Bioinformatics: Recent Trends in Programs, Placements and Job Opportunities Report to the Alfred P. Sloan Foundation Grant # 2002-5-58 BCMB June 2004 Grant C. Black Department of Economics Andrew Young School of Policy Studies Georgia State University [email protected] Paula E. Stephan Department of Economics Andrew Young School of Policy Studies Georgia State University [email protected] We would like to thank Eleanor Babco and Fred Fox for their help and encouragement and Donald Hemenway of AAAS for providing electronic copies of ads from Science. Asmaa El- Ganainy provided research assistance on this project. Cynthia Blasdell and Ruxue Xia assisted with the design of the web survey. EXECUTIVE SUMMARY The current study collected information on conditions in the bioinformatics labor market during the period 2000-2004. A three-pronged approach was used. First, all 74 known academic programs offering degrees in bioinformatics were surveyed in the spring of 2003 with a response rate of 59.5 percent; second, position announcements were analyzed for all bioinformatics- related ads placed in Science during the three-year period ending in 2003; position announcements on two internet sites were examined as well. Finally, a pilot study of seven firms was also conducted, exploring the firms’ perception of the state of the bioinformatics labor market. Results of the study are compared to those from a 1999 survey of academic programs and a 1998 study of position announcements. We find: (1) A dramatic increase in the number of academic training programs as well as the number of individuals enrolled in programs. To wit, during the period between 1999 and 2003 the number of known programs grew from 21 to 74 and the number of students enrolled in bioinformatics programs that responded to our survey grew from 169 to 881. (2) Little change, except at the master’s level, between the 1999 survey and the 2003 survey in the number of reported placements from academic programs. (3) That the number of newly minted individuals trained in bioinformatics during the five-year period is fairly small relative to the number of position announcements. (4) A decrease in the number of advertised positions regardless of the level of experience or level of training. By 2003 the number of positions advertised in i Science was below the 1997 number; the number of positions posted on www.bioplanet.com declined between 2002 and 2003. The seven company interviews provide qualitative support regarding the softening of the market. (5) A change in the source of demand, with a shift to academe and away from industry, especially for those with a Ph.D. (6) That salaries in the bioinformatics market have risen during the 1999 to 2004 period but that the substantial wage premium associated with bioinformatics jobs in the past has virtually disappeared and is unlikely to be offered under the current market conditions. Our analysis leads us to conclude that the bioinformatics labor market has gone through considerable change in recent years. In the context of a relatively fixed supply of specialists, salaries for individuals with skills in bioinformatics soared in the 1990s. Strong demand and the concern that the “seed corn was being eaten” led to the creation of numerous new training programs in bioinformatics. These training programs are now beginning to generate graduates. Many of these graduates assumed they would go to work in industry, yet positions in industry appear to be on the decline, and many of the positions that are available are for individuals with considerable experience. The strongest area of demand in recent years has been from academe, seeking faculty to staff new programs and to broaden research expertise. Unless conditions in industry change dramatically in the next few years, it is likely that many trainees from these programs will have difficulty finding jobs in industry. ii I. Introduction The vast amount of biological data that has become available sine the early 1990s has made computational methodologies in the life sciences increasingly important in research. This in turn has created the need for scientists with interdisciplinary skills in computational science and biology and has led to the emergence of bioinformatics as a distinct field. Little is known about the labor market in bioinformatics, including training opportunities, except for the previous work by Stephan and Black (1999a, 1999b, 2000). In 1998 they analyzed position announcements placed in Science as a measure of demand, and in 1999 they surveyed all known academic programs related to bioinformatics to collect information on characteristics of the training programs, starting salaries of recent graduates, and the identity of institutions hiring these graduates. The information collected by Stephan and Black in their earlier studies is now at least five years old. In a field that is fairly stable, data of this age can often present a reasonably reliable indication of current conditions; in an emerging field like bioinformatics, such data give little indication of the market’s current condition. For instance, approximately 50 more institutions have begun to offer training programs since the 1999 survey. In addition, the media has continued to report that the field still offers opportunities and that a shortage of qualified individuals persists (Chabrow 2004; Henry 2001, 2002; Park 2001). And programs and articles cite a National Science Foundation estimate that the United States will have 20,000 bioinformatics jobs to fill by the year 2005.1 At the same time the biotechnology industry has faced a considerable economic downturn and restructuring. In effect, there has been little 1 The NSF projection is quoted on a number of web sites. See, for example, the joint bioinformatics track through the departments of biological sciences and computer science at San Jose State University (http://www.cs.sjsu.edu/faculty/khuri/Bioinformatics/description.html), an article in Forbes, March 15, 2001, (http://www.forbes.com/2001/03/15/0315malone.html), and an article in Time, April 29, 2002, (http://www.time.com/time/business/printout/0,8816,233967,00.html). 