Sunday, April 29, 2007

IITs and women: A recap


This post is going to be long. Very long. Don't complain that I didn't warn you!

Also, this post is unlikely to have anything that I have not said before. But I still want to write this post because I want to get all the arguments and data -- and links, too! -- in one place.

A couple of things before I start. First, I want to say a big "Thank You" to all of you who came by and commented on the previous two posts. The number of comments on this post is a record for this humble blog. The comments on the bleg post, where many of you gave me some data on the proportion of women in your engineering class, were interesting, and I found them very useful in refining my own thinking on this issue. And finally, my special thanks go to Bill for writing a very thoughtful post on JEE's bias against women.

Second, just as low numbers of women in science and engineering (which become even lower in IITs as well as at higher levels of education) bother me, I am sure they bother many of you too. Given these low numbers, and given that we would like women's participation in engineering (and other fields, too!) to grow, we are interested in finding the underlying causes, so that suitable remedies can be found. The lack of agreement between Vivek (and some commenters) and me (and some other commenters) is essentially in the latter realm -- of understanding the reasons for women's under-representation; but the gap in our positions will continue to shrink as we keep this discussion going, and as we get more data to look at. On some observations, there's already some agreement: that our society doesn't value women's education as much as it does men's, and that high quality educational opportunities (IITs, AIIMS, NLSUI, etc.) are too few.

With the preliminaries out of the way, let's move forward.

1. First, the numbers

Vivek has unearthed an interesting document -- a report from the Indian National Science Academy (INSA) -- about the status of women in science (and engineering) in India [[I too had linked to it in a recent post]; it covers both higher education and employment. For the present purposes, the key data are in Tables 2, 3 and 4 (on pages 7-8) in Part I (pdf), covering college enrollment.

Among the college-goers, only 39.4 percent were women; more pertinently, just 21 percent of the engineering students were women. The only fields that do worse (in terms of women's participation) are veterinary science (20.9 percent), law (20 percent) and agriculture (17.4 percent). In most other fields (science, commerce, arts, etc), women's participation was around the overall average: commerce and management (36.5), science (39.4), medicine (44), and arts (44.2). Education is the only field in which women (51.2 percent) outnumbered men. All this is from 2001.

If you look at the trend from 1996 to 2001, the fraction of women increased in all fields. While percent of women went from 36 to 39.4 across all fields, it went from 16 to 21.5 in engineering. An encouraging sign.

Let's move down to Classes X and XII. At the level of Class XII, boys outnumber girls, but only by a slim margin in CBSE. While it is possible that in other exam boards boys may outnumber girls by a bigger margin, it is also true that girls outperform boys in these exams. So, when you look at the stream of students coming out of Class XII, boys may enjoy, at best, a small advantage in numbers.

But at the college level, the INSA report shows clearly that men have a 6-to-4 edge; indicating that more girls than boys drop out of the education pipe at this important transition point.

Let us now turn to the data specific to engineering. The subjects that a student has to take at Class XII to get into engineering are mathematics, physics and chemistry (MPC). It is possible that the boys/girls ratio gets skewed in favour of boys in classes with MPC. Further, in most states, there are many, many single-sex schools. It is possible that boys' schools in those states outnumber those meant for girls, introducing some additional skew in favour of the boys. I can't put a figure on the exact extent of this skew in Class XII with MPC.

When you come to engineering itself, the distribution of women is uneven. Some disciplines, such as computer science and electronics engineering, seem to have a fairly high fraction of women (30 to 50 percent); see the comments on the previous post (specifically, here, here, and here). On the other hand, in 'traditional' engineering subjects such as mechanical engineering or my own field, metallurgical engineering, there are typically very few women.

Given the trend during 1996-2001 when women went from 16 percent to 21.5 percent, and given that the fields in which women participate in large numbers are also the fields that have been experiencing higher growth rates in output, I wouldn't be surprised if the fraction of women in engineering is close to (or, even above) 25% now.

[Aside: A couple of commenters have alluded to reservation for girls in engineering colleges [Orissa, AP]. Frankly, I was unaware of it. In these states, girls do go into disciplines such as mechanical engineering and metallurgical engineering in not insignificant numbers!]

Given that engineering is a discipline that continues to grow quite rapidly, and given that girls compete successfully to get into the more rapidly growing disciplines, we come to the rather happy conclusion that the number of women is increasing at a faster pace (albeit from a low base) than that of men.

