Many institutions give tenure, chairs, raises, and other professional advancement based at least in part on numbers and placement of publications. For that purpose, timeliness of publication is less of a problem, but with a six year tenure clock at many places and five year lags, timeliness of acceptance the quality rating is a big problem. The sorting and branding function isn't working that well either. But having journals outsource quality evaluation was always an imperfect institution. Maybe we should just have star ratings instead -- seriously, start up a website devoted to crowd-sourcing working paper evaluation. Or, perhaps tenure committees will have to actually start reading papers. I don't think the journals see this as their main function either. They're set up to publish papers, not judge people's tenure, so improving journals as a tenure granting mechanism will be a hard sell.
There is some good news that this data point represents, relative to state of journal publishing 15-20 years ago. (See Glenn Ellison's superb "The slowdown in the economics publishing process," JSTOR, undated, one of my proudest moments as a JPE editor.)
Journals had gotten in to a rut of asking for round after round of revisions. Now there is a strong ethic of either rejecting the paper, or doing one round of revisions and then either publishing with minor changes or not. Related,
Journal editors have become, well editors. The referees provide advice, but the editor thinks about it, decides which advice is good and not, and makes the final call. Editors used to defer decisions to referees, which is part of the reason why there were endless revisions. This change is very good. Referees have little incentive to bring the process to a close, and they don't see the pipeline of papers to the journal. They are not in a good position to find the right balance of perfection and timeliness.
In my case, editors were very active. The referees wrote thoughtful reports, but largely made similar points. In fact, the strongest advice to reject came at the JME. But the AER and QJE editors were not impressed in the end by the paper, and the JME editor was.
So, with this state of affairs in mind, how might we all work to improve journals and the publication process?
I will take for granted that greater speed, and making journals more effective at communication and not just archiving and ranking is important. For one reason, to the extent that they continue to lose the communication function, people won't send articles there. Already you can notice that after tenure, more and more economists start publishing in conference volumes, invited papers, edited volumes, and other outlets. (blogs!) The fraction willing to take on this labor of love for journal publication declines quickly with age. Research productivity and creativity does not take quite such a parallel decline. (I hope!)
Always the free-market economist, I note that conference volumes, edited volumes, and solicited papers in regular journals seem to be healthy and increasing, which is a natural response to journal slowdown. This is a way to get papers in print more quickly. In the early days of the internet I had a rule never to publish in volumes, as they disappeared to library shelves and could not be found electronically. Now many of them have solved that problem. The NBER macro annual and Carnegie-Rochester conferences are good examples. The Review of Finance editor recently solicited my "Macro-finance" essay which therefore sped through publication. My active editors are also often taking a more active role in soliciting promising working papers. This helps to break the editor-to-paper match. But this isn't an ideal state of affairs either. Conference volumes tend towards commissioned work. Original work by people out of the social network of the conference organizers and editors has a tough time. (The market responds, organize more conferences.)
Suggestion one:
Around any economist cocktail party, there is a lot of whining that journals should do x y and z to speed things up. I start with what you and I can do. It is: do unto others as you would have them do unto you. If you complain about slow journals, well, how quickly do you turn around reports?
My recommendation, which is the rule I try to follow: Answer the email within a day. Spend an hour or two with the paper, and decide if you will referee it or not. If not, say so that day. If you can give a quick reaction behind your reason, that helps editors. And suggest a few other referees. Often editors aren't completely up to date on just who has written what and who is an ideal fit. If you're not the ideal fit, then help the editor by finding a better fit, and do it right a way.
If you agree to do a report, do it within a week. If you can't do it this week, you're not likely to be able to do it 5 weeks from now, and say no.
More suggestions:
Journals insist that you only send to one journal at a time. And then wait a year or more to hear what they want to do with it. Especially now that we are moving towards the editor-centric system, and the central question is a match with editor's tastes, why not let journal editors share reviewer advice and compete for who wants to publish it? By essentially eliminating the sequential search for a sympathetic editor, this could speed up the process substantially.
I don't know why lower-ranked journals put up with this. It's the way that the top journals get the order flow of best papers. Why doesn't another journal say, you can send it to us at the same time as you send it to the AER. We'll respect their priority, but if they don't want it we will have first right. The AER almost does this with its field journals. But the JME could get more better papers faster by competing on this dimension.
The journals say they do this to preserve the value of their reviewer time. But with shared or open reviews, that argument falls apart.
We advocate competition elsewhere. Why not in our own profession?
Update: An email correspondent brings up a good point:
The New-Keynesian Liquidity Trap is published in the Journal of Monetary Economics -- online, print will be in December. Elsevier (the publisher) allows free access and free pdf downloads at the above link until November 9, and encourages authors to send links to their social media contacts. You're my social media contacts, so enjoy the link and download freely while you can!
The paper is part of the 2012-2013 conversation on monetary and fiscal policies when interest rates are stuck at zero -- the "zero bound" or "liquidity trap." (Which reprised an earlier 2000-ish conversation about Japan.)