1 accurate information on the changing state of the bioinformatics market, including its current condition, over the past five years. To address this dearth of information regarding the bioinformatics labor market, this report presents findings from a recent survey of academic training programs and an extensive analysis of recent position announcements. All academic programs included in Stephan and Black’s original survey as well as all programs identified since the 1999 survey were invited to participate in a new survey designed to collect information on the characteristics of current training programs and the experiences of their recent graduates. The analysis of position announcements includes all bioinformatics related ads placed in Science in 2000 through 2002, as well as recent electronic ads listed on select internet employment boards. A pilot study of seven firms was also conducted. The seven firm interviews had two goals: (1) to provide preliminary information concerning the perception of firms regarding the state of the bioinformatics labor market and (2) to provide information concerning the best way to approach firms if the study were to be expanded to a larger, more representative set of firms. II. Academic Training Programs In March 2003, the 74 academic institutions with known bioinformatics-related training programs at the undergraduate, graduate, and/or postdoctorate level were asked to participate in a web survey.2 Appendix A lists the institutions targeted for this survey and identifies those that replied. The survey was designed to elicit standardized information across institutions while offering a convenient web-based format to encourage participation and enhance data collection and analysis. Appendix B includes a copy of the survey questionnaire. Forty-four institutions 2 Programs were identified from (1) institutions targeted in the 1999 survey, (2) professional science master’s programs in bioinformatics sponsored by the Alfred P. Sloan Foundation, (3) extensive internet searches, and (4) talking with individuals involved in established bioinformatics training programs. 2 (59.5 percent) responded to a part or all of the survey; 30 did not.3 Targeted institutions were initially contacted via email to provide information and to request their participation in the survey. Approximately one week later, institutions were sent an email providing a hyperlink to the web survey and instructions. Three weeks later, a postcard was mailed to every targeted institution, thanking those that had participated and encouraging non-respondents to participate. Continued non-respondents were then contacted by telephone to increase the response rate. Eighteen of the twenty-one known programs targeted in the 1999 survey replied to the 2003 survey (85.7 percent response rate); the response rate was considerably lower for the newer programs with 26 of the 53 replying (49.1 percent). One of the 27 non-respondents in the latter group specifically declined to participate due to the newness of the program; another non- respondent has a Ph.D. track in genomics and bioinformatics that was only implemented in 2001- 2002. The three institutions that participated in the 1999 survey that did not reply in 2003 are Boston University, The University of Connecticut, and Northwestern University. At the time of the 1999 survey, only one of these institutions (Northwestern) had a formal program in bioinformatics. Boston University had plans to initiate a formal program at the M.S. and Ph.D. levels in the fall of 1999. We conclude that the survey was reasonably successful at collecting information on established programs in bioinformatics. Data collected from the survey provide less accurate counts regarding newly established programs. This should not affect the accuracy of certain measures, such as graduation and placement counts, since many of the non-responding programs are too new to have graduated anyone; it does bias downward our counts of the numbers of individuals enrolled in these programs. 3 The 59.5 percent response rate is comparable to that for the 1999 survey of 61.9 percent. The 1999 survey information was enhanced by web-based research on non-respondents. The large number of programs surveyed in 2003 limited our ability to do web-based research for the 30 non-respondents. 3 A. Characteristics of Programs The number of academic training programs related to bioinformatics has grown substantially from the late 1990s to the early 2000s. Figure 1 shows the number of training programs by degree level, comparing the 1999 data with that for 2003. The highest growth was at the master’s level, with over six times as many programs in 2003 as in 1999. This is in part due to the creation since the late 1990s of twelve professional science master’s programs in bioinformatics, all initially funded by the Alfred P. Sloan Foundation.4 The number of doctoral programs grew to 24 in 2003 from 9 in 1999. There are still relatively few programs at the bachelor’s level though the number has more than tripled between 1999 and 2003. The small number of postdoctoral training programs undoubtedly reflects in part the informal nature of many postdoctorate training positions. Figure 1 Number of Training Programs 35 30 25 20 15 10 5 0 Bachelor's Master's Doctorate Postdoctorate 1999 2003 4 Nine of these twelve Sloan programs replied to the survey. 