Women are much better forecasters than men, and a weighted average of forecasts in Fairfax Media’s BusinessDay economic survey easily outperforms the Treasury, an analysis of 20 years of Australia’s longest-running economic survey has found.
“Answers to the BusinessDay survey, when weighted by past performance, provide the most accurate macroeconomic forecast in the country,” said researcher Warwick Smith. “Treasury would do well to set up a system like it.”
The BusinessDay survey has been running for almost 40 years. The first edition, published in The Age in January 1979, predicted economic growth of 3.4 per cent, inflation of 6.25 per cent, and an interest rate on government bonds of 8.25 per cent.
A year later BusinessDay reported that those first forecasts had been “way off beam”. Early in 1979 the Shah of Iran was toppled by Ayatollah Khomeini. By the end of that year 52 Americans had been taken hostage in Iran and the price of oil had doubled, pushing Australia’s inflation rate to 10 per cent, the government bond rate to more than 10 per cent, and economic growth to 4 per cent. Australia was an exporter of oil.
As has been the case in every BusinessDay survey since, the chief economists of each of the big four banks’ took part, as did 25 other economists from industry and academia including Professor Neville Norman of Melbourne University who took part in the first at the age of 32, who continues to take part today, aged 72.
Professor Norman attended the working lunch convened by former Age business editor Terry McCrann in 1978 to toss around ideas for the survey. He said in those days the Treasury published few forecasts and private consultants charged their clients and kept their forecasts secret.
“I wanted people to know what was being forecast and why, and I wanted some of the forecasts to span more than a year ahead,” he said. “Terry doubted the readership would take forecasts more than a year ahead. I suggested dividing the forecasters into groups: bank economists, academics and and others, a distinction that prevails today. Terry later reported that a surprisingly high proportion of the people we recommended had agreed to join the panel.”
An analysis of 10 years of the early survey results compiled by Professor Norman in 2003 showed that the panel’s average forecasts were often more accurate than the Treasury’s. The hardest thing to get right was exchange rates.
A new analysis of 20 years of survey results to presented to the Australian Conference of Economists in Canberra this month finds that, when weighted by past performance, the forecasts of the BusinessDay panel have been more accurate than those of the Treasury overall and more accurate on most metrics.
Economist Warwick Smith of the Per Capita think tank, and Melbourne University mathematician Anca Hanea and prediction specialist Mark Burgman, examined more than 15,000 forecasts prepared for the BusinessDay survey between 1996 and 2016.
They found that a simple average of the BusinessDay forecasts performed worse than the Treasury on 11 of the 17 jointly forecast values.
But when weighted for past success, by giving forecasters that had previously performed well on a question a higher weighting for that question, the BusinessDay forecasts outperformed the Treasury in all but five fields.
The weighted average was better at predicting the budget balance, the consumer price index, unemployment, wage growth, household spending, housing investment, the terms of trade, the current account deficit and world economic growth.
The five topics in which Treasury held its own were Australian GDP, Australian domestic demand, net exports, Chinese economic growth and OECD economic growth.
“The survey harnesses the wisdom of crowds,” Mr Smith said. “Different people come at problems from different perspectives and with different methods. Each brings in information, some of it subconscious.”
“Treasury modelling, on the other hand, needs to make sense in its own terms.”
“What we have done is to weight more highly the forecasters who have done extremely well in the past on a topic, and to underweight those who have done badly so that the worst forecasters on each question contribute almost nothing, and the best a substantial amount.”
Thirteen of the 26 forecasters taking part in the 2018 mid year survey published in today’s BusinessDay expect the Reserve Bank to lift its cash rate in the year ahead, 10 expect it to leave the rate steady, and three expect it to cut rates further. None expect the economy to go into recession in the coming financial year, although taken together they assign a 21 probability to a recession in the next two years.
Ten are women.
The study finds the women in the BusinessDay panel have a substantially better forecasting record than the men, being right 6.5 per cent more often.
“The finding about women is robust,” Mr Smith said. “I get the impression that women are more likely than men to skip questions where they are not confident in their answer. If so, that would help explain their greater accuracy.”
The study also finds that forecasters working for banks and and industry groups are more accurate than academics, something the authors attribute to academics being more likely to operate on their own without forecasting teams or computer models.
“The rely more on gut feeling, but having that in the survey adds diversity,” said Mr Smith.
Taking all the forecast questions together the weighted average performed better than 80 per cent of the individual forecasts, where as the Treasury performed better 60 per cent.
“I think Treasury would do well to set up a system like it that runs alongside their models,” Mr Smith said. “It is missing out on information that the BusinessDay survey captures."
Mr Smith will outline his findings at the conference on June 12.