Thursday, November 11, 2010

Is chemistry employment cyclical or counter-cyclical?

Original graph: C&EN, 11/1/10, 88 (44), pp 38.
Modified by Chemjobber
  Anon111010653a asks "is chemistry somewhat sheltered from economic crises, or do they underrepresent unemployed chemists?" in regards to the graph (modified by CJ) to the left from Susan Ainsworth's article in last week's issue of C&EN.

My take home from the graph is this: you can see the last three recessions (90-91, 2001, and 12/2007-) represented in red in the modified graph.* In my experience, this is evidence that modern unemployment in chemistry is at least somewhat correlated with the overall unemployment picture.

Anon's real question is "what does the gap mean?", i.e. "Why isn't the ACS unemployment rate also 10% right now?" My answer: honestly, I don't know. It's certainly not the "sheltering" hypothesis in my opinion. Whether the ACS salary survey undermeasures unemployment is, of course, a different question.

*Important to remember: unemployment is a lagging indicator of the economy.


  1. CJ, I think that we need to be careful with these comparisons, as you point out at the end of your post. The ACS's survey is almost certainly faulty. Is it faulty in the same way the national surveys are faulty? Ugh. I wish we knew. And, unfortunately, it's the best info that we have.

  2. Would it be a more prudent to compare the ACS figures to college graduates?

    Check this out. This data below ends in Sept 09, and would compare to the end of the figure above.

    Compared to college grads ACS members appear to have done better in 09, but it was less than 1% better in terms of the unemployment rate.

  3. The ACS numbers come from them sending out questionnaires to a random sampling of their member base. So a) they're only talking to ACS members and b) it may not be a truly representative sample of those members. I think it probably skews low in regard to unemployment numbers, since out of work chemists are probably less likely to be ACS members. (At least that seems to be the current vibe.)

    The other thing to remember it takes them over a year to analyze and publish the data after they've collected it. So these unemployment numbers are really representative of last March. As in, 2009.

    So yeah, crystal of NaCl there.

  4. "I think it probably skews low in regard to unemployment numbers, since out of work chemists are probably less likely to be ACS members."

    The same holds for chemists who have 'retrained' away from Chemistry, no? What I mean is that the number of unemployed chemists is a dynamic quantity.

  5. Sadly, what does anyone have to gain from putting out reasonable data, ironically, relating to the employment of scientists?

    shudder ...

  6. @2:47pm
    If reasonable employment data was avaliable it would slow down the rate of entry into PhD programs. If supply of chemists decrease then so will chemist unemployment/underemployment.

    But then maybe it wouldn't have that affect. When I went to graduate school I didn't look up unemployment rates of chemists. I just knew I liked chemistry, so off I blindly went to grad school like so many others.

  7. Maybe a better comparison would be a rate of layoffs of chemists versus other fields. That data might be available from the Department of Labor. It might not since chemistry is such a small field that noone outside of it really understands (so it is clumped with other non-relevant fields). It would be interesting to see how our field has been holding up against others the past few years.

  8. The ACS data is not worthy of the name. It conforms to no standards of survey methods or response rates (anything below 80% is highly suspect), and wouldn't pass preliminary peer review in any field. This is outside of the fact that it only reports ACS members (already a highly skewed population of working chemists). Those numbers are so suspect they shouldn't even be reported! The canard about "even bad data is better than no data" may or may not be true...

  9. "It conforms to no standards of survey methods or response rates (anything below 80% is highly suspect)"

    That's a fairly bold statement, E-s. Can you elaborate?

  10. The sampling frame (members only) is already highly biased, but moreover the response rate is nothing short of abysmal. Federal funding agencies require an 80% response rate for any fielded survey; below this you need extensive extra analysis, and the results are highly suspect. Most sampling pros would refuse to even report a survey with a response rate as low as the yearly ACS survey.

    A good ref is “Survey Research Methods” (Fowler, 2009).

  11. Good job, Eka-Si.

    There is a systemic error that is involved in the comparison of the two data trends by Dr. Jobber:

    The Bureau of Labor Statistics is measuring whether "anyone" is employed or unemployed. So, for example, an electrician, baker, lawyer or chemist who can't find a job in his/her trained field, but who is working in another one, e.g., Starbucks, Wal-Mart, etc. still counts as "employed".