The bottomline

The proportion of women decreases as they move from Class XII (where they have perhaps a small disadvantage in numbers) to college (where they are outnumbered 4 to 6). In college, their proportion is smaller in engineering (21.5 % in 2001, and perhaps 25% now). Their proportion is much greater than this average in fast-growing engineering disciplines than in traditional ones.

The society's impact is fairly self-evident in this trend: women's education receives less family attention and support, and certain engineering disciplines are seen as less desirable for women. Vivek made essentially these points in his first post on this subject; while he has been right about these societal effects, I believe that he overestimates their influence, leading him to deny that JEE might have a built-in bias against women. It is this we turn to next.

* * *

2. Let's look at JEE now, shall we?

Women constitute between 2 and 8 percent of IITs undergraduate student body. In 2005, this fraction was 6.3 percent (321/5092). In 2006, however, "there [was] only marginal increase in successful female candidates as compared to JEE-2005." The total number of successful candidates, on the other hand, went up from 5092 to 6343. Clearly, the fraction of women among the successful candidates in 2006 was likely about 5 percent. [Here too, I would like to see some studies giving figures for all the IITs since their inception.]

Thus, in a random selection of students in 2005 or 2006, women are three to four times less likely to be found in IITs than in other engineering colleges. This, by itself, does not constitute an evidence for (a plausible) bias in JEE against women. For that, we have to show that the fraction of women in the eligible pool of candidates is far higher than that in the IITs.

[A related argument is the following. Think about the trends in the number of women in engineeing colleges: it is increasing quite rapidly (albeit from a small base). What about the IITs? Their absolute number hasn't gone beyond 500! ]

Since IITs are premier engineering institutions in the country, its students would be from the pool of top students in Class XII. What do we know about this cohort? Consider the CBSE results for Class X in 2006: among those with 90% marks and above, there were nearly as many girls as boys, and among those with 95% marks and above, girls outnumbered boys 9 to 8. [I would love to get my hands on a detailed study of Class XII results!]

A second metric, which Falstaff prefers, is the fraction of women among those taking the JEE. In 2005, it was about 15 percent (30,000 out of 200,000), and in 2006, it was about 20 percent (60,000 out of 300,000). [Given the past history of women's success in JEE, Falstaff also wonders if women self-select out of this exam -- particularly in the second attempt. I too think it's possible; but again, there's no data.]

A third metric is the fraction of women among those entering engineering colleges in a given year. From the INSA report, we know that as of 2001 it was around 21 percent; I would put the current figure at around 25%.

If you use any of these metrics, the fraction of women in IITs is far lower than what one would reasonably expect from a consideration of the eligible pool of candidates. Does this establish a plausible bias in JEE against women? I think so.

Does this bias represent a second level of handicap over and above that imposed on women by our society? I think so.

[Heck, even a very badly skewed model proposed by Falstaff leads to an estimate of 13 % for the disadvantaged group!]

* * *

3. Is JEE (in its current format) really necessary?

In principle, there is no need for any entrance exam if we had some way of normalizing the students' performance in exams conducted by India's many different Boards. The students' (normalized) marks themselves could be used for rank-ordering them, and they could then be assigned to go to different institutions based on their rank and their choice.

Thus, the use of an entrance exam (other than, of course, as a mechanism for normalizing across many exam boards) needs a clear articulation of the need for it. For example, one should be able to say, "engineering requires certain special skills, and not only does JEE test students on those special skills, the board exams do a poor job of it."

Remember, this utility of JEE cannot just be asserted. It should be backed up with data from serious, large and long-term studies, which I don't believe the IITs have done [in case you are interested, here is my rant on this topic]. But I know of one study (and I don't know how broad a data-set it used), that produced some very interesting findings:

  • ‘‘There is a strong correlation between the marks of Classes X, XII and the CGPA during B Tech. The correlation factor is close to 1.’’ This means, the chances of a good student in school doing well in B Tech is almost 100 per cent.
  • ‘‘There is little correlation between marks in Class X, Class XII and AIR.’’ That is, good performers in school are not likely to get good AIR in the present selection system.
  • ‘‘There is little correlation between AIR and CGPA.’’ This means, toppers in the JEE are not at the top during their B Tech programme.

This (probably limited) study is a strong indictment of JEE as a filter-exam. It indicates that the exam cannot claim any superiority in discovering those special people with even more special skills which are to be honed further in the IITs. I'm not sure if we can define such special skills in a sufficiently precise and testable manner that will be useful for all the programs that IITs offer: from computer science to mining to agricultural engineering. [The US universities seem to be happy with standardized but relatively general-purpose exams such as SAT for admitting students into all kinds of undergrad programs.]

