Search This Blog

Tuesday, March 28, 2017

GATE Results: State of engineering education

GATE 2017 results are out, and once again, there is disappointment all over. This news article is stating that only 16% of those who gave the GATE exam could qualify. "Qualifying" means that these students have the capability to seek admission to an MTech program in the country. I cannot say about all engineering disciplines, and my comments are based on my familiarity with Computer Science/Information Technology education in the country.

We don't yet know the detailed statistics, but we do know that in the last couple of years, percentage of CS qualified candidates was lower than the combined percentage of qualified candidates in all disciplines. Let us assume the percentage of qualified candidates in CS in GATE 2017 to be about 14%.

Note that this 14% is after declaring candidates from reserved categories as qualified at much lower cutoffs. If we were to use the same cutoff for everyone (since that really reflects the preparation levels in Computer Science for all graduates), then this number could be as low as 11-12%. The reserved category students are the majority of the students giving GATE exam.

Are all these students really capable of doing an MTech (which, by the way, typically requires that you know only a few courses in CS). Anyone who has interviewed for MTech admissions, even in tier 2 institutions, would tell you that a lot of them are unable to answer many questions, and their marks in GATE are really a result of sustained coaching effort and not really understanding the subjects. Give them the same question that came in GATE paper in a different format, and they are unable to answer. Give them the same question 4 months after the GATE exam, and they claim that they have forgotten. So the percentage of students who can go through a poor quality MTech program in a Tier 2 institute is perhaps half of these qualified candidates. So may be around 6%. (A lot of hand waiving here. No scientific study. Just my impressions.)

Of course, not everyone gives GATE. How many of the "good" graduates not give GATE versus how many of the "poor" quality graduates not give GATE. I suspect that more than 1 lakh is fairly representative sample, but we may add a couple of percentage to our numbers, just to take care of this. So may be about 8% graduates are capable to doing at least a poor quality MTech program at a Tier 2 institution.

Note that this is roughly the capability required to do routine software related jobs in service industry, and NASSCOM, the industry body which represents the IT service industry, has been saying now for at least a couple of decades that less than 10% of our graduates are employment ready. (There is another class called "employable" which are those who can be readied with substantial training. With that the numbers go up.)

What should be the goal of a 4-year BTech (CS) program? Should the goal be to learn just 4-5 courses in a half-baked fashion which will enable you to do a low-quality job in IT service industry, or get admission in a poor-quality MTech program in a Tier 2 institute. If that was the goal, then about 10% of the graduates are meeting that goal, and 90% are not.

However, as media reported recently, HCL, the 4th largest IT service company will start hiring smart 12th class students and train them on those few things that they really need. The fact of the matter is that what can make you qualify in GATE (CS) is something that you can learn in a few months, you don't need a 4-year program, and that is exactly what the coaching industry is caching on.

In an earlier blog long time ago, I had pointed out that 50% of the graduates actually got a GATE score of 0 or negative in that year. It means that a majority of our CS/IT graduates are no better than a class 1 student (note I am comparing with someone who has just entered the school, not someone who is about to leave the school).

If we consider the goal of 4-year BTech (CS) program to not just learn a bit of programming, data structures and algorithms, but also understand how operating system works, how networks operate, how machine learning can be used to solve problems, and actually solve some of those problems, then the percentage of graduates who satisfy these program goals are much smaller than 10%. We don't have such numbers, but my gut feeling is that the numbers could not be more than 3-4%. Which means that out of about 10 lakh graduates in CS/IT domain, may be 30-40 thousand graduates meet what should be the requirements of a 4-year BTech (CS) program.

(I wonder if GATE would be willing to share the detailed data on each candidate's response to each question, and which colleges these candidates come from, etc., so that we can do a more detailed analysis, instead of making guesses.)

Thinking about who these 30-40 thousand graduates are, I would guess that these are primarily from IITs, NITs, IIITs, some of the state universities, deemed universities, etc. If we consider the entire eco-system of affiliated colleges in the country, there is perhaps less than 1% graduates who meet the requirements of a BTech program. And these 1%, I am sure, are largely because the students were inherently good, and the colleges could not spoil them despite best efforts. (Of course, there are always exceptions to the rule, for example, NSIT in Delhi.)

Now, think about it. AICTE is supposed to regulate all affiliated colleges. And the result is that may be around 1% of the graduates of those colleges meet the goals of a 4-year BTech (CS) program. 99% of those who are getting degrees are not good enough for those degrees. What if we were to disband AICTE. What if the government were to suddenly decide that all these affiliated colleges have complete autonomy starting now. They can decide whom to admit, whom to recruit as faculty, what to teach, whom to give degrees, and so on, without any interface with any government department or agency. They can choose to have accreditation, if they so desire. They can decide their fees, and so on. So basically leave everything to the market forces. I am sure a very horrifying thought experiment for most people. But would market fail us even more. Would market forces result in 100% of the degrees being worthless, or is there a chance that may be instead of 99% of the degrees being worthless, the number may reduce to 98% or even lower.

