Thursday, April 16, 2020

Too Few Ventilators, Too Many Cabbages: Just in Time Versus Just in Case Inventory--Pandemic Edition

by Michael C. Dorf

Thanks to shelter-in-place orders or their equivalent in most of the US, most hospitals now appear likely to avoid being catastrophically overwhelmed by a flood of patients in the way that it appeared they were about to be in New York City less than two weeks ago. That's the good news. The really bad news is that the success of these measures is being used perversely as an argument by Donald Trump and his followers as a reason to abandon them prematurely. 

But even if public health sanity prevails, there will be continued bad news. The floods of patients that did materialize and are likely to continue to materialize in places where the COVID-19 pandemic has yet to peak are severely straining hospital personnel and supplies. Every heartbreaking story of patients, doctors, nurses, and others falling sick and in some instances dying due to a lack of sufficient medical or protective equipment is a terrible tragedy. Worse, each such story appears to be an avoidable tragedy that, through bad planning, we failed to avoid.

How might some of these tragedies have been avoided? Part of the answer has to do with the difference between so-called "just in time" and "just in case" production and inventory systems.

The just in time concept more or less originated in Japanese manufacturing--at Toyota plants in particular. Automobile manufacturers from the time of Henry Ford typically generated stockpiles of necessary parts and materials. If you needed to make a hood for a Model T, the warehouse delivered the sheet metal to the factory floor, where it was turned into a hood. The warehouse was kept pretty full, so in normal circumstances production would not be delayed due to a lack of sheet metal. Sounds good, right?

Well yes and no. Pre-just-in-time manufacturers stockpiled not just raw materials but also special-purpose parts. That meant that the factory workers would not run out of wheel axles or pistons, but if those were highly specialized parts it also meant waste. Let's say that you're GM, Ford, or Chrsyler making large heavy cars but then the oil embargo of the 1970s causes gasoline prices to spike, which shifts demand to smaller more fuel-efficient cars. You will eventually shift your production, but in the meantime, you have on your hands too many special-purpose parts for which you no longer have any use. That's not just a hypothetical example, of course. The penetration of Toyota and other Japanese car makers into the US market was partly a result of the greater manufacturing flexibility of the Japanese firms using just-in-time production (also sometimes called "lean" production).

As the term suggests, just-in-time production means that materials and parts arrive just when they're needed. They are not stockpiled. That has the advantage of avoiding wasted stockpiles. It also has another advantage that might at first look like a liability: just-in-time manufacturing stresses existing processes. If the carburetor parts don't show up when needed, that can delay the entire vehicle's assembly. But as a result, managers of just-in-time plants are constantly discovering weak links in their processes and improving them.

Beginning in the 1980s, firms all over the world began to adopt just-in-time procedures, and they soon spread from manufacturing to other domains. Grocery stores were an obvious candidate, given the risk of spoilage, but so were other retail establishments. Think about the risks to clothing retailers due to changing fashions. Indeed, the just-in-time philosophy can be applied beyond the physical world. In the 1990s, financial institutions sought and obtained relief from what they regarded as some unduly onerous capitalization requirements. After all, a requirement that a bank have reserves just sitting in a vault or its equivalent is a requirement for a kind of waste.

As that last example is meant to show, however, there are serious dangers that can arise out of applying just-in-time methods where they do not belong. Money held in reserve is not actually wasted money, in the same way that money spent on a fire insurance policy is not wasted, even if your house never catches fire. An insurance policy or a reserve fund makes money available just in case it is needed due to a sudden disruption like a fire or a financial crisis. If an automobile plant has to stop the line for a couple of days because it has a carburetor shortage, no one loses their home or life savings.

Seen through this lens, we might understand some of the risk to health care workers and patients due to inadequate supplies as the result of our health care system having increasingly moved to just-in-time methods when just-in-case methods were needed. Having just enough hospital beds, ventilators, masks, and gowns for ordinary usage leaves our health care system extremely vulnerable when a shock hits. We have come to understand what we should have known all along: that unused beds and equipment are not wasted; like capital reserves for a bank or a homeowner's insurance policy, they are (or rather ought to be) there just in case.

To be clear, I am not offering an all-purpose version of the morality tale of the Ant & the Grasshopper in which it is always better to be the well-prepared ant than the happy-go-lucky grasshopper. Inventory management advisors recognize that even within a single well-run business, there is typically a place for both just-in-time and just-in-case approaches. My point here is that one needs to give some thought to whether a particular business, part of a business, or other institution can run on just-in-time methods without risking an eventual catastrophe. The answer, with respect to health care, is no.

To that, I would add three further points.

(1) International supply chains are often employed with just-in-time inventory systems. A car that is assembled in Michigan or Tennessee will use parts manufactured in many other countries. More relevant to the current crisis, hospitals throughout the U.S. use pharmaceuticals made in India and China. High demand due to the spike in coronavirus outbreak here and limited supply due to demand in the countries of origin mean critical shortages. The length and international character of supply chains thus exacerbates the problems that can arise from just-in-time systems during crises when just-in-case stockpiles would serve better.