Thus, I believe JEE should be replaced with an exam that has as its primary purpose the simple job of normalizing the scores across exam boards. Yes, this does imply that the questions must be standardized (pre-tested on a sample of students, with more difficult questions carrying a greater weight). More importantly, it also implies that it must stick to the syllabus that's actually used in Classes XI and XII -- both in breadth and depth.

[Aside: Some -- notably, RC -- have argued that since JEE is a 'selection machine' that has no prior conception of differences among candidates, it's not biased. This is a pretty lax yardstick with which to judge the fairness of a selection machine; by this yardstick, even a lottery is a fair machine! More seriously, however, we do have many other selection machines -- er, entrance exams -- that do a far better job of allowing women to succeed.]

* * *

4. How will the new format help?

The main problem with JEE -- that makes it biased against women, students from poor families and rural areas -- is that it is pitched at such a high level that students need external help through coaching centres. Some have asked how the need for coaching affects women. I see at least two ways: (a) coaching centres don't come cheap, and to the extent families value women's education less than they do men's, women get less money to spend on coaching, implying that many women who could benefit from coaching get insufficient amounts -- or none -- of it; and, (b) 'reputed' coaching centres are far away (Kota, Hyderabad), or they have classes during early mornings or late evenings (and one centre in Mumbai even tried all-night classes!): all these are clearly unfriendly for women.

Would a modified version of JEE really help in getting more women into the IITs? To the extent that performance in JEE tracks that in CBSE exams (for example), it is clear that more -- many, many more! -- women will make it to the IITs.

* * *

5. How about your objection about the coaching centres?

Will the new form of JEE reduce the corrosive role of coaching centres? Several (including Vivek and Tabula Rasa) seem convinced that it won't. The argument goes like this: "since the competition is so intense, even a small advantage due to these centres will help; thus, if boys have an advantage now, they will be able to carry it over to the new regime too." Not so fast! Consider this: schools themselves can train students for tackling both board exam and the new JEE; teachers can offer extra classes for motivated students.

I am willing to concede that the clamor for every possible advantage -- gettable at a price, of course -- will ensure that coaching centres will not go away. But, the extent of help can certainly be minimized by an exam that tests students at their level.

[Why doesn't this -- coaching at schools by the regular teachers -- happen now? In the current set up, with an extremely tough and high-level JEE, teachers and schools are unable to play this role, because they are ill-equipped to handle the college-level material demanded by JEE. (In fact, the format of JEE invites students to devalue their regular school education!)]

* * *

6. Anything else?

Yes, in its normalizing avatar, the new, improvedTM JEE can also become a uniter. It can replace all the other entrance exams. Currently, an aspiring engineering student needs to take the current JEE, AIEE, and one or two state-level entrance exams as well. They all have different emphasis, and require different kinds of preparation. In the new scheme of things, JEE is all there is. Our youth will be spared the pain of juggling conflicting requirements of many entrance exams.

What about concern that solving JEE level questions is a joy for some students? I suggest these exalted souls participate in Olympiads, the intellectual counterpart of the Olympic Games. When they do, it will be because of the love and joy of solving some super-problems; not because some frustrated professors thought up a crazy way to terrorize hundreds of thousands of innocent students. Though these Olympians do not need any incentives, we can still think of some: I suggest that the IITs set aside, say, 500 seats under the "Olympiad quota".

The rest of the students can get on with their lives in relative peace.

* * *

As I said, I have collected here several different strands that were scattered in various posts and comments. Thank you for reading all the way down here; you deserve an award!

30 Comments:

  1. Rahul Siddharthan said...

    The point about the correlation between board exam performance and IIT CGPA, and the lack of correlation between the above and the IIT-JEE AIR, is very interesting.

    About the coaching centres, I disagree with you that teachers in schools would be able to do the job with a "standardised" JEE. The JEE, of any kind, must focus on problem solving and testing the understanding of basic concepts, which our schools don't teach (or only a few good teachers do). But the good students can, and do, pick it up on their own. (I don't know the current situation, but the CBSE text books for classes 11 and 12 that were newly introduced in my time -- 1990-91 -- were written by some of the best scientists in India; many teachers complained about them, and they certainly had their rough spots, but an interested student could learn a lot from them.)

    I don't expect coaching centres to go away; I'd be satisfied if a bright student can get through without their help.

