World Reputation Rankings 2013 methodology

Poll position: experts think these are the best in the business

The excellent response to the third round of the annual Academic Reputation Survey gives an even more accurate picture of scholarly opinion

The Times Higher Education World Reputation Rankings are created using the world's largest invitation-only survey of academic opinion - a truly unique piece of research.

The Academic Reputation Survey, available in 10 languages, uses United Nations data to ensure that it is-properly distributed to reflect the demographics of world scholarship. It is also evenly spread across academic disciplines.

Those invited to take part are statistically representative of both their country and their discipline.

The questionnaire, administered by polling company Ipsos MediaCT for THE's rankings data supplier Thomson Reuters, targets only experienced, published scholars, who offer their views on excellence in research and teaching within their disciplines and at institutions with which they are familiar.

The 2013 rankings are based on a survey carried out in March and April 2012, which received 16,639 responses from 144 countries. When polled, the respondents on average had been working in the academy for 17 years.

With 13,388 answers to the first Academic Reputation Survey in 2010 and a further 17,554 in 2011, just under 48,000 detailed expert responses from more than 150 countries have now been collected in just three annual rounds.

There is a balanced spread of responses across disciplines. In 2013, the most (22.1 per cent) have come from the social sciences, followed by engineering and technology (21.3 per cent), physical sciences (18.0 per cent), clinical subjects (15.4 per cent) and the life sciences (12.7 per cent), with the arts and humanities polling the lowest (10.5 per cent).

The spread across the regions is also well balanced: 33 per cent of responses hail from North America, 17 per cent from Western Europe, 12 per cent from East Asia, 10 per cent from Oceania, 6 per cent from Eastern Europe, 5 per cent from South America and 5 per cent from the Middle East.

In the survey, scholars are questioned at the level of their specific subject discipline. They are not asked to create a ranking or to list a large range of institutions, but to name no more than 15 of those they believe to be the best, based on their own experience.

The survey data were used alongside 11 objective indicators to help create the World University Rankings 2012-2013, which were unveiled in October last year. The reputation data are published alone each year to create the World Reputation Rankings.

Calculating the scores

The reputation table ranks institutions according to an overall measure of their esteem that combines data on their reputation for research and for teaching.

The two scores are combined at a ratio of 2:1, giving more weight to research because feedback from our expert advisers suggests that there is greater confidence in respondents' ability to make accurate judgements about research quality.

The scores are based on the number of times an institution is cited by respondents as being the best in their field. The number one institution, Harvard University, was selected most often. The scores for all other institutions in the table are expressed as a percentage of Harvard's, set at 100. For example, the University of Oxford received 73 per cent of the number of nominations that Harvard received, giving it a score of 73 against Harvard's 100. This scoring system, which is different from the one used in the World University Rankings, is intended to provide a clearer and more meaningful perspective on the reputation data in isolation.

The top 100 universities by reputation are listed, but Times Higher Education has agreed with data supplier Thomson Reuters to rank only the top 50 because the differentials between institutions after that point become very narrow. The institutions that make up the second half of the table are listed in groups of 10, in alphabetical order. Scores are given to one decimal place, but were calculated to greater precision.