Spread too thin: The impact of lean inventories

This paper investigates the macroeconomic consequences of widespread just-in-time (JIT) inventory management, documenting a fundamental trade-off: JIT raises firm profitability and reduces micro-level volatility in normal times, but renders the economy significantly more vulnerable to large unanticipated shocks.

The empirical analysis draws on a novel dataset of approximately 200 publicly listed U.S. manufacturing firms for which the author identifies JIT adoption years using narrative records from SEC filings and historical news archives. Firm-level balance sheet data come from Compustat Fundamentals Annual (1980–2018), merged with county-level weather event data from NOAA. Four empirical facts are documented. First, JIT adoption is associated with a 13% decrease in inventory-to-sales ratios and a 9% increase in sales. Second, JIT adopters experience a roughly 7% decline in sales and employment growth volatility. Third, JIT adopters are approximately 25–30% more cyclical than non-adopters: a 1% increase in GDP growth predicts an additional 0.47 percentage point increase in JIT firm sales growth above the non-adopter baseline of roughly 1.6%. Fourth, a weather disaster predicts an additional 3% decline in JIT firm sales and employment relative to non-JIT firms.

To explain and quantify these facts, the author builds and structurally estimates a dynamic general equilibrium model with a distribution of heterogeneous final goods firms that differ in idiosyncratic productivity, inventory holdings, and JIT adoption status. Materials must be drawn from inventory stocks; new orders are subject to stochastic fixed order costs. JIT producers draw from a first-order stochastically dominated order cost distribution relative to non-JIT firms. Adopting JIT requires an upfront sunk cost and a smaller continuation cost thereafter. The model is estimated via simulated method of moments (SMM) targeting 11 moments (adoption frequency, inventory-to-sales ratios, covariances, and spike frequencies for both firm types), with nine parameters to be estimated.

In the estimated model steady state, JIT adoption delivers a 9–10% increase in output, a 40% decline in the aggregate inventory-to-sales ratio (close to the observed 35% decline in nonfarm inventories-to-final-sales from 1980 to 2018), a 1.3% increase in firm value, a 1.3% increase in measured TFP, and a welfare gain of 1.43% in consumption equivalent terms. These gains arise because lower order costs allow firms to better align material input use with realized productivity, smoothing inventory cycles.

The vulnerability side is quantified through an unanticipated supply disruption calibrated to match the 3.4% drop in real U.S. GDP between 2019 and 2020. In response, the JIT economy experiences an approximately 0.40 percentage point excess output contraction relative to the no-JIT counterfactual, amounting to roughly 13–15% more output lost. The mechanisms are stockouts — firms that fully exhaust their inventories and cannot produce — and hoarding behavior, whereby firms that retain some inventory draw stocks down more slowly to preserve buffers, reducing material input use. Both channels reduce production relative to the counterfactual. The excess output loss is estimated at approximately $100 billion, comparable to state and local government allocations under the CARES Act.

JIT nevertheless remains welfare-improving even under this shock. For a social planner to prefer a no-JIT world, the negative productivity shock to the intermediate goods sector would need to be nearly 14% — an order of magnitude larger than the calibrated COVID-19 shock. The trade-off is robust across alternative order cost distributions, parameterizations, partial anticipation scenarios, and stockout cost specifications.

Q: What is the central trade-off identified by the paper? A: JIT adoption reduces fixed order costs, enabling firms to place smaller and more frequent orders, which raises sales, reduces micro-level volatility, and increases firm value and welfare in normal times. However, because JIT firms hold fewer inventories, an unexpected aggregate shock increases the likelihood of stockouts and hoarding behavior, producing a deeper aggregate output contraction relative to an economy without JIT. Firms do not internalize the prospect of large shocks when making their private adoption decisions, generating the externality at the heart of the trade-off.

Q: How does the paper measure JIT adoption, and how large is the sample? A: The author constructs an adoption dummy for approximately 200 publicly listed manufacturing firms by exhaustively reviewing SEC filings and historical news archives for keywords including “JIT,” “just-in-time,” “lean manufacturing,” “pull system,” and “zero inventory.” Each document is individually analyzed to confirm the adoption year and to ensure it refers to the firm itself rather than its suppliers. More than half of observed adopters in the sample adopt prior to 1990, and nearly all adopt before 2000. The final Compustat-linked sample covers about 5,017 unique manufacturing firms from 1980 to 2018.

