Counting linguistics job ads and dissertations
« previous post | next post »
Heidi and I both posted recently about the job market in linguistics (my post, and her collaborative post with Shannon Bischoff). There was some discussion in the comments about whether the numbers were truly comparable and whether we should base conclusions on just one year of data. This post attempts to make amends, as it were, by setting Linguist List job data from the last five years alongside the Proquest dissertation data that Shannon collected:
Update (2009-06-09): See this post for arguably more accurate counts for dissertations.
A few comments:
- For syntax, phonology, and semantics, the disparity between the number of dissertations and the number of jobs is somewhat troubling, but we should not read it as an indication that these fields are hopeless to get into. Rather, it suggests that such students would do well to combine their work with experimental or computational research, to put themselves in the running for a wider range of jobs. This interdisciplinary approach is likely to produce stronger research too!
- For the leftmost three fields, the disparity runs in the opposite direction, making them look like paradise for new graduates. This could be misleading, though. If Proquest encourages more specific categorization for these very general fields, then that could drastically affect the counts we have.
- The orange and blue bars are close for a number of fields. My take on this: those fields don't have enough researchers. Increased competition might do them good.
Here are the numbers:
| Field | Jobs ads | Dissertations | Ratio of ads to dissertations |
|---|---|---|---|
| Applied linguistics | 475 | 38 | 12.5 : 1 |
| Cognitive science | 47 | 20 | 2.35 : 1 |
| Computational | 517 | 110 | 4.7 : 1 |
| Psycholinguistics | 89 | 56 | 1.6 : 1 |
| Phonetics | 126 | 79 | 1.6 : 1 |
| Phonology | 146 | 290 | 0.5 : 1 |
| Syntax | 142 | 749 | 0.2 : 1 |
| Semantics | 116 | 262 | 0.4 : 1 |
| Sociolinguistics | 149 | 152 | 1 : 1 |
Geoffrey K. Pullum said,
January 29, 2009 @ 11:48 am
I want to make a brief comment about this, addressed to linguistics doctoral students. I speak as one who simply would not listen when people said "You can never become an academic in linguistics because job vacancies are just too scarce." For me, the prospect of not spending a life in linguistics was unpleasant to think about, so I refused to think about it. This is the way to think about that enormous blue column over Syntax, the worst piece of data in the bar graph above. If you think it is saying you can't make it, you are overlooking something important. You are missing a key variable. What's being left out is just how good you are determined to be.
This is the way to look at that blue spike: The numbers show there are only about 20 job vacancies in syntax for each 100 syntax dissertations. But that is great for you! It means that on average you only have to be better than the next four syntax dissertation-writers who come along, and (assuming broadly merit-based hiring) one of those jobs has your name on it! You know that at every conference you've seen papers presented by students who are worse than you aim to be. You have seen real turkeys mumbling away up there at the mike, presenting analyses that don't fly, or even waddle. You can damn sure be better than any random four of them. Can't you?
Don't look at the gross numbers. Look at the field in the competition, and where you think you can place. You only need to stay out of the bottom 80 percent! You can do that! If you work and train harder, you can probably do better than that, and get one of the really good jobs. It isn't about how many dissertations are being written and accepted; it's about how many dissertations are being written by people who just aren't going to look as good as you are when it comes time for a job talk. There are lots of them. All you need to do is think harder than they did, and write better, and give better talks. And you can do that. Yes, you can. Even in syntax (and in the other areas it's easier), you can get a job teaching and researching in an area you find fascinating. Go for it.
Eric Baković said,
January 29, 2009 @ 12:41 pm
Ditto to what Geoff says above.
And to add another point that I hope is encouraging to current linguistics doctoral students: having been involved in several searches since arriving at my current institution, I can confidently say that the number of people you will be competing with for each individual job will be nowhere near on the order of e.g. 750 for syntax, 290 for phonology, and so on. (That said, I think that the relative number of people you will be competing with for each subfield probably correlates well with the above graph — that is, the competition in syntax is probably numerically stiffer than it is for phonology, and so on.)
I do still wonder whether the numbers of applicants for individual jobs, at least at public institutions like mine, are a matter of public record. It'd be nice to have at least a smattering of data of that type to put the graph above into some perspective. I would look into this myself, but I'm too busy/lazy (bazy? lusy?) to do so. Sorry.
Rubrick said,
January 29, 2009 @ 2:07 pm
I wish more people would adopt Geoff's attitude. I mean, just think how wonderful it would be if everyone was in the top 20%!
Laura Kalin said,
January 29, 2009 @ 2:18 pm
Thank you, Geoff and Eric! Those are very encouraging words :)
To be honest, as an aspiring syntactician, my first impression of the table presented in this post was entirely positive: "Wow! I am definitely going into the right field. Syntax is so exciting… look how much more research there is going on in syntax compared to phonology (about a third as much), semantics (about a quarter as much), and phonetics (about an eighth as much)."