(2) In some contexts just-in-time methods might be preferable but impossible to implement. Consider agriculture. Even as the animal agricultural supply chain weakens in some respects, farms as a whole are over-producing. Apparently when people buy prepared meals at restaurants they create greater demand for food than they do when they prepare meals at home. Why? I suspect some combination of more waste and larger portions in restaurants, as well as different items being purchased (more boxes of pasta at home, more onion rings in restaurants). Whatever the reason, apparently the closure of the vast majority of restaurants except for scaled-back delivery and takeout has greatly reduced demand for agricultural products; as a consequence, farmers are simply dumping what they cannot sell (or give away to food banks, which have limited cold storage capacity). Needless to say, that's going to be very costly for farmers. If they could have produced food "just in time," they could have avoided the waste and associated cost.

There have been efforts to adapt just-in-time methods to agriculture, but they have necessary limits. Cabbages (to use an example that's mentioned in the news story linked in the prior paragraph) need to be planted about three to six months before they're harvested. Six months ago, sars-cov-2 (the coronavirus that causes COVID-19) had not yet infected a single human. True, by three months ago, the National Security Council had already warned the President of the impending public-health cataclysm, but the nation's farmers received no such warning. Farmers need to forecast demand. Even in normal times, that's a challenge. Farmers need to consider all sorts of uncertain factors--especially weather, which affects one's own production as well as production by competitors, which in turn affects price. But at least in normal times one can be fairly confident about the likely range of product quantity to aim at in order to maximize profit. In normal times, the inability to fully use just-in-time production methods in agriculture adds cost but acceptable cost. By contrast, the current abnormal demand shock will make the waste cost a matter of economic survival for many farmers.

(3) I do not mean to say that just-in-time methods have no place in the provision of public or essential services. (I regard health care as an essential service that ought to be provided to the public on a universal basis, as it is in every other developed country, but let's put that aside.) It would be wasteful for every hospital in the country to stockpile the maximum number of ventilators it might need in the event of a pandemic respiratory disease like COVID-19, given that money spent on ventilators leaves less money for equipment that could be used for other potentially lifesaving purposes. To have enough ventilators "just in case," the system as a whole needs to have enough ventilators. So long as there is sufficient coordination to move them where needed in a crisis, and so long as the crisis doesn't peak everywhere simultaneously, the total just-in-case number for the system will be lower than the sum of the number each locale would need if on its own.

Here's an analogy. Suppose that insurance regulators (or, in the absence of effective regulation, responsible insurers themselves) are trying to decide how much money needs to be available per policy. In many contexts, individual risks are uncorrelated. Automobile insurance is a good example. Although there can be multi-vehicle collisions, on the scale of large numbers, automobile accidents are uncorrelated. If an insurer were writing only one automobile insurance policy with maximum coverage of $1 million, it would need to set aside $1 million in order to be able to pay out in the event of a claim. But of course, major automobile insurers write millions of policies and so can use actuarial models that, due to the law of large numbers, are quite accurate. The average driver will have an accident roughly once every 18 years, resulting in a claim that averages about $27,000. (I've updated claim size for inflation from the source cited.). Thus, a prudent insurer will have available assets of about $1,500 per policy per year (that's $27,000 divided by 18). That figure is a little less than the cost of an average insurance policy, which makes sense, taking account of insurers' administrative cost and profit.

Insurance works to reduce the amount that needs to be held in reserve just in case per insured customer by spreading and smoothing risk over large numbers. Government can and should take parallel actions with respect to the risks imposed by local institutions having insufficient just-in-case supplies in the event of a cataclysm. Most cataclysms are at least somewhat local. A hurricane might create severe shortages of necessary supplies in Texas, Louisiana, Florida, or Puerto Rico but not in the rest of the country. So long as the federal government coordinates a response by rapidly moving supplies from where they are abundant to where they are desperately needed, the national stockpile of supplies needed "just in case" will be smaller than the stockpile needed if each jurisdiction is left to its own devices.

Seen in this light, the federal government's ineffective and haphazard relocation of supplies and personnel in response to the pandemic counts as one of the many ways in which it has performed badly. COVID-19 could have struck the whole country all at once--in which case the analogy to uncorrelated risk in automobile insurance would completely break down. But it did not. Critical need arose first in Washington State, then in California, then in New York and Louisiana. As drastic public-health measures throughout the country have taken hold, the risk of overwhelming state and local resources has substantially subsided, but it could re-emerge where those measures are relaxed too early. Should that happen, the federal government ought to be ready to move personnel and resources where they need to be deployed. Unfortunately, it does not appear ready or willing to do so.

Postscript: I am aware that typically uncorrelated risks can become correlated in unusual circumstances. That can be disastrous, but occasionally such unexpected correlation works out well. For example, as a consequence of shelter-in-place orders, people are currently driving much less and thus getting into fewer accidents. Because insurance premiums were set based on pre-pandemic levels of driving, they are now too high. Accordingly insurers are issuing partial refunds