  2. Vivek Kumar said...

    @Abi: A quote from the Indian Express article to which you have linked:

    "And the questions for the JEE will be strictly from the syllabus and standard of the Class XII, encouraging a synergy between Class XII examinations and the JEE."

    Syllabus and standard.. both. Isn't this what you are asking for from a new & improved JEE?

    How should the JEE format change beyond this point?

    And shouldn't we be giving the current format of JEE some time in order to generate meaningful data trends? 2007 is only the second year of this JEE.

    One more thing to explore: BITS Pilani used to normalize scores from different Boards across India. Then they ditched it in favour of an entrance exam (AIEEE now, I think). May be you could get some information on normalization and the reasons for ditching it from some current/past students.

    Fact I know about BITS' normalization in 1997: Whatever they did, a fellow with 94% in CBSE was placed at 1000th place or so on the waiting list. I had 90%, and I didn't even make it to the waiting list. Then we both got into IIT.

    Long post, so more thoughts in the days to come. Thanks for keeping it going.

  3. Vivek Kumar said...

    "A reliable procedure known as normalization is used to bring students from different boards on the same scale of comparison. Admission is competitive and the minimum normalized percentage required for admission is higher than 97%."

    From here

    No more details are provided.

  4. Saurabh said...

    1. It is not correct to use CBSE success of girls as a criteria that they should similarly succeed in IITs. As mentioned in a later section of the blog, this correlation between CBSE and AIR does not hold, even for guys.

    2. JEE might be too tough than it should be, but surely that doesn't make it "tougher" only for one gender - and easy for other - especially if one does not doubt the intellectual abilities to be equal among genders. In case the difficulties for girls are due to social pressure etc, that again does not support the assertion of the JEE exam itself is biased.

    3. If one says that girls would get in if JEE only if were made easier, it puts an unwarranted question mark over their intelligence, one that is unwarranted and incorrect.

  5. Anonymous said...

    ‘‘There is little correlation between AIR and CGPA.’’ This means, toppers in the JEE are not at the top during their B Tech programme.

    This is not very surprising. Most of the top JEE ranks opt for Computer Science. But it is highly likely that the topper of some "low" branch has a better GPA than many of the students in CSE.

    JEE rank is just an easy way to decide who gets which branch. Wish the administration would be willing to take other factors into consideration while admitting students. (But I stray from the topic.)

  6. barbarindian said...

    So, would you support BJP's proposal of 20% reservation for women?

  7. Gaurav said...

    A slightly different point. Schools in the US use CGPA as well as standardized tests, among several other criterion. A research design expert at my univertsity told me this is because standardized tests measure mental ability, but not long term motivation. School grades also measure long term motivation.

    Maybe instead of just setting school grades as a lower cut-off, IITs and even IIMs could factor them in a composite measure.

  8. Gaurav said...

    A few years back the CAT was readministered due to leaks. A few years prior to that, the same thing happened for JEE. Do you know if any reliability analysis was performed on that data? If it wasn't, then it should be.

    GRE and GMAT have a Cronbach's Alpha that is pretty high. At least for those two years, we will see reliability measure of the ranking, if not selection itself.

  9. Pratik Ray said...

    regarding correlation between JEE and BTech GPA:

    Consider:

    1. The sort of preparation required for board exams and entrance exams differ drastically.

    2. The ranks are rather "clustered" within a given discipline. The difference between the student with the best JEE rank and the worst JEE rank within a branch, I guess would be at most 5 marks or so out of 300. Clearly, with such a slim margin, on another day the ranks could so easily have been reversed.

    4. That some of the students "sacrifice" their board exams for JEE (while obviously quite a few dont)

    Now, it is fairly easy to see that one cant hope for a good correlation between JEE and BTech GPA. When a student preparing well for both the exams, and one preparing for just the JEE get the almost the same rank (or even if the fellow preparing for both gets a slightly lower rank) they cant really be considered at par.

    ---------------------------

    By the way, it isnt such that IIT JEE is the only such entrance exam where this lack of correlation with JEE rank exists. Such a lack of correlation exists in state entrance exams too (its just that most people dont bother to look at these less glamorous exams).

    Just an example, once again from my BTech batch in NIT-R. The topper in BTech had the second last rank in the state entrance exam amongst all the gen category students in the dept. The 2nd topper had the 3rd last rank. The total class strength was 42. Both of them, by the way, scored around 75-77% in the board exams and the topper also happened to be a Math Olympian.

