Sunday, January 01, 2012

PISA 2009 update -- results from India, UAE, and a few other countries

The OECD’s Program for International Student Assessment (PISA) tested an additional 10 countries/regions on the PISA 2009 test. Known as the “PISA 2009+” test countries, they include:

• Costa Rica
• Georgia
• 2 States in India (Himachal Pradesh and Tamil Nadu)
• Malaysia
• Malta
• Mauritius
• 1 State in Venezuela (Miranda)
• Moldova
• The United Arab Emirates

Strangely enough you can’t find the report on the OCED website, you need to go to an Australian education website for it.

I was excited to see the update because in previous blog posts that discussed the PISA test I was always curious to see how India would do given the emphasis Indians place on education. Would India score high marks despite its relatively low GDP per capita?

It turns out its marks were -- not so good. Here is a table with some select countries for comparison. The first number is the reading score, followed by mathematics and science:

556 600 575 Shanghai (highest results of all participants)
536 541 554 Finland
520 529 539 Japan
524 527 529 Canada
500 487 502 United States
493 496 501 OCED average
494 492 514 UK
464 445 454 Turkey
431 421 438 UAE
425 419 416 Mexico (lowest scoring OECD country)
414 404 422 Malaysia
413 381 402 Columbia
402 371 383 Indonesia
372 368 379 Qatar
370 365 369 Peru (second lowest of all the countries tested in the original PISA 2009)
337 351 348 India (Tamil Nadu)
317 338 325 India (Himachal Pradesh)

314 331 330 Kyrgyzstan (lowest scoring of all the countries tested in the original PISA 2009)

Ummmm, wow. India scored way lower than I expected. Nearly bottom of the table, except for Kyrgyzstan, and Himachal Pradesh scored lower in science than them.

So what happened? I'm not sure yet. I thought at first maybe it would be the social economic correlation (wealthy Indians would test way higher) but the PISA report notes:

In Himachal Pradesh-India and Tamil Nadu-India the relationship between socioeconomic status and reading performance, as measured by PISA, was very weak. This may be because students in these populations perform very poorly in reading and have low socioeconomic status, as measured by PISA, and therefore the relationship between socioeconomic status and reading cannot be adequately detected using the PISA scales.

Translation: “everyone taking the test in India was poor by our standards so we couldn't measure the difference between rich and poor”. Thanks guys.

The India test results were very different than the results Indian schools in Qatar achieved so for now I'm going on the assumption that overall education in India may still be very poor, only those who can afford to send their children to decent private schools are getting a good education. Any Indian readers please feel free to leave a comment to clarify what happened here.


Anonymous said...

The Indian results are unsurprising because it corroborates earlier studies all coming to a broadly similar conclusion. That Indian education is mediocre and overall level of human capital in India is very low. Please refer to the TIMMS test administered by Jishnu Das and Tristan Zajonc or the Wipro educational initiative results.

Your skepticism is unfortunately untempered by rationality. To expect high Indian results based from the selection of the highest scoring school in Qatar is the result of several cognitive biases. Chiefly there is the issue of selection bias, by comparing the most unrepresentative school in the nation with OECD averages. To apply your logic, I can pick the highest scoring school in the US or China in the PISA tests and it will make even the highest scoring Indian (Qatar) students, look like drooling idiots, relatively speaking.

For example Al Khur Interntational school only admits the children of the employees the state LNG monopoly. These aren't just the children of affluent Indians, these are the children of Indians that are, probably, greater than two standard deviations above the socio-economic mean of India. To be that selective in the US, you would have to administer such a test to only the children of Harvard, Yale, Princeton, MIT, etc alumni.

In fact, the Qatar results for Indian students, rather than showing the possibility of high Indian achievement actually comes to the opposite conclusion. That even a highly selected group of Indian students are just above the OECD mean (itself dragged down by relatively poor performing OECD nations). What does that say about the rest of India's human capital +/- one standard deviation from their mean? Nothing positive.

Glen McKay said...

please refer to my post of Jan 3, 2012.

rec1man said...

Calculating Average Indian IQ from PISA

TN raw math PISA score = 351
TN implied IQ = 100 - 1.5 x 15 = 78

HP raw math score = 338
HP implied IQ = 100 - 1.62 x 15 = 76

Indian Avg IQ based on raw IQ = 77


Next step is to remove the bias caused by the PISA sample having
75% bilingual kids

TN mono-lingual = 378
Implied IQ = 500 - 1.22 x 15 = 82

HP mono-lingual = 401
Implied IQ = 500 - 15 = 85


Next there is a 40 point difference between scores for 'Village' and scores for 'Large city'
In HP and TN, The village category is over-represented by a factor of 4
Even worse, in HP, City and Large City are entirely removed from the survey sample

So adding an urban correction of 20 ( half the village-large city difference )

TN semi-urbanised mono-lingual = 378 + 20 = 398
Implied IQ = 85

HP semi-urbanised mono-lingual = 401 + 20 = 421
Implied IQ = 100 - 0.79 x 15 = 88

Current Indian IQ = 86


Next we look to the future as malnutrition is removed

The only Indian kids who go to govt school is for the mid-day meal,

If they are not starving they go to private school

Private schools score 45 more than govt schools and thats the future as poverty reduces


HP - future - semi-urbanised- mono-lingual = 401 + 20 + 45 = 466

Implied IQ = 95

TN - future - semi-urbanised - mono-lingual = 378 + 20 + 45 = 443

Implied IQ = 91

Future Indian IQ = 93

Given the huge bias in sampling towards over-representing the lower end IQ,
by the poverty pimp NGOs, I am certain that none of the CBSE or Cambridge schools
that serve the top 15% are included in the survey

And they have an entirely different IQ profile and cause an IQ bulge at the top end