4 As seen in Figure 1, most of the academic programs are relatively new. Of the 10 bachelor’s programs, 4 started since 2000 and another 2 began in 1999. Among the master’s programs reporting start dates, 7 started in 1999, another 7 in 2000, 4 in 2001, 4 in 2002, and 1 in 2003. Seven Ph.D. programs were created in 1999 alone and seven more have been created since 2000. The reader is reminded that these counts are biased downward given that new programs were less likely to reply than were older programs. Unlike the other programs, little growth has occurred at the postdoctorate level since 1999. Institutions participating in the survey were asked if additional levels of training were currently being considered. Sixteen institutions (36.4 percent of the responding institutions) indicated that they had no intention of expanding their current programs and another fourteen offered no indication of their plans. Over one third of the responding institutions, however, were considering additional training at different degree levels than currently offered. Five indicated interest in creating Ph.D. programs in bioinformatics or computational biology.5 Four were considering graduate certificate programs to provide graduate level training without a full master’s program. Possible master’s training was being considered by four institutions, and three institutions were considering bachelor’s training. Three institutions were considering additional training but did not specify at what level. Institutions were asked the degree to which they thought their programs met student demand for bioinformatics/computational biology training at their institution. Twenty-five institutions indicated their programs mostly or completely met demand for training at their institution. Five institutions reported that they somewhat met demand, while only one institution indicated their program did not at all meet demand. Several institutions did not reply to the 5 Institutions may have indicated more than one level of training being considered so the counts by degree level sum to more than sixteen, the number of institutions reporting that they were considering the addition of new levels of training in their programs. 5 question. When asked about the strength of demand for bioinformatics programs at their institution, 30 of the responding institutions perceived demand to be moderate or great, with 22 of those describing demand as great. Comments from respondents describing demand for training at their institution and their consideration of adding additional levels of training to their current programs included: ●“An externally funded pre-doctoral training program is pressingly needed.” ● “There is clear demand for a curriculum in bioinformatics at the undergraduate level that we have yet to address but are now working on.” ● “Demand is strongest at the undergraduate level.” ● “There is no way that we could accept all 170 grad applicants who applied this year – 16 was all we had room for. We expect the pool to be even bigger next year….” ● “We are not addressing one great area of demand – training master’s level students or training at a pre-master’s certificate level…. We have decided to concentrate on Ph.D.-level training, and do not satisfy this great need.” ● “We are not really meeting demand of those who are not in our MS degree program. We are actually working hard to keep the classes from over-enrollment with those in other programs.” ● “Students in other programs would like to have more training in this area.” ● “Since [our current program] is only a summer program and since we are required to attract at least half of our students from outside the state, [our program] cannot meet the growing demand for an undergraduate training program in bioinformatics and computational biology.” The establishment of these training programs has not come without challenges. Thirty of the forty-four institutions described a problem or challenge when asked to describe the most difficult challenge faced in establishing their program. The most common challenge was a spectrum of issues broadly related to administration and university support of a program. For example, several programs had difficulty receiving approval from university administrators. 6 Others faced limited administrative resources and difficulties structuring programs across academic units. The next most frequently reported challenges revolved around program funding, securing faculty involvement, and curriculum development. Across these challenges runs the underlying thread of interdisciplinarity. The interdisciplinary nature of bioinformatics programs makes for practical hurdles in their creation within the traditional university environment. It can be challenging to recruit faculty, allocate course credits and students, obtain resources, assign administrative responsibilities, and develop curriculum across departments and colleges at a given university. The difficulties surrounding the development of interdisciplinary programs are not symptomatic only of newer programs but also of early training programs, arguably contributing to the sluggish response of universities to industry’s growing demand for graduates with bioinformatics training in the 1990s (Stephan and Black 1999). B. Financial Support No dominant pattern of financial support is evident from the surveyed bioinformatics programs. Seven institutions reported that their bachelor’s bioinformatics program is fully funded internally. Fourteen of the master’s programs were reported to be at least 95 percent internally funded. Six institutions fully funded their doctoral programs internally. Only one postdoctoral program reported to be fully funded internally. More commonly, many institutions have relied, to some extent, on external sources of funding to support their bioinformatics programs. Table 1 lists the reported sources of external funding for responding bioinformatics programs. External funding is less prevalent for bachelor’s and master’s programs. The breadth of external support is also less broad for these programs. For institutions reporting sources of 7