Actuarial Exams (Gender Issue)

The BBQ Pit
1

One central feature of an actuarial career - at the outset, at least - is a series (about 9 these days) of extremely difficult exams. These must be taken and passed to progress in the field. But they are extremely difficult - the percentage of people who pass a given exam tends to be in the range of 30% to 50%. And the average IQ of the test-takers is pretty high, so indeed it’s safe to say that the exams are quite difficult.

The percentage of correct answers needed to pass varies from exam to exam, but generally tends to be in the range of 60%. It is near impossible to get all the answers right.

The early exams are all completely multiple choice - 5 choices per question. Later exams have some multiple choice and some essays.

So here’s the issue. In the good old days (meaning a few years ago) there was a “guessing adjustment”. Which means that anyone who had no clue as to the answer could leave it blank and pick up 1/5 point. The thinking was that there was no reason to leave the marks in such cases subject to random luck. At the same time, it was worthwhile to guess in cases where you could narrow down the answer to less than 5 choices - the odds in such cases increased to the point where they were worth more than the 1/5 point.

But now it is changed. There is no adjustment - everyone must guess. This means that differences in how one’s random guesses happen to play out can make the difference between passing and failing. To see to what extent this is true, I computed the following table:

10	11	12	13	14	15

0 10.7% 8.6% 6.9% 5.5% 4.4% 3.5%
1 26.8% 23.6% 20.6% 17.9% 15.4% 13.2%
2 30.2% 29.5% 28.3% 26.8% 25.0% 23.1%
3 20.1% 22.1% 23.6% 24.6% 25.0% 25.0%
4 8.8% 11.1% 13.3% 15.4% 17.2% 18.8%
5+ 3.3% 5.0% 7.3% 9.9% 13.0% 16.4%

I’m not sure how readable it will be, but the columns represent the amount of guessed answers, and the rows the amount that are right by pure random fluctuation. To use the case in which 10 answers are pure guesses (there are generally about 40 questions), there is a 37.5% chance of getting 0 or 1 right, and a 32.2% chance of getting 3 or more right. This means that two groups of about 1/3 of the population each will be separated by two answers based on pure random luck. And 2 correct answers on a 40 question exam are 5% of the total. If you assume - as I do - that the majority of the marks are clustered somewhere near the pass mark - either slightly below or slightly above, this would imply that a large percentage of the determination between passing and failing is determined by luck. In light of the difficulty of the exam, the amount of preparation needed, and the implication for a person’s career, I think this is unacceptable.

So why do the actuarial societies do this? The answer is women. Psycologists have found that women tend to shy away from risks in these situations. Which makes them apt to settle for the 1/5 of a point “guessing adjustment” even in situations in which they can narrow down the field of answers. The removal of the guessing adjustment is an attempt to “level the field”.

I think is not a valid justification. The fact that a disproportionate percentage of the people who happen to have an illogical fear of risk-taking are women is not a significant point, IMHO. Everyone has an even chance under the “guessing adjustment” system. Anyone who happens to have a problem functioning within the parameters of of the fairest, most logical system has their own problem. And whether more men than women have this type of problem or vice versa is irrelevant. Or should be, anyway. Alas.

2

Izzy, although I was chairman of the Examination Committee for the Casualty society 20 years ago, I do not have much to contribute. The Casualty exams do not have the same structure, and I never got involved in the early, joint exams.

This Message Board doesn’t allow us to set our spacing. I have found that one can get some control of the spacing by separaring adjacent numbers with a series of dots…

I certainly agree with your point. I had been unaware of the change in scoring (or maybe I just didn’t pay attention.) It’s not even clear whether the change helps women. For those who felt they did fairly well, not guessing may have been a good strategy.

