General Motors had a pretty good quarter in the last months of 2017. Sales of its SUVs surged, surpassing those of competitors and fueling fat gross margins. It forecasts strong earnings in 2018. And of course it’s staring at a healthy tax cut, courtesy of the administration of President Donald Trump and Republicans in Congress.

Naturally, when the time came to do its earnings call, the company reported a $5.15 billion loss for the quarter.

Actually, if you know a bit about accounting, this is entirely natural. In lowering the company’s corporate taxes, the government suddenly made the ability to avoid those taxes less valuable.

Between 2005 and 2008 (when the government stepped in to bail it out), GM lost roughly $80 billion. This is a gruesome number. But the grim cloud had a silver lining: Those staggering losses created something called an “NOL carryforward.” NOL stands for “Net Operating Loss.” NOL carryforwards are a tax asset arising from those sorts of losses. You may remember the mini-scandal that NOLs created for candidate Trump during the 2016 election, because exposed tax returns showed he had used them to offset his profits and pay very little income tax.

In tax law, NOLs have a special meaning because they can offset money earned in other years. A loss “carryforward” can offset income made in subsequent years; a loss “carryback” can offset income made in previous years.

GM’s enormous losses generated an enormous future benefit, allowing the company to earn income without paying taxes on it. Quite properly, accounting rules require the firm to carry that future benefit as an asset on the company’s books.

But now that the company’s tax rate is going down, that benefit is no longer so beneficial. A “get-out-of-taxes-free card” is obviously more valuable when your tax rate is 35 percent than when it is 21 percent. So the asset’s value has to be written down, which in turn shows up as a charge against net income.

Welcome to the wonderful, weird world of accounting, where something positive, like lower taxes, can show up as a negative on your books. Luckily for GM, the market wasn’t fooled. Although GM has suffered somewhat from the market’s recent general decline, its stock got a nice bounce off the earnings call where it announced that multibillion-dollar loss.

Washington could take a lesson from this very sensible reaction: Accounting is useful, but it isn’t real.

Generally accepted accounting principles do not aim to give us a real, true, complete, unerringly accurate picture of a company’s net worth and earnings prospects, because that’s impossible. (To offer just one example, say you are a vineyard owner operating next to a tannery whose by-products are, unbeknown to you, slowly poisoning your land and vines. You may be scrupulously fair in representing the worth of your winery, but you will still be wrong.)

And that’s just one of the fascinating quandaries that beset accountants:

Should you account for revenue and expenses when cash moves in or out the door, or at the moment when you know that you owe (or are owed) a certain amount?

When you have a big pile of inventory that has been acquired over a long period of time, how do you record the cost of those goods when they’re sold? Is it the price of the first shipment you got, or the most recent, or some average of your costs over time?

Should real estate be valued at what you paid for it, or what you could sell it for today?

There is no obvious right answer to any of these questions (or the many more complicated ones I’ve left unexplored). But before a firm can issue a financial statement, we have to try to come up with something. So accountants do their best to stick to reasonable, “conservative” principles — conservative not in a political sense, but in the sense that they give managers the least leeway to make up numbers that please them.

Those “conservative” numbers are not necessarily more accurate than we could get in another way — some Platonic ideal accounting standard would probably have reflected GM’s improved expectations, not the diminishing value of its past losses. Yet these numbers are consistent, allowing investors to compare companies, and shareholders to feel reasonably secure that corporate managers aren’t cooking the books.

Washington has its own form of these dilemmas. Yet unlike the markets that saw through the confusions and reacted positively to GM’s losses, we often don’t recognize them.

Take the way Democrats gamed the Congressional Budget Office’s scoring process to get President Barack Obama’s health care law rated as reducing the deficit — and then how journalists treated those numbers as if they were a likely prediction about the future, rather than the product of a particular set of scoring rules. (That phenomenon recurred during the Republican attempt to repeal Obamacare.)

Or take the somewhat bizarre way that the government accounts for student loans, making them appear to be a profitable arbitrage opportunity rather than a substantial risk.

Or take the perennial arguments over the nature of the Social Security Trust Fund.

Discussions of these issues are frequently marred by the belief that government numbers reflect the best and truest guess analysts can make at a system’s future revenue and outlays. Of course, government analysts would like their numbers to be right and true. But just as in the private sector, they frequently publish numbers that more accurately reflect the standards they have adopted to calculate the numbers than the true state of the world.

If those standards routinely depart from reality, then they should be adapted (something that the CBO and other government branches regularly do, within the limits set by law and considerations such as keeping continuity so that you can compare one set of figures with a previously published set). But the process of adaptation is slow, and in the meantime, we will regularly get some odd results.

Just like stock traders, we therefore have to be alert to those oddities — and do our best to figure out what the numbers really tell us, rather than what they might suggest from a casual glance.