WEDNESDAY, JULY 29, 2009, 11:00am: Parallel Session 102:  STEM Pressures from Birth to Globalization

Jac Vennix and Daniel Horschel chaired.

Full Report:
Workforce Modeling for the National Institutes of Health (NIH), by J. Chris White (presenter), Margaret Rush,and Dr. Walter Schaffer
Mr. White provided an overview of a project he is working on at NIH with Dr. Walter Schaffer regarding the NIH pool of funded Principal Investigators.  The motivation for the project came from looking at the average age of funded investigators for NIH projects from 1980 to 2005.  The average age has shifted from about 39 years old in 1980 to about 49 years old in 2005, with a much flatter distribution (i.e., more older investigators).

The purpose of the project was to develop a model to help understand how this aging of the NIH workforce occurred and what possible actions could be taken to mitigate the effects.  Mr. White described a basic system dynamics aging model that was used as an agent in the SimBLOX Model Creator simulation framework, which is an icon-based, drag-drop-connect interface for working with multiple models at a macro level.

Results from the simulation tracked historical results extremely closely, with an RMS value of 0.38% being the largest deviation.  The model was then used to project forward the “as is” situation.  This meant that the demographics and quantity of new investigators coming in remained the same, as did the demographics and quantity of older investigators leaving.

Several different options were analyzed for demographics/quantity of incoming new investigators and exiting investigators.  Three main demographic profiles were used:  baseline (current demographics), ESI 35 (a demographic that made the average age of incoming new investigators 35 years old), and ESI 38 (a demographic that made the average age of incoming new investigators 38 years old).  Four different quantities were used for the incoming new investigators.  This yielded 12 simulations.  Tables below show the average age of the pool of investigators 5 years out (year 2011) and 10 years out (year 2016).  The yellow values are the as-is scenario.  Clearly, as more younger investigators are brought in, the average age of the pool shifts significantly.  However, details showed that sometimes these shifts were too aggressive.

The conclusion of this project was that the system dynamics-based aging chain was a valid, logical, and consistent method for modeling the pool of investigators, even though it was very simple.  It proved better than previous models and looked to be a good foundation from which to build further models.  The current model, as simple as it was (with no real feedback loops), was still able to show the potential impacts of certain policies.  While it didn’t show an optimal policy (mainly because a desirable shape of the age distribution for the investigators is not known), the basic model was enough to show that several policy options under consideration were undesirable.

The successful results of this first project has led to a follow-on project with NIH in which the basic model is expanding to incorporate important external variables (e.g., annual NIH budget, success rates for grant applications), as well as feedback loops (e.g., as the budget changes, how do people respond).

America Disrupted:  Dynamics of the Technical Capability Crisis, by Dan Sturtevant (presenter) and J. Bradley Morrison

Mr. Sturtevant started with the basic question, “Is American competitiveness in engineering moving forward?”  Data show that while the percentage of students getting degrees has increased over time, the percentage of engineering degrees has decreased from a peak of almost 8% in 1985 to about 4.5% in 2005.  We also rank in the bottom half of the list of twelfth grade average scores for math and physics. 

If STEM jobs offer higher wages on average, with higher satisfaction ratings on average, and low unemployment, these types of jobs should be viewed as attractive.  In essence, the current situation is a violation of the simple law of supply and demand.  There must be something else going on.

Mr. Sturtevant found that, because educators are often paid low wages compared to other jobs, teacher quality has been decreasing since the 1950s (i.e., there are not as many teachers that are certified/qualified to teach STEM subjects).  This drives the attractiveness of teaching STEM subjects down, which leads to lower quality teachers, which creates poor engineers in the end.  An interesting point came up that before the women’s liberation movement, STEM teachers were predominantly male.  It was decided that a better forum for this type of discussion was over several beers.

A fact about math education is that it is a chain of learning:  once you drop off it is very difficult to get back on track, because each level builds on the previous level.  Miss a few levels and there is a lot of rework that needs to be done.  Consequently, many teachers choose to teach because they DO NOT have to use math.  How can they be good math teachers?  Considering that there is a strong correlation between teacher quality and student performance, how can there be good engineers?  Below is the final conceptual model for which simulations were run.

Through various simulations, Mr. Sturtevant was able to show several “tipping points” in system behavior. One possible fix, he noted, is to pay teachers the same as engineers.  This would lead to more teachers of higher quality, which would lead to more and better engineers.

Key messages:

  1.  We systematically invest least in whatever skills the economy desires most.
  2. Teacher quality is a long-term constraint to growth for the high-tech sector.
  3. Math knowledge is structurally unique.  We must create an intact pipeline because the low-quality teacher is the bottleneck.  There is a trade-off between immediate response and high leverage.
  4. We could move the system into a fundamentally better pattern of behavior, but only after long delays.  Most policy proposals on the table today will not get us past the tipping point allowing us to do so.

Science, Technology, Engineering, and Mathematics (STEM) Career Attractiveness, by Andjelka Kelic (presenter) and Aldo Zagonel
Ms. Kelic’s presentation focused on career attractiveness for STEM.  Our goal is for people to enter and stay in the STEM workforce.  How can this be done?  One of the problems is that there are competing viewpoints.  Some say that there is a shortage of STEM jobs.  Others say there is a shortage of STEM workers.

Ms. Kelic showed causal loop diagrams from each perspective to try to gain understanding of how these two seemingly incompatible views could co-exist.

It turns out both “camps” are correct.  Employers want to keep costs low, so the employers tend to remove older employees that have higher salaries and replace them with younger and less experienced employees who command lower salaries.  A “revolving door” phenomenon occurs as people are “kicked out” after a few years and replaced.  These displaced employees then tend to go to other fields.  There is a perceived shortage of STEM jobs because older people are being removed, and there is a perceived shortage of STEM workers because enough new, younger workers cannot be found.

Ms. Kelic then described the current state of the system. A simulation model developed by Ms. Kelic and others at Sandia Labs was used to run various simulations to test the potential fixes.

The best option was to apply all four policies simultaneously.  With this scenario, it would take about 20-25 years to get back to the 1980 levels for the size of the STEM workforce, the level of K-12 STEM literacy, the attractiveness of STEM jobs, and the domestic job availability (as opposed to off-shoring jobs).

J. Chris White