BTW, maybe the old system was a subtle way of testing students’ knowledge of probability theory. Those who didn’t understand enough to guess when they had narrowed down the field of answers got properly dinged. :slight_smile:

3

10        11       12       13       14       15
0 10.7%  8.6%   6.9%  5.5%   4.4%   3.5%
1 26.8% 23.6% 20.6% 17.9% 15.4% 13.2%
2 30.2% 29.5% 28.3% 26.8% 25.0% 23.1%
3 20.1% 22.1% 23.6% 24.6% 25.0% 25.0%
4  8.8%  11.1% 13.3% 15.4% 17.2% 18.8%
5+ 3.3%   5.0%  7.3%   9.9% 13.0% 16.4%

4 

	10	11	12	13	14	15
0	10.7%	8.6%	6.9%	5.5%	4.4%	3.5%
1	26.8%	23.6%	20.6%	17.9%	15.4%	13.2%
2	30.2%	29.5%	28.3%	26.8%	25.0%	23.1%
3	20.1%	22.1%	23.6%	24.6%	25.0%	25.0%
4	8.8%	11.1%	13.3%	15.4%	17.2%	18.8%
5+	3.3%	5.0%	7.3%	9.9%	13.0%	16.4%
5

Bloody American actuaries and their multiple choice exams. Bunch of Larry Lightweights.

And nine exams? Nine? You cheeky fucks. There’s nine exams here just to get through the easy first set of exams.

My solution? Can the multiple choice - it doesn’t test much anyway.

pan

6

So, some of the actuary students can’t figure out when it’d be statistically better to guess? Not good form for an actuary student, surely.

7

This…is a pit rant?

I feel like I should have my notebook out.

j

8

I should clarify that I am not saying that the new system “helps women” in a sense that it gives them an edge. Rather it helps people who have a problem with multiple choice tests, by stopping them from hurting themselves - a disproportionate percentage of such people are thought to be women. In a sense, the old system “discriminated” against those who harmed themselves, the curent system discriminates against random people. The SOA has adopted for the latter in the interests of harming a proportionate percentage of men & women.

BTW, Kabbes, the SOA has had various amounts of tests at different times - at some point they also had quite a lot. The current amount is the resut of rolling up all the smaller ones into fewer, bigger ones. Guess the Brits couldn’t with with a lot of material at a time.

And as for multiple choice not testing anything - no true actuary would make such a silly statement. :smiley:

Tansu, it’s actually not so simple as you put it - in some cases it’s better to take the credit even if you can narrow down the answers. Suppose, for example, that you are pretty sure (based on historical precedent) that the pass mark won’t be higher than 60%. And you are absolutely sure that you’ve successfully answered 20 correctly (out of 40). By taking the 1/5 point on each of the remaining 20 questions, you add 4 points and guarantee yourself a passing mark. By guessing - even if you can narrow down some of the answers - you leave open the possiblity of failure.

Interestingly, this is part of a larger principle of “Utility Theory”, which concerns the fact that a percentage chance of getting (or losing) something is not necessarily equal to the percentage x the potential reward (or loss). This principle is the underlying foundation of the very concept of insurance, and as such is itself part of one of the actuarial exams. (I believe december was alluding to this in the final sentence of his third paragraph).

But the SOA is obviously concerned about the cases where this is not so. I would suggest that this phenomenon is not the result of a lack of knowledge about the best strategy, but rather to an ingrained psychological hangup, to which - contrary to what many may believe - even actuaries may be subject.

9

For a while the Casualty Actuarial Society had 10 exams. Then they “partiationed” some of them, so that, in effect, there were 14 or 15. That was terrible. I used to do some teaching, and most of the male students were already partially bald.

With only 9 exams now, things should be back to normal, I hope.

10

Izzy, no way can you on one hand tell me that some of your early exams are wholy multiple choice and on the other hand claim that they test more. I don’t care how much material is involved - multiple choice mostly tests your ability to work backwards and forwards at the same time and your ability to pick the right answer out from a selection of possibilities - not necessarily your ability to start from scratch with no clues.

Not to mention the fact that there is nothing for method - the most important part of the test in my opinion. And essay questions are a total non-starter.

