Beware the thumb on the scale of government data
To make decisions, our politicians need data that only the government can collect. Political tampering with that data marks the beginning of the end for fact-based policy — and for a nation claiming to serve the needs of the people.
Imaginary facts are bad enough. Cooked statistics are much worse.
For me, the most disturbing headline in the last few days was not that the President watched Fox News and saw an imaginary terrorist attack in Sweden, or that he erroneously thought his electoral victory was larger than any President since Ronald Reagan. It was this, from the Wall Street Journal:
Trump Administration Considers Change in Calculating U.S. Trade Deficit
Tweak in counting exports could bolster president’s case for redoing Nafta, other trade deals
WASHINGTON—The Trump administration is considering changing the way it calculates U.S. trade deficits, a shift that would make the country’s trade gap appear larger than it had in past years, according to people involved in the discussions.
The leading idea under consideration would exclude from U.S. exports any goods first imported into the country, such as cars, and then transferred to a third country like Canada or Mexico unchanged, these people told The Wall Street Journal. . . .
Several economists interviewed by the Journal were uneasy with fully excluding re-exports from exports but not imports.
“As a statistician, you generally want symmetry,” said Steve Landefeld, former BEA director. “If you’re going to begin to exclude re-exports from the U.S. export figures, you probably for reasons of symmetry” would want to adjust import figures as well.
So the administration is considering skewing the export numbers, but not the import numbers, because that would make the trade deficit look wider.
For good policy, you need consistent data
Consider, for a moment, all the data that the government collects and how we use it.
For example, the Bureau of Labor Statistics (BLS) collects and publishes a vast amount of data about employment. It’s where we find out if the unemployment rate has gone up or down each month, and by how much. It also calculates the labor participation rate (the percentage of potentially employable workers that are working) and the total number of jobs in the economy.
These figures include uncertainty — in fact, the government goes back every month and revises the jobs figures from previous months. But the methodology remains consistent. This is what allows us to know just how much the economy has recovered, how many manufacturing jobs have been lost, and how slowly wages are increasing. The Fed relies on this data to determine whether to tweak interest rates.
Imagine, for a moment, that a president wishes to make things look like they are improving under his administration. So imagine that he tells the BLS to count unemployment differently — to, in effect, put its thumb on the scale.
How are we supposed to decide appropriate policy in a world like this? I want policy based on facts, and theories drawn from those facts, and that means we have to be able to count on the data. I may not trust the politicians, but the data bureaucracy is supposed to be apolitical.
Of course, it’s not just unemployment. The Commerce Department’s Bureau of Economic Analysis calculates not just the trade deficit, but the Gross Domestic Product. Is the economy growing, and if so, how fast? Are we in a recession? To determine these things, we need a consistent measure of GDP.
The BLS computes inflation. That’s a pretty important number to measure if you’re trying to determine interest rates or the health of the economy.
I want to know just how much money the IRS is collecting, because the size of that number is crucial for policy. There can be no lying about how much tax we collect.
The Census collects data on all of us every decade, which determines the number of representatives each state gets in Congress. But it also uses surveys between census years to measure everything from business activity to consumer spending to health data.
The Centers for Disease Control measures the spread of disease. I want my health treatments based on actual data, not skewed data that favors one party of another.
I don’t actually give a damn about the trade deficit. But I care a lot about data integrity.
People who collect and crunch the data rarely get any recognition. Well, I’m standing up for the stat geeks and the data they collect, because once they get politicized, we can no longer evaluate the decisions our politicians make. That scares the crap out of me.
What we need from government data collection
Only the U.S. government is large enough to collect and manage this data. There is no one else who can calculate GDP or trade deficits, or conduct the census. Only the government can compel people and businesses to participate. Only the government can correlate the information with decades of past data.
These are the principles we must insist on:
- The government will continue to collect data to measure the state of our nation, including all the major sources of data that feed into policy.
- If we collected data in a certain way before, we will continue to collect it in that way for at least another ten years, so we can compare statistics appropriately with past years.
- If there is a good reason to make a change, the government will collect the new statistics alongside the old for at least ten years, so we can see how the two measures differ and understand why.
- Anyone in the executive that wishes to make a change must explain why the new measure is more accurate, not just more favorable to a policy.
- Congress oversees and regulates this data collection. Changes without bipartisan agreement should set off alarms.
While I am no expert on government data collection, I know that it’s the bedrock of government policy. If any partisan politician puts a thumb on the scale, that’s a threat to democracy. And I care a lot more about that than imaginary terror attacks in Sweden.
The ‘mushroom treatment’ is a well established practice of authoritarian management.
“Keep them in the dark, cover them with horse manure and if they stick their heads up, cut them off.”
Policy driven facts are an extension of ‘alternate facts’. There is a need for calling lies out regardless of how they are ‘engineered’. There are many ways of recasting raw data to influence the results obtained by subsequent analysis – statistical or otherwise. No need to limit concerns to a specific methodology but urgent and important to stop the corruption of decision processes at every level.
Important point. The impulse to make things look better is always there and often acted upon.
The Obama administration, we learned late in its tenure, changed the way it counted deportations to seem tougher on immigration enforcement.
Every state and federal administration I can remember has come up with new ways to measure achievement in public schools. The scales fell of my eyes some years ago when a friend said, “It’s just a new way to measure 70 percent.”
Unemployment or labor force participation? Which gives a better picture? It depends.