    On the other hand, summarizing the criticisms of the ACS data: (1) it's selective for ACS members; (2) ACS members who give up on Chemistry will likely drop out of the ACS + seek employment outside of the field and (3) the response rate isn't acceptable.

    Furthermore, the national workforce consists of a large proportion of "service" jobs, i.e. those that can't be outsourced - in contrast to science, engineering and semi-skilled manufacturing professions. And so, logically one would expect the unemployment rate for _outsourceable_ workers to be higher than for _non_-outsourceable workers. Shouldn’t this be reflected in Dr. Jobber's comparison? Or is the number of really long-term unemployed so significant as to overwhelm this trend?

  12. Hi, Eka:

    Thanks for pointing me to a very interesting and handy book. Here's what it has to say on the matter, in the concluding paragraph of the relevant chapter:

    Nonresponse As a Source of Error

    Nonresponse is a problematic, important source of survey error. Table 4.2 demonstrates the great potential of nonresponse to effect results. Yet, although we can calculate a rate of response, we usually do not know the effect on nonresponse on data. The Keeter et. al (2006) study illustrates a survey with a comparatively low response rate producing results that are very similar to one with a much higher response rate, but the Groves (2006) analysis reminds us not to be complacent.

    The key problem is that we usually lack good data about when nonresponse is and is not likely to be biased with respect to the content of the survey. Certainly one unintended positive effect of the increasing concerns about response rates is to increase pressure on researchers to collect data about nonresponse bias when they can. However, it is hard to do. In the absence of such data, perhaps the strongest argument for efforts to maximize response rates is credibility. When response rates are high, there is only a small potential for error due to nonresponse. When response rates are low, there is great potential for important error; critics of the survey results have a reasonable basis on which to say the data are not credible.

  13. The ACS employment numbers always appear rosier than reality. First it is a survey of ACS members who return surveys not trained chemists trying to do chemistry for a living. About half the ACS members are tenured professors who last time I looked are never unemployed. Those guys seriously pollute any unemployment numbers.

    The detail we all need is: 1) Industrial chemists wanting to be employed as chemists but who are unemployed or in temporary non chemistry professions (i.e., employed as greeters at WalMart). 2) Unemployed chemistry graduates still waiting to be employed by industry or academia but unable to find any chemistry jobs other than post docs. 3) Chemists over age 50 wanting to be employed in a chemistry job, but forced into early or even late retirement or “consulting” for lack of chemistry jobs. 4) Chemists who want to be employed in the chemistry profession but left the profession entirely to find employment. 5) Chemists who want permanent employment in a chemistry job but who employed by temp agencies. 6) Chemists who are working at a wage much lower that the historical prevailing degree adjusted wage for chemists. 7) The answers to the prior three questions broken down by chemistry specialty, age, location and degree level.

    Now, good luck finding any of that information on any ACS web site! I will gladly contribute to a fund to get this kind of information for all us to see.

  14. "About half the ACS members are tenured professors who last time I looked are never unemployed."

    A3:08a, this is somewhat off. I believe the splits are closer to 70/30. Nonetheless, the issue remains.

  15. Chemjobber,

    You are, of course correct, but the real question is what % of those ACS surveys come from tenured professors? Just how polluted is the ACS data by these guys?

    What we really need to see is the % of industrial and nontenured chemists who want to be employed as chemists but are either unemployed, under employed (post doc), have taken work outside of their profession (sales clerk), forced to retire or forced to consult.

  16. A7:27a:

    I believe it's about the same, but I don't know. I will check with my sources.

    Re underemployment: it's worth pointing out that the ACS salary survey does separate out part-time workers and postdocs.

  17. I consider chemists who want to work in a traditional full time job as a chemist, but who must work as post docs, temp employees, piece meal workers, catch-as-can workers, consultants and part time employees all under employed chemists.

    I would like to know the % of chemists in this group.

  18. A7:27a:

    I would, too. Surely you understand the general difficulty of these measurements; nevertheless, I will be attempting something of the sort in the near future, I hope.