    I admit a couple of cases dont prove anything, but in my observation, such lack of correlations between state entrance exam ranks and BTech ranks are the norm as well. Therefore, its not the IIT JEE which is just at fault. Even the "easier" exams show the same trend. So an easier IIT JEE wont necessarily lead to better correlations as long as the format isnt the same as board exams. As an aside, do note, that most of the BTech semester exams are similar to board exams in nature,i.e. the emphasis is on having a decent idea of the subject rather than pose challenging problems.

    This lack of correlation will keep springing up as long as you have one set of students preparing only for an entrance exam, and another set preparing for both entrance and board exams. The way out would be if you could have just a single set of exams. However, one just needs to look at the current state of board exams to understand that they dont really test the fundamentals. They are too many "State the Newton's laws" type of questions instead of application oriented questions.

    Let me ask another question. What really constitutes a good/deserving student? Is he the one who can solve tough problems using known and standard (thanks to coaching centers) techniques (IITians!) or is he the one who can "describe" and "memorize" stuff the best (board exam toppers) or the one who has a bit of knack for both the above mentioned qualities (though not necessarily in equal measure as IITians or board exam toppers), but has the ability to think out of the box? Note: I dont deny that many of the IITians (though not all) and many of the board exam toppers (though not all) would fall in this third category.

    If the purpose of entrance exam / normalization is to find out the really "good/deserving" students, this is a question we need to answer unequivocally at first.

  10. chitta said...

    Abhi:

    In your point 3 you said

    "Thus, I believe JEE should be replaced with an exam that has as its primary purpose the simple job of normalizing the scores across exam boards. Yes, this does imply that the questions must be standardized (pre-tested on a sample of students, with more difficult questions carrying a greater weight)."

    ===

    I wonder what you mean by "more difficult questions". Could you please elaborate on that how those "more difficult questions" would be different from questions in current JEE exams and yet would be able to select 5K students out of several hundreds of thousands that sit for JEE.


    Thanks
    Chitta

  11. Tabula Rasa said...

    Consider this: schools themselves can train students for tackling both board exam and the new JEE; teachers can offer extra classes for motivated students.

    Come on Abi, how many people sit for the board exams on the strength of classes taken in school alone? Everyone who can afford it gets tuitions -- some people have tuitions right from junior school onwards. Even at the "excellent" Delhi high school I attended, I'd probably have trusted only one and *maybe* two teachers to coach me outside class.

    Let's reform the school system first.

  12. Unknown said...

    This is becoming a forest of words that we are getting lost in. Guys, the problem is not JEE - its a wonderful exam to identify sharp guys. The problem is not coaching - some kind of training is always there for every performance linked event in life. The problem is the huge demand versus abysmally small supply. We need to increase the number of IITs. Once we that, you will see most of the problems will get solved automatically!

  13. Arun said...

    Abi,
    I hope you've considered the fact that many states have a women's quota for Engg. And that contributes a lot to the 20-25%.

  14. Anonymous said...

    >> Aside: Some -- notably, RC -- have argued that since JEE is a 'selection machine' that has no prior conception of differences among candidates, it's not biased. This is a pretty lax yardstick with which to judge the fairness of a selection machine; by this yardstick, even a lottery is a fair machine! >>

    Abi,

    Ironically, a lottery is considered the world over as the ultimate benchmark for fairness.

    In my view, there are only two completely fair systems in the world:

    - a lottery (for selecting multiple winners)

    - and a toss (for selecting a single winner)

    Any contest (such as exams) are the worst examples of fairness, because they are designed to be in favour of those who prepare for it.


    Can we use a lottery then for as a "selection machine" for IITs ? I see nothing wrong with that proposal, provided the intention is to select lucky students from a given population.

    While the essence of your post, that women are not in IITs makes for good debate. Your attempts to investigate bias in the JEE selection machine are not along the correct lines. The JEE selection machine cannot distinguish between girls and boys, so the machine itself is clean and squeaky. Maybe what you want is to replace this clean and squeaky machine with another clean and squeaky machine that will make it more favorable to women. Fine, that is a reasonable argument.

    This may seem to be trivial semantics, but it is important to be clear that you are suggesting an alternative selection machine. That is quite different from proving bias in the current machine.

    To illustrate the selection machine point further, if you had accused the IIM selection machine as biased in some way against women or rural students - that needs further investigation. This is because of the fact that the GD and Interview rounds can quite clearly distinguish.



    Your point about the JEE acting as a normalizer is well taken. Before we get there, a lot needs to be done to align various state board syllabi.