I guess at this point in my linguistics career (I'll be entering grad school this coming fall, pending acceptances…), I'm much more interested in the number of dissertations than job postings, because the job-search seems so far away. To me, more dissertations equates to more innovation, more resources, and more interesting avenues to pursue.
Oh how I love being young and naive ;)
Jonathan said,
January 29, 2009 @ 2:31 pm
Could it be that linguistics departments in PhD-granting instutitions have more syntacticians who want their students to follow in their footsteps, but that these same departments cannot absorb that number of PhDs? Or could some of the computational jobs be open to specialists in syntax too, providing a hidden escape valve?
Sili said,
January 29, 2009 @ 2:55 pm
Prof Pullum,
Norwegian has just started offering discount flight to Glasgow twice weekly.
Would you mind terribly if I popped by and smacked you a big wet one? (I'd bring along a coupla lobsters for you to juggle, of course.)
Unfortunately, I'm one of those mumbling dodos who just phoned it in most of the time, but hey-oh …
Chris said,
January 29, 2009 @ 3:34 pm
Hmmmm, the computational linguistics numbers look odd to me. I find it hard to believe there were more phonology dissertations than computational linguistics (I could be wrong). How were the dissertations counted? If a CS grad student did a machine learning diss using natural language corpora for training, was that counted? The annual Human Language Technologies conference is quite large and attracts mostly CS not linguists (shame on us for not attending this, btw) and it seems to me they should be counted. And they're the ones getting the jobs (try this little experiment: put "linguistics" into Monster.com then "Natural Language Processing").
Rick said,
January 29, 2009 @ 5:46 pm
I left a comment about knowledge engineering jobs on the first post in this series, having worked as a linguist in industry for some time now, and I realized that one thing that has not been considered (at least with respect to industry positions) is that one ad does not necessarily equal one position. One of the companies I worked at hired primarily linguists for our group, topping out at around 40 individuals, who were generally hired in groups of 5-7. This is not typical, but is another factor to keep in mind.
Mark Liberman said,
January 29, 2009 @ 5:55 pm
Chris: I find it hard to believe there were more phonology dissertations than computational linguistics (I could be wrong). How were the dissertations counted?
As I understand it, the dissertations that were counted were those that ProQuest identifies as in "Linguistics". I'm not sure how this category is assigned — given the numbers, it surely does leave out a lot of what would in an alternate universe be called "linguistics" dissertations, but in our world are done in departments of computer science, psychology, electrical engineering, speech & hearing, etc. Note also that in some countries, Phonetics is counted as a separate discipline rather than as a sub-field of linguistics. All this is the result of the historical balkanization of the discipline defined logically as the study of speech and language. (For some further discussion, see A series of unfortunate events", 1/26/2005; slides for a later version of the cited talk are here.)
Of course, the job-category counts in the graph also leave out many jobs that are available to people trained in psycholinguistics or speech technology or NLP or clinical applications, but were not advertised on Linguist List. It's hard to know what the balance between dissertations left out and jobs left out is. At least in recent times, though, the market in phonetics and computational linguistics has been favorable to job seekers, especially those who are willing to look outside of academia.
Mark Liberman said,
January 29, 2009 @ 7:57 pm
A bit of checking verifies that the dissertation counts in "interdisciplinary" areas are indeed likely to be way low.
For example, today Ryan McDonald gave a talk at Penn on the topic "Mining and Summarizing Opinions from the Web". His 2006 Penn dissertation was "Discriminative Training and Spanning Tree Algorithms for Dependency Parsing"; it was surely an example of computational linguistics, but it wasn't in ProQuest's inventory under Linguistics.
Erwin Chan's 2008 dissertation was "Structures and Distributions in Morphology Induction"; also surely an example of computational linguistics, and also not indexed by ProQuest as Linguistics.
This suggests that CL dissertations outside of linguistics departments are generally not indexed by ProQuest under Linguistics, and so I'd guess that the count of relevant dissertations in that category is an order of magnitude too small.
On the other hand, I've just checked the last couple of non-academic "linguistics" jobs for which I've written letters of recommendation for linguistics students. And neither of them was advertised on Linguist List, as far as I can tell.
In a better world, a professional organization like the LSA would arrange to do the (significant but feasible amount of) work required to compile accurate statistics of this kind. All that said, I believe that the numbers in the graph probably point to the truth in some approximate sense — there are (or recently have been) more jobs than qualified candidates in computational linguistics, for example.