At the time, new-Keynesian models and modelers were turning up all sorts of fascinating results, and taking them seriously enough to recommend policy actions. The Fed can strongly stimulate the economy with promises to hold interest rates low in the future. Curiously, the further in the future the promise, the more stimulative. Fiscal policy, even totally wasted spending, can have huge multipliers. Broken windows and hurricanes are good for the economy. And though price stickiness is the central problem in the economy, lowering price stickiness makes matters worse. (See the paper for citations.)
The paper shows how tenuous all these predictions are. The models have multiple solutions, and the answer they give comes down to an almost arbitrary choice of which solution to pick. The standard choice implies a downward jump in the price level when the recession starts, which requires the government to raise taxes to pay off a windfall to government bondholders. Picking equilibria that don't have this price level jump, and don't require a jump to large fiscal surpluses (which we don't see) I overturn all the predictions. Sorry, no magic. If you want a better economy, you have to work on supply, not demand.
Today's thoughts, though, are about the state of academic publication.
I wrote the paper in the spring and summer of 2013, posted it to the internet, and started giving talks. Here's the story of its publication:
September 2013. Submitted to AER; NBER and SSRN working papers issued. Blog post.
June 2014. Rejected from AER. 3 good referee reports and thoughtful editor report.
October 2014. Submit revision to QJE.
December 2014. Rejected from QJE. 3 more thoughtful referee reports and editor report.
January 2015. Submit revision to JME.
April 2016. Revise and resubmit from JME. 3 detailed referee reports and long and thoughtful editor report.
June 2016. Send revision to JME
July 2017. Accept with minor revisions from JME. Many (good) comments from editor
August 2017. Final revision to JME
September 2017. Proofs, publication online.
December 2017. Published.
This is about typical. Most of my papers are rejected at 2-3 journals before they find a home, and 3-5 years from first submission to publication is also typical. It's typical for academic publishing in general. Parts of this process went much faster than usual. Three months for a full evaluation at QJE is fast. And once accepted, my paper sped through the JME. Another year or two in the pipeline between acceptance and publication is typical.
Note most journals count average time to decision. But what matters is average time to publication of the papers they publish, and what really counts is average time to publication in the journal system as a whole.
Lessons and thoughts?
- Academic journal publication is not a useful part of communication among researchers or the communication between research and policy.
Anyone doing research on zero bound in new-Keynesian models in the last 4 years, and carrying on this conversation, interacted with the working paper version of my paper (if at all), not the published version. Any work relying only on published research is hopelessly out of date.
Interest rates lifted off the zero bound quite a while ago, so in the policy conversation this publication at best goes into the shelf of ideas to be revisited if the next recession repeats the last one with an extended period of zero interest rates , and if we see repeated invocation of the rather magical predictions of new-Keynesian models to cure it. If the next recession is a stagflation or a sovereign debt crisis, you're on your own.
Rather than means of communication,
- Journal publications have become the archive,
the ark, the library, the place where final, and perfected versions of papers are carved in stone for future generations. (Some lucky papers that make it to graduate reading lists more than 5-10 years after their impact will be read in final form, but not most.)
And this paper is perfected. The comments of nine very sharp reviewers and three thoughtful editors have improved it substantially, along with at dozens of drafts. Papers are a conversation, and it does take a village. The paper also benefitted from extensive comments at workshops, and several long email conversations with colleagues.
The passage of time has helped as well. When I go back to a paper after 6 months to a year, I find all sorts of things that can be clearer. Moreover, in the time between first submission and last revision, I wrote four new papers in the same line, and insights from those permeate back to this one.
So, in the end, though the basic points are the same, the exposition is much better. It's a haiku. Every word counts.
But such perfection comes at a big cost, in the time of editors and referees, my time, and most of all the cost that the conversation has now moved on.
The sum length of nine referee reports, four reports by three editors, is much longer than the paper. Each one did a serious job, and clearly spent at least a day or two reading the paper and writing thoughtful comments. Moreover, though the reports were excellent, past the first three they by and large made the same points. Was all this effort really worthwhile? I think below on how to economize on referee time.
Of course, for younger people
- Journal articles are a branding and sorting device.
Many institutions give tenure, chairs, raises, and other professional advancement based at least in part on numbers and placement of publications. For that purpose, timeliness of publication is less of a problem, but with a six year tenure clock at many places and five year lags, timeliness of acceptance the quality rating is a big problem. The sorting and branding function isn't working that well either. But having journals outsource quality evaluation was always an imperfect institution. Maybe we should just have star ratings instead -- seriously, start up a website devoted to crowd-sourcing working paper evaluation. Or, perhaps tenure committees will have to actually start reading papers. I don't think the journals see this as their main function either. They're set up to publish papers, not judge people's tenure, so improving journals as a tenure granting mechanism will be a hard sell.
There is some good news that this data point represents, relative to state of journal publishing 15-20 years ago. (See Glenn Ellison's superb "The slowdown in the economics publishing process," JSTOR, undated, one of my proudest moments as a JPE editor.)
- Journals are doing fewer rounds, more desk rejection, more one round and up or out.