    The biggest problem I see with the coaching centres is the cost and accessibility. I see them as private gates controlling access to a public resource.

    This is why I like ideas that promote a cat-and-mouse game between the JEE and the coaching centres. Today, the two are lined up evenly as if in secret collaboration. That is dangerous.

  15. Anonymous said...

    I can't figure out how to trackback, but here's my take:

    http://thetomatoes.wordpress.com/2007/04/30/coming-late-to-controversy/

  16. Anonymous said...

    Some questions regarding your analysis :

    1) When you mention successful candidates, do you mean candidates who managed to get into IIT only or those candidates who managed to get through JEE and ended up in IT BHU or ISM Dhanabad ?

    2) Again, wrt women candidates did you mean the percentage of women getting into IITs or those clearing JEE ?

    3) An article in wikipedia mentions the total intake through JEE in 2005 and 2006 as 4935 and 5444 respectively. These numbers are inconsistent with your figures. It is, of course, possible that the figures in wikipedia are erronous.
    In that case where can one find the exact figures ??
    The article is here:
    http://en.wikipedia.org/wiki/IIT-JEE#Seats

    4) ISM Dhanabad has no intake of women as a policy. As mentioned in the above article, intake was 402 and 658 students in 2005 and 2006 respectively, roughly about 10% of the seats. That leaves women eligible for only 4533 and 4786 respectively, at least according to the numbers in the wikipedia article.

    Did you take this factor into account while calculating the percentage of successful women candidates ?

    -Abhijit

  17. Anonymous said...

    Interesting thoughts Abi. I need to go through the other posts on this topic.

    Personally, I don't believe that the ability to crack near-unsolvable problems in limited time is an appropriate measure of intelligence. But the professors might know better than me.

    The mushrooming of coaching classes is a recent phenomenon. The percentage of women in IIT was not great even in 70s or 80s. I don't think the proportion went down in last 15 years or so. At worst, there has been no improvement. It is however important to disincentivise the role of coaching classes, because of their unfair distribution and monetary gateway.

    By the way on the topic of tough exams, the chinese National college entrance exam is one of the toughest at its level. And there too they make similar criticisms. Although, Iam not aware of proportional number of women engineers in China.

  18. chitta said...

    An interesting article somewhat related to the topic under discussion:

    http://timesofindia.indiatimes.com/CIC_asks_IIT_to_disclose_cut-off/articleshow/1981857.cms

  19. Abi said...

    Rahul: Yes, I too would be quite happy if a bright student can get through without any coaching. I'm also with you on the need for JEE to be problem-oriented.

    Vivek: Whatever I have said is based on info available until last year. And the changes in JEE that you noted were implemented last year itself! But you are right: may be they are moving in the right direction, so I'm willing to suspend my judgement until this year's results are out. Let's hope for some good news.

    The BITS-Pilani thing is something that has been bothering me. I will tell you about something else: in my 'bleg' post, take a look at the data from the NITs. Women seem badly outnumbered in these places, even in computer engineering. Thus, there is something strange about entrance exams per se, that put women at a disadvantage. This is worth exploring.

    Wanderer: The problem is not that JEE is tough. The problem is that it has questions at such high levels that they can be answered only with the help of coaching. A student's school education alone will not do.

    Vishnu: I don't know the details of the IIT-study, but I presume it includes the first year, when everyone goes through the same curriculum.

    Gaurav: Your suggestion that the JEE score can (somehow) incorporate the students' performance in high school exams is a very good one. The rationale is excellent.

    Sorry, I have linked to the only study that I know of. I agree that there should be lots of serious studies done on entrance exams's efficacy in identifying people who will do well.

    Pratik: Your point about the generality of the lack of correlation between the rank and CGPA is interesting. Your other point about the 'noise' in the JEE results is very valid.

    Here's my (personal) answer to youre question: Solving tough problems is not necessary, but ability to apply one's knowledge to new but simple situations is essential. This is the kind of skill that a bright student should be able to pick up quite easily. Coaching may only help this student with some minor additional skills (exam taking strategy, time management, etc), but this advantage cannot be huge, if the problems that are in the entrance exams are relatively simple.

    Chitta: I just mean what SAT or GRE does. It 'knows', beforehand, the difficulty level of every question it poses to a student. I don't know the policy of SAT or GRE, but I wouldn't include any question that cannot be answered by at least 25 percent of target students. Currently, JEE seems to thrive on questions that can be answered by less than 10 percent.

    TR: Let me confess: it's extremely difficult to convince you! I throw up. You win!