What’s the point in testing a larger quantum of material if your test only scratches the surface of understanding?

No wonder we have people here who get fed up with trying to become FIA and go off to qualify in America instead.

Feh and fie, I say, feh and fie.

pan

(Hey, I’m just teasing you really. I’m sure merkin actuaries must also know something. And we can’t all be the elite :p)

11

How do women do on the tests compared to men anyway?

By the way, when I was in college, I took a muliple choice test that was scored as follows: Next to each answer choice, you put the probability that you thought that choice was correct. To figure out the amount of credit you got for each question, they took the probability you assigned to the correct answer and put it on a logarhythmic curve. If you put 100% next to the correct answer, you’d get one point. If you put 20% next to the correct answer, you got zero points. If you put 0% next to the correct answer, you got negative infinity points. It made for some interesting strategizing.

12

So now you have to arbitrarily decide how “sure” you are too? People can’t even figure out the simplest of probabilities - how the hell are they going to do this?

13

Just by vaguely observing the list, my impression is that women constitutute around 30% to 50% of new Fellows of the Casualty Actuarial Society.

14

Although (as I’m sure december appreciates) that statistic doesn’t tell us a great deal without knowing:

a) how many females are joining as new students; and
b) how those women would have done based on a different system.

From the (British) Institute of Actuaries handbook 2000/2001 (I seem to have lost 2001/2002), as at 31/7/2000 29% of our students and 14% of our fellows were women. This is on a non-multiple choice basis :wink:

Unfortunately though this statistic gives another little bit of the whole, by itself it is probably even more prone to distortion than december’s, since it doesn’t say anything about new fellows and it still says nothing about new students, let alone the proportions by sex of students that make it to fellows.

However I did learn that prior to 1920 women weren’t allowed into the Institute at all. Fancy.

pan

15

I sure can tell you that. In fact, I will go further. Your reasoning strikes me as something that might be applicable to high school level math, or something of a similar elementary level. By the time you get to the more advanced levels of actuarial science, the principles and formulas are complicated enough that there is no way whatsoever to “work backwards”. Either you know how to approach the issue or you don’t. And there’s no “method” to test - if you know the answer it’s because you know the method. (The only advantage multiple choice does give is that if you get an answer that is not one of the choices, you know you have to rework the problem. But this is a small advantage, and besides, the answers are frequently given as ranges).

If this seriously represents your assessment, it would indeed tend to suggest that actuarial exams in the UK are on a much more basic level than those in the US and Canada. (So there!)

As for statistics of males and females, I am not under the impression that their decision was based on observations about the actuarial exams themselves. (In fact, I don’t think they know who is male or female). Rather, the idea that females are reluctant to guess is widely thought by psycologists in general (I had previously come across this notion when studying psycology) and the actuarial society bought into it as applied to themselves.

16

Also, British female actuaries (Fellowettes? Fellowesses?) are cuter. That’s because they’re younger. The sexiness factor demonstrates that our exams are harder than yours.

17

One of the problems is your initial assumption, IzzyR. You assume that the majority of people hover just above or below the passing mark. If the tests were designed that way, then yes, it would be through sheer luck that some people passed and others didn’t.
However, that’s only on the assumption that the peak of the bell curve falls squarely on the 60% line. This doesn’t have to be the case. If I wanted to design a test which was difficult, I’d make it so that the curve hit the 55% line. Likewise, an easy test I’d try to design to hit at the 65% line. That way, more than 50% would pass.

But no matter how you score it, people that guess and hit near the 60% line will have luck determine their outcome. There’s nothing to do about it.

18

I wouldn’t assume that the peak of the curve falls squarely on the 60% line. But as the pass percentage (i.e. the percentage of passers) tends to be between 30% & 50%, it’s not unreasonable to believe that the peak of the distribution is not far off the pass mark, whatever it is.

Well, true, for “people that guess”. But by eliminating the guessing adjustment you are forcing everyone to guess, and making everyone subject to luck.