Journals had gotten in to a rut of asking for round after round of revisions. Now there is a strong ethic of either rejecting the paper, or doing one round of revisions and then either publishing with minor changes or not. Related,
- Journal editors are more decisive.
Journal editors have become, well editors. The referees provide advice, but the editor thinks about it, decides which advice is good and not, and makes the final call. Editors used to defer decisions to referees, which is part of the reason why there were endless revisions. This change is very good. Referees have little incentive to bring the process to a close, and they don't see the pipeline of papers to the journal. They are not in a good position to find the right balance of perfection and timeliness.
In my case, editors were very active. The referees wrote thoughtful reports, but largely made similar points. In fact, the strongest advice to reject came at the JME. But the AER and QJE editors were not impressed in the end by the paper, and the JME editor was.
So, with this state of affairs in mind, how might we all work to improve journals and the publication process?
I will take for granted that greater speed, and making journals more effective at communication and not just archiving and ranking is important. For one reason, to the extent that they continue to lose the communication function, people won't send articles there. Already you can notice that after tenure, more and more economists start publishing in conference volumes, invited papers, edited volumes, and other outlets. (blogs!) The fraction willing to take on this labor of love for journal publication declines quickly with age. Research productivity and creativity does not take quite such a parallel decline. (I hope!)
Always the free-market economist, I note that conference volumes, edited volumes, and solicited papers in regular journals seem to be healthy and increasing, which is a natural response to journal slowdown. This is a way to get papers in print more quickly. In the early days of the internet I had a rule never to publish in volumes, as they disappeared to library shelves and could not be found electronically. Now many of them have solved that problem. The NBER macro annual and Carnegie-Rochester conferences are good examples. The Review of Finance editor recently solicited my "Macro-finance" essay which therefore sped through publication. My active editors are also often taking a more active role in soliciting promising working papers. This helps to break the editor-to-paper match. But this isn't an ideal state of affairs either. Conference volumes tend towards commissioned work. Original work by people out of the social network of the conference organizers and editors has a tough time. (The market responds, organize more conferences.)
Suggestion one:
- Adopt the golden rule of refereeing
Around any economist cocktail party, there is a lot of whining that journals should do x y and z to speed things up. I start with what you and I can do. It is: do unto others as you would have them do unto you. If you complain about slow journals, well, how quickly do you turn around reports?
My recommendation, which is the rule I try to follow: Answer the email within a day. Spend an hour or two with the paper, and decide if you will referee it or not. If not, say so that day. If you can give a quick reaction behind your reason, that helps editors. And suggest a few other referees. Often editors aren't completely up to date on just who has written what and who is an ideal fit. If you're not the ideal fit, then help the editor by finding a better fit, and do it right a way.
If you agree to do a report, do it within a week. If you can't do it this week, you're not likely to be able to do it 5 weeks from now, and say no.
More suggestions:
- Reuse referee reports
Do we really need nine referee reports to evaluate one paper? I always offer editors of journals to whom I send rejected papers the option of using the existing referee reports, along with my response as to how I have incorporated or not their suggestions. Nobody has ever taken me up on this offer. Why not? Especially now that editors are making more decisions? Some people mistakenly view publication as a semi-judicial proceeding, and authors have a "right" to new opinions. Sorry, journals are there to publish papers.
Why not open refereeing? The report, and author's response, go to a public repository that others can see. Why not let anyone comment on papers? Authors can respond. Often the editor doesn't know who the best person is to referee a paper. Maybe a conference discussant has a good insight. At least one official reviewer could benefit from collecting such information. Some science journals do this.
Some people would hate this. OK, but perhaps that should be a choice. Fast and public, or slow and private.
While we're at it, what about
- Simultaneous submission. Competition (heavens!)
Journals insist that you only send to one journal at a time. And then wait a year or more to hear what they want to do with it. Especially now that we are moving towards the editor-centric system, and the central question is a match with editor's tastes, why not let journal editors share reviewer advice and compete for who wants to publish it? By essentially eliminating the sequential search for a sympathetic editor, this could speed up the process substantially.
I don't know why lower-ranked journals put up with this. It's the way that the top journals get the order flow of best papers. Why doesn't another journal say, you can send it to us at the same time as you send it to the AER. We'll respect their priority, but if they don't want it we will have first right. The AER almost does this with its field journals. But the JME could get more better papers faster by competing on this dimension.
The journals say they do this to preserve the value of their reviewer time. But with shared or open reviews, that argument falls apart.
We advocate competition elsewhere. Why not in our own profession?
Update: An email correspondent brings up a good point:
- Journals should be the forum where competing views are hashed out.
They should be part of the "process of formalizing well argued different points of views -- not refereeing "the truth." We dont know the truth. But hopefully get closer to it by arguing. [In public, and in the journals] The neverending refereeing [and editing and publishing] process is shutting down the conversation."
When I read well argued papers that I disagree with, I tend to write "I disagree with just about everything in this paper. But it's a well-argued case for a common point of view. If my devastating report does not convince the author, the paper should be published, and I should write up my objections as a response paper."
0 yorum:
Post a Comment