    Phani: Do you really mean that there is nothing wrong with JEE? Gosh, I have wasted the whole of last week to come to this?

    RC: Thanks for the link to your insightful post. I agree with you about coaching centres being private gates to public resources. Though I have been talking about the bias against women, I have alluded to (in several places) my concern that JEE -- through the coaching centre mechanism -- discriminates against poor students and those from rural areas.

    Tipsy Toes: Thanks for that link. I will get back to you in a while.

    Anon (Abhijit): All the data about successful candidates in this post is about JEE rankholders. It's not about those who actually joined IITs and other institutions.

    Anoop: Thanks for that info about the Chinese version of JEE! Several Asian countries seem to believe in brutalizing their young: Japan, Taiwan, etc. In the West, I know of France which has a pretty tough exam to select students for their Grandes Ecoles.

    Chitta (again!): Wow! That is a pretty explosive development. I really liked the name of the person who filed the complaint: Eklavya!

  20. Anonymous said...

    Abhijit said:
    4) ISM Dhanabad has no intake of women as a policy.

    As a graduate from the Indian School of Mines, I can vouch for the fact that ISM has had women undergraduate students since 1985. However, the number has traditionally been small.

  21. Anonymous said...

    You keep ranting about the average performance of girls in X / XII and compare it with their relatively less successful numbers in JEE, when what should be compared it their performance in NTSC exams, that have more of a problem solving nature than the different board examinations, that emphasize rote learning and regurgitation.

    How many girls prepare for the NTSC exams ?

    How many girls clear these exams ?

    Some here claim that JEE only tests how motivated you are to clear a test, not sustained motivation. Yet, if one needs 3 years of hard work to get through JEE, what more evidence of sustained motivation does one need ?

    Others claim that a standardized test will solve all problems, but what happens if 20,000 people have scored 100% in the standardized test and you have only 6000 seats to offer ?

    Others claim that AIR bears no correlation to GPA, which is simply not TRUE !! In the first year of studies, when all subjects are common to all students, the GPA of students correlates extremely well to their AIRs ! While people in CS and elec (relatively higher AIRs) breeze through, people in say, MSc chem find the first sem tough. so much so, that at the end of the first year, MSc walas breathe a sigh of relief, and CS wallas breathe a sigh of easy pickings gone. Once the students are segregated into individual branches though, performance does not correlate to AIR so well. But thats understandable, because most students are clustered very close together in their performance to have gotten into the same department.

    >> schools themselves can train students for tackling both board exam and the new JEE; teachers can offer extra classes for motivated students.

    Oh abi, I can almost see the ivory tower you are perched in. :-) Not all schools have a teaching staff that can cope for a differentiated level of teaching, i.e. teach students of different capabilities the same subject at a level thats equal to their capabilities. Thus, you have removed the alleged bias provided by coaching centers to affluent schools that can afford extra teaching staff and offer extra hours to their students.

    But hang on, isnt this what coaching centers do in the first place ? They test students to pick those with better scholastic abilities, and then administer extra/individualized classes to students at different motivation/scholastic levels, although they do charge fees. Is that what your objection is ? The Capitalistic exchange of money for services rendered ? This kind of explains a lot.. :-)

    Sudeep

  22. chitta said...

    Abi:

    You said: I just mean what SAT or GRE does. It 'knows', beforehand, the difficulty level of every question it poses to a student. I don't know the policy of SAT or GRE, but I wouldn't include any question that cannot be answered by at least 25 percent of target students. Currently, JEE seems to thrive on questions that can be answered by less than 10 percent.

    ==

    I have similar worries as the previous poster: What if after the first 50 or so ranks there are 10,000 people in the 51st rank
    and there are 5000 seats.

    Chitta

  23. Anonymous said...

    Dr. Baral,

    I am not sure that JEE still asks questions that can only be answered by < 10% of target students, that used to be the case in my times more than 10 years ago. AFAIK, The pattern of JEE has been changed to an all objective question paper.

    Sudeep

  24. chitta said...

    Sudeep:

    I was quoting Abi and was in a sense questioning his proposed solutions.

    Chitta

  25. Tabula Rasa said...

    abi:
    no no, please don't throw up! :-D
    someone said a couple of days ago that you were diluting your case by introducing arguments that were at times weak and at other times irrelevant. as i've said before, i think we have a basic disagreement of *opinion* as to whether coaching classes are a necessary "evil" or not, and whether an entrance exam is to be held as culprit for some of society's most deep-rooted biases. ultimately, unless there are data, aka *facts* available, people will simply keep talking at each other. imagine what this discussion would have been like without vivek's contribution.

