Death to inventories! (Long live inventories.)

Scrunchies. Parachute pants. Corporate inventories. All were way out of fashion by the end of the ’90s. Two of three have made a comeback.

The stunning return of the scrunchie, we leave to others; this is the story of inventories.

The ’80s and ’90s were an ambitious era of supply-chain innovation, featuring jargon-y flavors like JIT (just-in-time) production, Six Sigma, and lean manufacturing. Competition and inspiration from Japan played a central role, especially the production practices developed at Toyota.

Across these philosophies, inventory was in the crosshairs—a potential symptom of inefficiency and waste. Why buy and store supplies you don’t need, or assemble products that gather dust?

“The idea was that you could have your inputs shipped just in time for when you need them in the production process,” said Maria Jose Carreras Valle, a research analyst with the Minneapolis Fed. “Not a day before, not a day later, so you can minimize the inventory you have.”

Through the ’90s and into the mid-2000s, inventories fell as firms followed these streamlining mantras, keeping fewer inputs at-the-ready and fewer final goods on the shelf.

Then around 2004, U.S. inventories began to rise. And they have been rising ever since, more or less, across nearly every industry (Figure 1).

Carreras Valle, an economics Ph.D. candidate at the University of Minnesota, says that for all the ink spilled over the decline of inventories, few economists have looked into why they bounced back. For a new paper, she set out to explain the pattern for intermediate (or “work in process”) inventories, which have experienced the largest decline and the biggest rebound.

Inventories as “China insurance”

U.S. manufacturers only began seriously sourcing from China in the 2000s, sparked by China joining the World Trade Organization in 2001. China has since become by far the largest source of U.S. imports, followed by Mexico and Canada (and then a steep drop to fourth-place Japan).

But sourcing from China comes with a big hitch: the shipping.

While COVID-19 caused the mother of all migraines, shipping headaches from China are nothing new. In 2019, according to analysis by one maritime intelligence firm, around 75 percent of ships to the U.S. from Asia arrived off-schedule, with an average delay of three days. A different set of data places on-time shipping arrivals from the Far East to North America at just 45 percent.¹

The bigger challenge is the length of the trip, said Carreras Valle. “Yes, the delivery times are volatile. But it’s also the 30 days it takes for the ship to get from China to the U.S.,” she said. For importers, “every day your demand is changing. But you need to hedge for around a month of changes until you get the new shipment.”

The ’90s weren’t wrong: Stocking inventories is still inefficient, all else equal. But with the subsequent rise of China, manufacturers have found inventories a necessary evil as a form of insurance.

“In the same way people have a savings account in case things go wrong, firms have these savings accounts of goods in case something goes wrong, or you have a demand spike,” Carreras Valle said. The import-to-inventory relationship holds across U.S. industries. The greater the reliance on imported inputs, the higher level of inventories firms are choosing to stock (Figure 2).

Weighing price vs. reliability

Carreras Valle designed an economic model to simulate firms making delivery and production decisions over multiple time periods, amid countervailing forces. On one hand, transportation and information technologies continue to improve—a steady, anti-inventory force. On the other, prices from China are falling, making Chinese inputs increasingly competitive. Those low prices, however, carry the risk of long and unpredictable shipping.

Calibrated with decades of U.S. trade and shipping data, the model generates a U-shape similar to real-world experience. Carreras Valle finds the opposing forces—technology and China—explain 61 percent of the initial decline in inventories, and 34 percent of the subsequent rise.

It’s not the entire story of the U-shaped inventory curve, but it appears to be an essential part.

U.S. importers do have other choices. For example, use more air cargo. But despite a small shift toward air, the share of China-U.S. freight going by ocean remains strong—around 80 percent in recent decades, consistent across manufacturing sectors.

Another possibility: Source the inputs closer to home. In most cases, however, the allure of low Chinese prices has been too powerful. While inputs from China rose in recent decades, the data don’t show any counterbalancing shift to import more from our neighbors to the north or south (Figure 3).

Nor have American companies shown much inclination to return to making the inputs we have come to rely on from China.

Instead, U.S. firms flexed their inventories to continue sourcing from China. Sure, the shipping is a pain. But it’s nothing we can’t forecast and buffer against…right?

And then came COVID

As in so many other areas of life, no one was counting on a coronavirus. Delays once measured in days became weeks. Manufacturers, retailers, and Christmas shoppers watched anxiously as container ships stacked up offshore.

“You don’t insure for the worst case—that would be incredibly costly,” said Carreras Valle. “Firms were not necessarily hedging for a pandemic supply chain disruption, with the worst demand-spike you can think of.”

To some extent, the existing inventory insurance against the shipping risk from China surely helped cushion the initial blow. “But in terms of companies actually being able to pull through the worst shock, obviously what we see is that they weren’t,” Carreras Valle said. “We’ve seen stock-outs, price increases—all the things the model would predict, because firms only hedge for the expected level of the risk.”

She plans to extend her model and data to reflect the COVID-19 experience. Going forward, “I would expect this will change firms’ decisions,” Carreras Valle said, as they factor the potential for another global pandemic into their cost-benefit calculation.

But as long as Chinese prices remain low, however, don’t expect a broad shift among firms to start buying American (or Canadian or Mexican, for that matter). Instead, Carreras Valle’s model and 30 years of manufacturing data suggest that when companies run the numbers, they might lean even more heavily on a once-maligned business practice: Bulk up those inventories a little more, watch for ships on the horizon, and hope for the best.

A supply-chain childhood

Few economists can trace their fascination with global supply chains back to elementary school. But for Maria Jose Carreras Valle, who joined the Minneapolis Fed as a research analyst in 2018 while pursuing her Ph.D. at the University of Minnesota, her professional focus has everything to do with growing up in the central Mexican city of San Luis Potosí.

“We have this huge industrial zone in San Luis that has all these small factories that produce car parts,” Carreras Valle said. “GM and BMW set up there, knowing that they have all this support along the supply chain. Everyone is involved in that sector—you grew up with it, even if it’s not in your immediate family.”

San Luis is well situated, combining favorable labor costs with good highways to the U.S. border. But at a crossroads of global trade, things can change quickly. “There was a Ford plant that was supposed to come to my hometown,” said Carreras Valle. “Everyone was talking about it. It was an enormous investment, and they set aside a huge piece of land.”

Then, U.S. voters elected a new president with a commitment to renegotiate the North American Free Trade Agreement and stem a perceived loss of American jobs to Mexico. In 2017, the new plant planned for San Luis fell through.

“I lived through it, and everyone lived through it in San Luis with me,” said Carreras Valle. “My research is inspired by that experience.”

Endnote

1 On-time estimates come from proprietary reports by Sea-Intelligence and eeSea, respectively, as provided by the researcher, Carreras Valle. The 30-day average delivery time in the subsequent paragraph was calculated for the year 2018 by Carreras Valle based on proprietary shipping data from Freightos.