James said,
January 29, 2009 @ 10:33 pm
http://www.indeed.com/salary?q1=linguist&l1=&tm=1
This graph from indeed.com shows the salaries of linguists over the last year. Apparently we're doing pretty well as the average salary is said to be $130,000 (USD). But then when you look at the jobs listed below, few of them involve doing what we think of as "linguistics". Instead, they seem mostly to be translators.
You get a less satisfying average salary if you search for "linguistics professor" ($52K) and "computational linguist" ($68K).
Of course, I have no idea how reliable this really is. The site says that "Salary Search is based on an index of salary information extracted from over 50 million job postings from thousands of unique sources over the last 12 months. Many job descriptions don't contain salary information, but there are enough that do to produce statistically significant median salaries for millions of keywords".
John Sylak said,
January 30, 2009 @ 12:04 am
Another fact that I try to keep in mind is that "computational linguistics" is not a subfield in the same sense that syntax, semantics, phonetics, phonology, morphology, and pragmatics are. I'm not denying that it is a subfield (it is, especially because of NLP), but the fact is that if you do syntax, phonology, semantics, phonetics, or morphology using computational methods, you are in a certain sense doing computational linguistics. This is a good fact to keep in mind because, in my understanding, it means you can keep your theoretical passion and just gain some computational skills (e.g. coding, better understanding of math and stats) to make yourself more marketable.
Also, I haven't heard much talk of applied linguistics jobs. Perhaps it's because Language Log seems to be more popular among the academic crowd and the term "linguistics" is often used as a shorthand for "academic theoretical linguistics," but I like to think that there are applied linguistics jobs that are very (if not possibly more) satisfying than traditional teaching/research positions at various universities. After all, many of the jobs that applied linguistics takes on (such as second language teaching, language documentation and description, speech pathology, etc.) provide the raw data that theoretical linguistics uses in its research and could not do without. In addition, I think its safe to say that most linguists end up coming into the field after experience with some L2, and thus those teachers play a vital role for the field. Plus, there's the fact that since it is "applied" linguistics, one may even have more of a concrete result to show for their work at the end of the day (a student who can now speak X language to whatever degree, a person cured of a speech impediment, a descriptive grammar, etc.).
Chris said,
January 30, 2009 @ 1:43 am
Mark, thanks for the follow-up. I think you are right about the job/qualified candidate ratio. It's worth pointing out, I think, that non-academic computational linguistics jobs are likely to be more adversely affected by the current "economic downturn" than academic linguistics jobs. VC funding, the life-blood of industry CL, is virtually non-existent right now.
James, I think you've hit on an interesting factoid. Whenever I do a job search on Monster.com, CareerBuilder.com or some such site, I ALWAYS use the word "linguistics" and avoid the word "linguist" because of the prevalence of non-relevant jobs for translators and ESL teachers. It's not a perfect strategy, but it helps.
Also, note that many of those high salary "linguist" jobs are for temporary government contract related projects for Arabic, Korean and Chinese speakers who have a US security clearance, and those are few and far between. Scarcity of talent drives up wages.
kyle gorman said,
January 30, 2009 @ 3:08 pm
came here to say what mark already did. the same thing i think is true for "applied linguistics" and "cognitive science" positions: they're generally filled by people without "linguistics" on their diploma.
echoing geoff: you should only go to grad school if you're distressed by the idea of leaving your subject.
kyle gorman said,
January 30, 2009 @ 3:12 pm
@john: i think your story is a bit too rosy. use of computing does not a computational linguist make. papers at the ACL conferences (the premiere conferences for CL researchers) are expected to show a useful product using novel algorithms, which even years of learning the tools may not teach you to do. even if you implement somebody else's algorithmically-complex ideas in the aid of doing theoretical research (an example from my research: parsing or tagging a corpus of german child speech to look for morphological errors) that's far below the ACL line, and it's hard to imagine a computational linguistics hire that doesn't have an ACL publication.
lillyjaynes said,
February 10, 2010 @ 2:37 pm
Regarding non-computational, non-applied, non-industry-related linguistics:
Theoretical linguists rank only amid society's fluff. That is why there are so few "jobs". Of what value is fluff? Any linguistics "job" (excepting jobs in industry/applied linguistics where, as John Sylak observed, "one may even have…a concrete result to show for their work at the end of the day") is derivative and inconsequential. After all, as Geoff says in regards to the requisite labor for such employment, "All you need to do is think …and write…and give…talks." What does it matter if no one's directly, concretely, meaningfully impacted by what you think or write or yammer on about at some podium? What concrete results depend upon your input? Your value is negligible, as evidenced by the market's negligible need of your services. What difference do you make? You're all just talking to yourselves. You're a closed circuit.
If every theoretical linguist on the face of the earth suddenly disappeared, POOF!, would society even notice? Or care?
Oh, well, of course we'd miss those commentaries on NPR's Fresh Air! One can't imagine the world spinning without those.