Q: What are the firm-level efficiency gains from JIT adoption? A: JIT adoption is associated with a 13% decrease in inventory-to-sales ratios and a 9% increase in sales; the corresponding standard deviation changes are –16% and +4%, respectively. Adopters also experience a roughly 7% decline in both sales and employment growth volatility, and a 5% increase in sales per worker relative to non-JIT firms. JIT firms additionally show a roughly 20% standard deviation reduction in squared forecast errors, indicating improved predictability of profitability.

Q: How much more cyclical are JIT firms relative to non-JIT firms? A: A 1% increase in GDP growth is associated with approximately a 1.6% increase in sales growth for non-adopters; JIT adopters experience an additional 0.47 percentage point increase above this baseline, making them roughly 25–30% more cyclical. This elevated cyclicality is estimated from variation external to the firm and reflects the heightened sensitivity of lean producers to aggregate demand fluctuations.

Q: How are JIT firms affected by local weather disasters? A: On average, a weather disaster predicts an additional 3% decline in JIT firm sales and employment relative to non-JIT firms. Using upstream supply chain linkages from Compustat Segment files, a unit increase in the average number of disasters hitting a firm’s suppliers predicts a 7–8% decline in firm sales and employment, with a similar excess decline for JIT firms. These results parallel the strategy in Barrot and Sauvagnat (2016).

Q: What is the model structure, and how does the JIT adoption decision work? A: The model features a representative household, a representative intermediate goods firm producing materials with capital and labor, and a continuum of heterogeneous final goods firms that differ in idiosyncratic productivity (AR(1) in logs), inventory holdings, and JIT adoption status. Each period has three stages: adoption decision, order decision (conditional on stochastic fixed order cost draw), and production decision. JIT producers draw order costs from a distribution first-order stochastically dominated by the non-JIT distribution, meaning JIT firms face systematically lower expected order costs. Adoption requires an upfront sunk cost c_s; maintaining JIT requires a smaller continuation cost c_f (estimated at slightly more than one-third of c_s), generating hysteresis: conditional on being an adopter, the probability of remaining one is estimated at 94%.

Q: What moments are targeted in the SMM estimation, and how well does the model fit? A: Eleven moments are targeted to identify nine parameters: the empirical adoption frequency, plus five moments each for JIT and non-JIT firms (mean inventory-to-sales ratio, the covariance matrix of inventory-to-sales ratios and log sales delivering three moments, and the frequency of positive inventory-to-sales ratio spikes exceeding 0.20). The model successfully fits targeted moments; non-targeted regression coefficients reproduce a quantitatively similar reduction in inventory-to-sales ratios after adoption, a comparable increase in sales among adopters, and reductions in firm volatility of 4–5% versus 6–7% in the data.

Q: What are the estimated key structural parameters? A: The upper support of the order cost distribution among non-adopters is estimated to be an order of magnitude larger than that of adopters, implying JIT firms place orders about 45% smaller than non-JIT firms. The estimated carrying cost is about 20% of inventory value. The estimated share of non-adopters in the model’s steady state implies a mass of JIT establishments of approximately 0.40. The technology parameters for the idiosyncratic productivity process are consistent with prior estimates in the structural firm dynamics literature.

Q: What are the steady-state aggregate gains from JIT adoption in the model? A: Relative to a counterfactual economy with no JIT option, the estimated model delivers a 9–10% increase in output, a 40% decline in the aggregate inventory-to-sales ratio (close to the observed 35% decline from 1980 to 2018), a 1.3% increase in firm value, a 1.3% increase in measured TFP, and a welfare gain of 1.43% in consumption equivalent terms. The TFP gain arises because lower order costs reallocate resources toward high marginal product producers at the aggregate level.

Q: How is the unanticipated disaster calibrated, and what are its effects in the JIT versus no-JIT economies? A: The disaster is an unanticipated negative shock to aggregate productivity in the intermediate goods sector, calibrated to match the 3.4% drop in real U.S. GDP between 2019 and 2020. In response, the JIT economy experiences approximately a 0.40 percentage point excess output contraction relative to the no-JIT counterfactual, amounting to roughly 13–15% more output lost. This excess loss equals approximately $100 billion, comparable to CARES Act allocations to state and local governments.

