An Equilibrium Model of “Global Imbalances” and Low Interest Rates

Here’s the puzzle that kept a lot of macroeconomists up at night in the two thousand s: why did safe real interest rates drift down for decades while the United States kept running big, persistent current account deficits—and somehow, at the same time, the world kept stashing a growing share of its wealth in U.S. assets? Caballero, Farhi, and Gourinchas offer a single storyline that links all three. Don’t look first at saving gluts or single shocks. Look at who can manufacture financial assets, and who can’t. Their organizing idea is disarmingly simple. Real investment—planting projects that throw off cash—creates the financial assets that savers can hold. If one region is really good at turning projects into tradable, trustworthy claims and another isn’t, then when the second region grows or its local assets collapse, its savers reach across borders for safe storage. The United States becomes the main shelf where the world puts its savings. That pushes up the U.S. share in global portfolios and pushes down the world interest rate. And because those inflows are claims on U.S. residents, the United States runs external deficits. One mechanism, three facts, moving together. To make this concrete, they boil the global asset market down to a tree. Not a metaphor you can’t touch—a bare asset that pays a steady dividend. The return on the tree is the dividend yield plus the capital gain: in words, the interest rate equals the cash flow you get per dollar of asset value plus how fast that dollar’s price is rising. And because saving has to sit in some asset, total wealth equals the total value of all trees. That identity is powerful. It means the world interest rate is the single price that clears a very simple market: the supply of financial assets that investment can generate versus the global demand to save. Now split the world into two regions. Call one "U" for the United States, the other "R" for the rest. They share the same growth rate and the same basic patience, so they face one world interest rate. But they differ in a crucial way: The United States can supply more safe, tradable assets per unit of investment. Region R is the constrained side; sometimes its ability to turn projects into assets deteriorates. Add one friction—home bias. Households try to satisfy their saving needs at home first, and only after exhausting local options do they buy foreign assets. With that in place, a negative shock to Region R's asset-supply capacity does three things at once. It tightens the supply of assets globally, so the world interest rate falls to coax more saving into the remaining stock. It shifts portfolios toward the United States because that’s where quality assets are easier to produce. And it leaves the United States with a persistent current account deficit because foreigners are buying claims on the United States and the income stream those claims demand must be paid. The logic mirrors the old Metzler diagram: the asset-supply curve shifts left, the saving-demand curve presses right, and the price—here, the interest rate—moves to clear the gap. Directionally, this is clean: the interest rate drops below the United States autarky rate and the United States current account goes negative at impact and stays negative. The punchline gets stronger if Region R is the fast grower. Imagine Region R’s economy expanding more quickly than the United States, but its financial system can’t create assets at the same clip. Demand for assets in Region R rises faster than local supply. Where do those savers go? Back to the United States. In their formal results, faster growth in Region R, or a higher propensity to save there, makes the United States' long-run deficit larger and pushes the world interest rate lower. It’s the same mechanism, just with the volume turned up. That’s the theory. Do the numbers come out in the right ballpark? The authors discipline the model with three anchors from U.S. data. First, the output-to-wealth ratio: using two thousand four balance sheets, U.S. household net worth was about forty-eight trillion dollars against twelve trillion of gross domestic product, or GDP, so the ratio of output to wealth sits near zero point twenty-five. Second, trend growth is set near three percent. Third, they back out how much of output is "capitalizable"—the part that can support asset claims—by targeting a six percent autarky interest rate. In their simple accounting, that pins the capitalizable share at roughly zero point twelve. Those three numbers lock down the baseline. Then they replicate a shock that actually happened: a big drop in the value of emerging-market assets around the Asian crisis. Think of stock price declines on the order of twenty-five percent in that bloc. In the model, that’s a fall in Region R’s asset-quality parameter to about zero point zero eight. They also revisit Japan’s earlier bust by calibrating a decline in Japan’s asset-creating capacity consistent with a roughly thirty percent Nikkei drop. These aren’t meant to be point estimates; they’re order-of-magnitude matches to real swings. Run the clock forward in a three-region world—U.S. plus a Japan block plus an emerging-rest block—and you see the mechanism at work. After the Japan shock in the early nineteen nineties, global demand tilts toward U.S. and emerging assets, nudging world rates up a hair—on the order of a tenth of a percentage point—for a few years. The U.S. current account deepens into deficit, about six percent of output in that early window, and U.S. net foreign assets fall to roughly negative one-fifth of output. It’s the allocation story: portfolios moving toward the United States as its assets absorb more of the world’s saving. The Asian crisis shock does the other half. When Region R’s asset supply collapses in nineteen ninety-seven, world rates step down—about sixty basis points lower—and U.S. external imbalances widen. The current account dips to around seven percent of output in deficit terms, and net foreign assets slide toward minus