Forecasting fails

The managers running U.S. factories — from auto parts to chemicals — are consistently getting their predictions of shipments very, very wrong: 43% of deliveries lie outside even their most optimistic or pessimistic scenarios.

And it costs them: Bad forecasts distort investment and sap productivity as factories plough money into expansions that never pay off or hold back when demand is about to surge.

That is the central finding of a new research by Eva Labro, the Michael W. Haley Distinguished Professor of Accounting at UNC Kenan-Flagler Business School, and University of Michigan professors Lindsey Gallo and James Omartian (MBA ’13, PhD ’18). They report their findings in “Internal Forecasts in Multi-Location Firms” in a forthcoming issue of the Journal of Accounting Research.

Their study is based on a U.S. Census survey that asked 24,000 factories to predict how much they would ship 18 months later, known as the total value of shipments.

The results are “awfully bad,” says Labro. “When we compare their predictions with reality, the gap is spectacular.”

According to their research, only 22% of results landed where managers thought they were most likely to land. Nearly half missed the mark completely: 23% fell short of their lowest estimate and 20% beat even their most optimistic forecast.

The forecasts were not meaningless. They captured the right direction of travel, but managers routinely misjudged the size of the swings.

Managers also proved too upbeat. They predicted shipments would rise 9.4% between 2015 and 2017, but the actual increase was less than half that, at 4.2%.

“Managers don’t even consider that outcomes could fall outside their forecasts,” Labro says. “They’re over precise, narrowing the range too much.”

The forecasting failures were worst at firms with multiple plants. Their managers forecast with more precision but less accuracy, missing by an extra 4.7 percentage points compared with single-plant factories.

The study builds on Labro’s wider research into how firms use internal information — and what happens when that information turns out to be wrong.

Labro sees her work as an effort to close the gap between research and business practice.

“I’m always looking for topics where practitioners struggle, but academia has said little,” she said. “Forecasting stood out as one of those areas.”

The study suggests plant managers compete for money from headquarters. Showing precision suggests certainty and — even if misplaced — helps them attract investment.

The head office, for its part, does not simply take these forecasts at face value. It compares plants and treats disagreements as a sign of uncertainty. That creates bad incentives: Managers who admit doubt look weak, while those who express certainty — even if wrong — win more resources.

“When managers show uncertainty, headquarters leans on other plants’ forecasts,” Labro says. “That discourages information gathering and leaves firms unprepared for rare but serious shocks.”

The stakes are high. Manufacturing depends on careful planning. Too rosy a forecast and companies risk wasted investment and excess inventory; too gloomy and they miss opportunities to expand, hire and capture sales.

Indeed, when plants missed their forecasts, productivity dropped — and the fall was clear at multi-plant firms but not at single factories.

Investment patterns mirrored the forecasts. Plants that forecast higher growth ploughed more money into equipment and staff. But when managers admitted more uncertainty or saw a bigger risk of weak growth, investment collapsed, and capital spending and hiring dropped.

For Labro, the message is this: Managers might try their best, but if the forecasts are flawed, everything built on them will be, too. “Forecasts are absolutely critical for budgets and plans,” she says. “And those plans are only as good as the forecasts are.”

The findings carry a broader warning. Multi-plant firms make up just over 20% of U.S. factories but generate more than 90% of manufacturing output.

Their reliance on bad forecasts could skew investment across the world’s largest economy. The result is a system where precision counts more than accuracy, and useful information gets lost.

Why do managers miss the mark so often? They shy away from considering extreme scenarios, says Labro. “We might not be as well prepared for these ‘black swan’ events.”

Her advice to factories’ headquarters is to accept that forecasting is hard, and don’t punish managers for admitting doubt. “An awareness of how hard it is to be good at forecasting is important,” she added. “Otherwise, you encourage a culture of false precision.”

Her message is clear: Bosses should stop demanding precision and start building systems that accept uncertainty.