Wednesday, June 10, 2009

(Why) Are Database Conferences in Decline?

This week, I took a look at the submission counts and acceptance rates for key database conferences. I started out by updating the statistics that Peter Apers has been maintaining for some time with the most recent data available. Here, I report my findings.

I view SIGMOD and VLDB as the flagship database conferences, and I believe that most members of the community will agree.

From 1993 to ca. 2002, SIGMOD received numbers of submissions in the range from 200 to 300, in a zigzag-like pattern. Then the number of submissions grew to the 400-to-500 range, peaking at 480 in 2007. The last couple of years have seen numbers that are similar to the 2004 and 2005 level of 431.

Now consider VLDB. The trend for the period 1993 to 2001 is that of a slight increase from around 300 to around 350, with 1999 being an outlier (390). Then follows a few years of marked increases in submissions. The last three years have each seen some 550 submissions. At 626 submissions, 2006 is an outlier.

The VLDB acceptance rates have been in the range from 13 to 19 percent since 1993, with only two years being outside the 14 to 18 percent range. There are no clear patterns. Similarly, the SIGMOD acceptance rates have generally been in the range from 14 to 18 percent during the past decade, with no clear trends toward increase or decrease.

So VLDB and SIGMOD are similarly selective. SIGMOD is seeing a slight decline in submissions, while VLDB is seeing a slight increase. VLDB is attracting the larger number of submissions.

Here, recent submission counts for several conferences are graphed:





I generally have no explanations for the various fluctuations in the submissions. Or for why SIGMOD seems to be decreasing slightly while VLDB seems to be increasing slightly.The relatively low figure for SIGMOD 2002 may be due to an overlap between the EDBT reviewing period and the SIGMOD submission date that year. And the locations of the conferences may be a factor. If anyone has good explanations for the figures, I would love to hear about it.

Next, I view ICDE and EDBT as the second-most prestigious database conferences.

For ICDE, the numbers of submissions generally were in the 250 to 300 range from 1993 to 2002. Then followed a gradual increase to 521 in 2005. The last three years have seen from about 550 to 650 submissions. It is noted that ICDE had few relatively submissions in 2006, the year where VLDB had relatively many (-is this a coincidence?).

The ICDE acceptance rate is characterized by a declining trend: From the 20-25 percent range early on to a situation where four of the last six acceptance rates have been in the range from 12 to 14 percent, with the two other being 19 percent.

EDBT also exhibits a slight growth from 1994 to 2002, ending slightly above 200. Then follows a strong growth that takes EDBT to 352 in 2006. And there was a slight decline in 2008. The acceptance rate has been at 17 percent for all years, with the exception of two (16 and 14 percent).

So we are seeing a pattern of slight growth until ca. 2002, then a significant growth followed by a decline or a slowing growth the last couple of years. I wonder why?

Here are the acceptance rates that correspond to the submission counts given above:




Finally, I want to mention CIKM. This conference has grown steadily over the years; and in 2009, CIKM received 847 submissions, which is the highest for a database research conference ever (yes, CIKM spans broader than databases). CIKM's acceptance rate has been as high as 40 percent, but it seems to have stabilized in the range from 15 to 18 percent.

It would be interesting to compare with non-database conferences as well.

Let me end by observing that there is of course more to a conference than the number of submissions and the acceptance rate. For example, the expectations associated with a conference wrt. the topics it accepts and its "toughness," affect the quality of the conference and the submission behavior, e.g., leading to self-refereeing, where researchers choose to send only their best papers to the "toughest" conferences.

PS. I chose to manage the conference statistics using Fusion Tables.

4 comments:

  1. Christian,

    While this analysis is interesting by itself I'd be also interested in seeing, in terms of the same metrics, what has happened with journals in the last 10 years. For instance, has the number of submissions to the top journals declined proportionally to the increase of submissions to the top conferences? What are the "typical" acceptance rates for the journals? (Granted one has to consider revisions, which do not exist in
    the conference system.)

    I think CIKM is being seen by many as a more inter-disciplinary (by definition) venue, which could explain the migration from core-DB venues (if they can still be characterized as such). BTW, has EDBT's acceptance rate really been such an outlier this year?

    I'm also particularly curious about one thing (which I don't think one could measure though), I think reviewers are becoming harsher (less patient?) and less-technical (too much to read?) in recent years.

    ReplyDelete
  2. Mario,

    Good observations. Database journal acceptance rates are hard to get, for technical and probably other reasons. I believe that they are higher than those of the conferences.

    CIKM has three tracks. It might be good to compare with the DB track instead of the three tracks combined, but I did not have those figures. I do believe that all the numbers I report are accurate.

    Finally, I do think the field would be better served by the review process if we could find a way to better reward good reviewing.

    ReplyDelete
  3. TODS publishes many statistics, but not acceptance rates.
    http://tods.acm.org/Statistics.html
    I'll ask to see if we can publish them or if
    there is merit to doing so.

    I agree that interdisciplinary conferences attract more submissions, such as WWW, which had an acceptance rate of 8% in 2006.

    One factor that may help to explain variations in individual conference submissions is location. CIKM 2008 was in Napa Valley. Perhaps
    that played a role...

    curtis

    ReplyDelete
  4. Curtis,

    I added a location column to the table. Based only on a quick look at a couple of conferences, I did not see a location pattern. that is not to say that there is one.

    ReplyDelete