three-quarters of output. These moves don’t snap back next year. If the deterioration in Region R is permanent, the model settles into a world of lower safe rates by roughly half a percentage point, a long-run U.S. current account near a three percent deficit, and net foreign debt stabilizing just shy of its annual output. If the shock is temporary—think a few decades—those steady states are milder: the current account deficit closer to two percent and the net foreign asset position around negative three-fifths of output. Different assumptions about persistence, same qualitative picture. They push the envelope further by splitting the rest into two—an "M" region that starts integrating in nineteen ninety and a remaining Region R. Give Region M weak asset-generation, faster growth, and a lower stand-alone interest rate, and the world looks even more like the data. Early on, U.S. current-account deficits near four percent of output and net foreign assets around negative fifteen percent sit comfortably within historical ranges, while the world rate drifts below its autarky benchmark. Over the very long run, if Region M keeps growing without building a matching financial infrastructure, the drag on the world interest rate persists, the U.S. current account remains negative—think roughly two to three percent—and net foreign liabilities swell. The U.S. share in global portfolios rises from low double digits toward the mid-teens, which is exactly the "U.S. as the world’s balance-sheet hub" the authors point to. Two extensions add realism and nuance. First, let firms invest in new trees, and allow U.S. investors to finance planting in Region R through foreign direct investment. This opens a new channel. In the short run, the prospect of foreign direct investment increases demand for assets—investors need to fund those projects—so the world interest rate falls even further on impact. Over time, as the new trees come online, the global supply of assets expands and the interest rate inches back up toward its autarky level. There’s a compact way they summarize the steady state after a Region R shock: the long-run rate equals the United States' autarky rate minus a term proportional to how much of the world’s initial output sits outside the United States and to the gap in asset quality between the United States and Region R. It’s a mouthful, but the intuition is clean: bigger outside world, bigger asset-quality wedge, a lower interest rate. The distribution of foreign direct investment rents then decides who pays for what. If a healthy slice of the rents accrues to U.S. intermediaries, the United States can sustain a small, permanent trade deficit—on the order of a few percent of output—financed by those intermediation earnings, even as the interest rate normalizes. Second, let the world produce differentiated goods and give consumers a home bias. That flips on exchange-rate dynamics that feel familiar. When Region R’s asset supply collapses, the U.S. real exchange rate jumps—about a ten percent appreciation in their calibration—then, over a decade or so, gives that back and drifts a touch below baseline. Why? On impact, global portfolios scramble for U.S. assets; their prices rise; so does the relative price of U.S. goods. Over time, as liabilities build and the United States services them, the real exchange rate depreciates. Crucially, the current account is still negative at impact and remains so, and the world rate still lands below the U.S. autarky rate. The two-good structure mainly amplifies the imbalances: slightly larger long-run net foreign liabilities—roughly half of annual output versus just under that in the one-good case—with similar pressure on rates. Step back, and the trio of stylized facts looks less mysterious. Regions differ in their ability to mint safe assets. When the underdeveloped side grows fast or suffers asset crashes, its savers reach for the U.S. balance sheet. That flow pushes down safe real rates, raises the U.S. share in global portfolios, and embeds persistent U.S. current-account deficits. By the mid-2000s, the weight of U.S. assets in the rest of the world’s portfolios had climbed markedly; in their calibration runs, the U.S. share in Region R’s portfolio jumps from single digits to roughly a third. And in the broader wealth picture, U.S. asset holdings tower over a sizable fraction of the rest of the world’s financial base—think on the order of one-fifth of its wealth and a large slice of its output—mirroring the model’s core prediction that U.S. assets become the global parking lot for savings. There are caveats, and the authors are upfront about them. The machinery is intentionally stripped down. Home bias does a lot of work. Baseline markets are otherwise frictionless, and the model targets current accounts and portfolio shares rather than consumption risk or crisis dynamics. The calibration aims for magnitudes, not tight fits, and the timing of short-run swings is not the point. But the comparative-statics engine—move asset supply and saving demand, watch the world rate and portfolio composition adjust—keeps finding the same directions as the data. That’s the virtue of parsimony here. If you’re wondering what to watch next, the framework gives a clear checklist. Growth and financial development outside the United States are the fulcrums. If large emerging markets deepen their ability to generate safe, tradable claims, the global shortage of assets should ease, safe real rates should lift, and the pressure on U.S. external deficits should soften. If, instead, growth in those regions outpaces their financial plumbing, the savings will keep coming, and the United States will keep wearing the world’s balance sheet. And it’s a flexible lens. You can layer in risk, official flows, or sovereign safety demand and still ask the same first-order question: who can create the assets the world wants? As Caballero, Farhi, and Gourinchas show, once you answer that, a lot of the last few decades falls into place—not as three separate riddles, but as one story about the global supply chain for safety.