If you’re an agricultural analyst, statistics are your lifeblood. But where do statistics come from?
Statistics come from raw data. Raw data comes from surveys. And surveys come from people talking to farmers.
But what if farmers don’t tell the truth?
Chances of that are small. And if producers do fudge the numbers, deliberately or unwittingly, there are lots of ways to account for that, statisticians say.
“We can usually tell if a response from someone just doesn’t fit in,” said Daniel Kerestes, livestock division head for the USDA’s National Agricultural Statistics Service.
Kerestes and other analysts attended a livestock statistics users seminar in Winnipeg Sept. 15.
Farmers and others sometimes express skepticism at agricultural forecasts and market analyses, especially when they turn out wrong.
Actually, though, the long-term accuracy of such forecasts is usually pretty good, said Shayle Shagam, a senior livestock analyst with the USDA’s World Agricultural Outlook Board.
“Part of the problem, of course, in generating any forecast is that there’s variables we just can’t account for.”
A prime example is weather. Shagam admitted USDA’s crop outlook for Australia has been off seven years in a row because of a drought that just won’t go away.
Unforeseen developments, such as BSE, changes in policy or fluctuating exchange rates can also throw forecasts off.
But statistics themselves (e. g., how many hogs you have) are immutable. And it’s in a producer’s interests to provide accurate data, said Kerestes.
“We feel they have a genuine interest in responding correctly because the data is used to make up their marketing decisions,” he said.
Kerestes noted agri businesses and investors use statistics to advise producers on what to buy and when to sell. As a result, farmers have a vested interest in giving surveyors the right numbers.
“It’s in their interest to have the best numbers out there because they’re not competing against their neighbours in the States any more. They’re competing against neighbours internationally. If they don’t make good marketing decisions, they’re out of business.”
Robert Plourde, who heads up livestock estimating for Statistics Canada, said his department has sophisticated ways of cross-checking and verifying survey data.
“We don’t ask only one question. We ask 10 or 15 questions,” said Plourde. “If the numbers don’t jibe, there’s a flag to us: have a look at that. Something is wrong.”
Statistics Canada produces quarterly reports on livestock, crops, farm income and prices. It is also responsible for the Census of Agriculture, conducted every five years. The last census took place in 2006. The next one will occur May 10, 2011, with results released exactly one year later.
Questionnaires and survey results always include internal checks to make sure the sampling is right, said Plourde.
“Everything has to make sense and add up.”
Shagam said surveys are not like public opinion polls, in which replies can vary depending on how a question is phrased. Surveys are objective and there’s no interpretation about the number of animals on a farm on a specific date, he said.
Kerestes said USDA’s response rate on most surveys is 85 per cent, whereas socio-economic surveys usually range between 20 and 25 per cent.
And, unlike them, there’s no room for feelings in an agricultural survey, he said.
“You don’t bring a personal bias in. You can’t adjust the number of farrowings based on feelings.”[email protected]