I think the time has come to ask AICTE and MHRD some tough questions. Why regulations?


Ranjith said...

If we leave everything to the market forces. I'm wondering, what's the difference between education institutions/ universities and job training centers.
If market forces are the only determinant criteria, what's the need for setting up universities with vision and mission.
Would like to know your views on this.

Milind Sohoni said...

dear prof. sanghi

while much needs to change at mhrd and aicte, much more needs to change at the elite IITs as well.

see which appeared in Current Science.

"However, the conduct and reporting of this examination is decided by the older IITs and Indian Institute of Science (IISc), Bengaluru. This curriculum is largely
based on the faculty expertise of the organizing institutions,existing undergraduate (UG) curricula, the type of graduate students that they require for their research purposes, and the amount of effort that they wish to put into it. For example, consider the topic of ‘concurrency’ in Computer Science and Engineering (CSE), which appears in two areas, viz. databases, and operating systems, and was a traditional research area of faculty members in these institutions. It is unclear if concurrency is industrially important, can be taught by a wide body of faculty members across our engineering colleges, and that its ramifications in both databases and in operating systems are important attributes of a typical CSE graduate of the country.

Moreover, as Tables 1 and 2 suggest, the IITs and IISc have been sending a minority of their graduates into engineering jobs and fewer still into engineering for domestic needs. The typical syllabus in GATE is what is taught at the IITs/IISc and which is largely theoretical, and needs about 12–14 core courses. For example, most graduates from IITB would not have visited a factory or a water supply system, let alone a firm in the small and medium sector, the mainstay of domestic engineering.
However, most colleges do not have the requisite faculty strength and expertise to teach the above 12–14 theoretical courses (though they may be placing more students
into domestic engineering), and students must take recourse to coaching classes ..."

also see my proposal to aicte/mhrd on regional engineering.

regards, milind.

Dheeraj Sanghi said...

@Ranjith, First of all, I am not saying that free market is the solution for all ills. I am only saying that free market is better than AICTE. That free market can not result in 99% failure. Second, free market can result in differentiated products. In fact, I see that all around me in every category of goods and services. US allows for-profit education, and yet, most of the top universities in private sector have chosen to remain not-for-profit. That too is free market.

Dheeraj Sanghi said...

Dear Prof. Sohoni, I have gone through both the publications that you mention in your comment. Thanks for sharing. I fully agree with you on all the points about GATE, including the syllabus representing what is taught at IITs, etc. However, the way I look at GATE is this. That students can hardly score 75% and more is a problem of GATE (all the problems you pointed out), but that 50% of the students score a 0 or less is a problem of engineering education. Irrespective of what is your ideal curriculum (given focus on regional engineering, hands on, supporting local industry, solving local problems - or whatever else you want to achieve through that ideal curriculum), at least some of those questions (indeed, a lot of those questions) should still be answerable. And hence I am concluding that the issue is that of quality of education and not so much as quality of exam.

iitmsriram said...

Duh, I thought the qualifying fraction will ALWAYS be around 15% since the current scoring scheme was introduced some years ago and this is taken as the target point. I don't think we need to worry terribly about the percentage of people qualifying (the past percentages are available at GATE website We should be looking at the raw scores (marks out of 100) being obtained by the top candidates, the average qualifying candidate and the last qualifying candidate, all also available at above link.

Dheeraj, I take it you have not set GATE question paper. There are extensive past statistics of which questions most people don't get, which ones most people get etc etc. and appropriate suggestions are made to the question paper setters. On account of this, the top scores have been increasing, but the bottom scores are still abysmal. Given all this, I don't know what we can conclude about anything based on GATE scores!

Dheeraj Sanghi said...

@Sriram, I have indeed set GATE question paper once many years ago, and that is how, I know that 50% of the students got 0 or less marks at least in one year. So the raw scores are absolutely terrible. This time, candidates with 25 marks have been declared qualified in general category (and general category students are in a minority). So, my guess is that only 10-11% of the students have got 25% marks or higher in the CSE paper. This is terrible. I would love to do more analysis if I get access to detailed data. But my gut feeling is that if we were to divide GATE questions into a few important courses - programming, Data Structures, Algorithms, Organization, OS, Networks, Databases, etc., and then see how many people have been able to answer correctly, at least 33% questions in each sub-discipline (without negative marks), that number would not be more than 5% of GATE candidates.

Ayan Banerjee said...

I think all the colleges will reach much higher education quality if they are regulated by "coursera" or "edx" as their examinations and course offerings which will be supervised by local professors.