Q: What are the two mechanisms through which JIT amplifies the disaster shock? A: The first mechanism is stockouts: because JIT firms hold fewer inventories, an unexpected spike in order costs makes them more likely to fully exhaust their existing stocks, leaving them with no material inputs and forcing them to forgo production entirely. The second mechanism is hoarding: firms that do not fully stock out face a higher shadow value of inventories and cut back on material input use to draw inventories down more slowly, reducing output even without a full stockout. Both mechanisms reduce material input utilization in the JIT economy, causing a sharper drop in sales relative to the counterfactual.

Q: Is JIT still welfare-improving when the COVID-19 shock is accounted for? A: Yes. A social planner comparing welfare across steady states would not prefer to eliminate JIT even accounting for the deeper crisis it generates. For the planner to prefer a no-JIT world, the negative productivity shock to the intermediate goods sector would need to be nearly 14% — an order of magnitude larger than the calibrated 3.4% shock. This implies that the welfare gains from JIT in normal times substantially outweigh the welfare costs of the deeper recession under a COVID-19-scale shock.

Q: How does the paper relate to the Great Moderation literature? A: JIT adoption is credited in prior work (McConnell and Perez-Quiros, 2000; Blanchard and Simon, 2001; Kahn et al., 2002) as contributing to the roughly 35% reduction in the aggregate inventory-to-sales ratio between 1980 and 2018 and to the broader decline in macroeconomic volatility. The estimated model is consistent with this: JIT adoption reduces firm-level volatility and, in the steady state, implies a reduction in aggregate inventory-to-sales ratios close to the observed magnitude. However, the paper documents that the same forces that smooth normal-times fluctuations amplify unanticipated large shocks.

Q: What robustness checks does the paper conduct? A: The paper considers alternate parameterizations (all robustly show the micro-macro trade-off), larger disaster sizes calibrated to UK and France 2020 contractions (JIT economy contracts ~10% vs. ~8.7%, a ~15% larger contraction), partial anticipation (a sizable excess output drop persists because the left tail of firm outcomes is truncated at zero profits), stockout costs (trade-off remains with ~1.2% firm value gain and ~10% excess contraction), and an alternative right-skewed beta order cost distribution (firm value gain rises to 1.8%, trade-off remains). An alternative CUSUM-based measure of JIT adoption identifying approximately 560 firms produces qualitatively similar empirical results.

Q: What is the subsample estimation finding on adoption costs over time? A: Comparing 1980–1989 and 1990–2018 subsamples, the upfront sunk cost of JIT adoption estimated from the 1980s sample is about 26% higher than in the later subsample, implying it has become easier to initiate JIT production over time. Steady-state output rises by about 3.4% in the 1990–2018 period relative to 1980–1989, and the excess output contraction under the disaster shock is about 15% relative to the 1980s counterfactual, close to the baseline estimate.

Just-in-Time (JIT) Production: A lean inventory management philosophy that minimizes the time between orders by committing to smaller and more frequent orders from suppliers, reducing costs of managing large material purchases and storing idle stocks; in the model, JIT is operationalized as drawing order costs from a distribution first-order stochastically dominated by the non-JIT distribution.

Stockout: The condition in which a final goods firm enters a period with no inventories (s = 0) and chooses not to place an order, leaving it without any material inputs and forcing it to forgo production entirely for that period.

Hoarding (in the disaster context): The behavior of firms that, facing a higher shadow value of inventories during an unexpected shock, cut back on material input use in order to draw down existing inventory stocks more slowly, preserving buffers at the cost of reduced current production.

Fixed Order Cost: A stochastic, labor-denominated cost that a firm must pay each period in which it places a materials order; JIT adopters face a systematically lower distribution of these costs, enabling more frequent ordering at smaller quantities.

Adoption Sunk Cost: The one-time upfront cost c_s a non-adopter must pay to initiate JIT status, which exceeds the continuation cost c_f paid by existing JIT firms to maintain their status; the gap between these costs generates hysteresis in the adoption decision.

Simulated Method of Moments (SMM): The structural estimation procedure used to identify model parameters by minimizing the weighted distance between model-simulated moments and their empirical counterparts; here applied with 11 targeted moments to identify 9 parameters in an overidentified system.

Micro-Macro Trade-off: The paper’s central finding that individual firms rationally adopt JIT for private profitability gains (1.3% increase in firm value, 1.43% welfare gain), while the aggregate economy becomes more fragile to unanticipated shocks (roughly 13–15% deeper output contraction) because firms do not internalize the systemic vulnerability created by economy-wide lean inventories.