  26. Abi said...

    Chitta: You seem obsessed with ranking, which need not be the main concern. When I say a standardizing exam, I don't quite mean something at the level of SAT where a LARGE number of students get a perfect score. One can think of something that's intermediate between something at the level of SAT and at the level of JEE, all the while ensuring that the new JEE strictly sticks to the syllabus of CBSE (or equivalent).

    TR: Oops! I certainly intend to throw up. Your comments are quite delicious, actually.That mangled phrase was either "throw my hands up" or "give up". I forget now ...

    You are certainly right about the difference in 'opinion'. I know you don't want to give even an inch here, but tell me, TR, which part of the 'data' that I presented are you unhappy with? Sure, a more detailed set of data (Class XII composition, the aspirations of students at this level, etc.) will allow us to refine our arguments and conclusions, but given what I have presented here, wouldn't you at least 'suspect' that there is a plausible 'gender bias' in the IITs' exam? Just curious, because you seem willing to give these institutions a clean chit.

    I also agree with you on Vivek's intervention. Ever since he started this lively discussion, I have been learning a lot of new stuff.

  27. chitta said...

    Abi:

    The ranking part was only for illustration that it is not an easy job to select 5000 out of hundreds of thousands.

    As the TOI article illustrates even the real tough IIT questions don't allow that and the IIT examiners seem to be using arbitrarily decided weights to decide on the selected 5000 and rank them. (Thats what they do.)


    Abi: I agree with your goals.
    I just wonder if what you want is doable with only changing the JEE exam. I am sure you know there have been debates inside IITs and they have been trying to tinker with the exams. I came across some old debates at http://www.iitk.ac.in/infocell/Archive/dirjuly3/directions.html
    http://www.cse.iitk.ac.in/users/dheeraj/ideas/newjee.html

    My gut feeling is similar to Vivek's; figure out how to (not in a year or 2) increase the number of seats from 5,000 to 50,000 or even more. That will partly solve a lot of the issues that you point to. But as others have said its easier said than done and will take years.

    So lets have that as a 5-10 yr goal and in between continue with the tinkering.

    cheers
    Chitta

  28. Sriram said...

    The discussion seems to ignore well documented differences between men and women on math/spatial reasoning, the kind of skill that has some bearing on dealing with exams like JEE or math olympiad. Its hardly limited to India. It is a world wide phenomenon. Such a disparity is noticeably absent in medicine and law, both equally or even more prestigious professions.

    Even small differences in means will translate into large discrepancies at the tails of two overlapping gaussians.

    This does not rule out the operation of social factors but even when all social aspects are controlled for (or equalized), I dont believe we can expect more than 15% presence of women in JEE ranks. A difference does not equal discrimination.

    The same sort of reasoning probably applies to groups that are over-represented in IITs (Iyers, Agarwals etc) compared to their numbers in the population. Relative ratios at the tails of distribution can be very different even with (relatively) modest mean differences (say half a standard deviation)

  29. Sriram said...

    author writes
    --
    Thus, in a random selection of students in 2005 or 2006, women are three to four times less likely to be found in IITs than in other engineering colleges. This, by itself, does not constitute an evidence for (a plausible) bias in JEE against women. For that, we have to show that the fraction of women in the eligible pool of candidates is far higher than that in the IITs.
    ---

    Your line of thinking is flawed/vague. So long as JEE is based on advanced reasoning about math/physics, it WILL be biased against women (on average). If you want to remove the bias, convert it to an exam pertaining to verbal ability, knowledge of history or biology. Every exam has a BIAS and it needs to have a BIAS to do the job (until such time the domain itself changes).

    Less selective exams like AIEE and CBSE will have greater proportions of women (even when subject matter is controlled for) simply because the comparison is between subsamples that are not on the extreme ends of the gaussian (or non-gaussian) distributions (as in the case of JEE, see my previous post). The more extreme the selection criteria, the greater the discrepancy between groups (based on gender, caste, region, what have you).

    http://www.lagriffedulion.f2s.com/ has excellent examples of valid statistical reasoning that you didnt seem to have picked up in your education as a metallurgist.

  30. Anonymous said...

    ok, i am a woman and did my BTech from an IIT.

    some of my observations based on women in my batch and seniors/juniors:

    most of us came from families with no sons. in other words most of us had sisters only.
    wonder if there is a correlation..