The Setup
On May 21, 2026, a 34,000-gallon tank of methyl methacrylate started overheating at a GKN Aerospace plant in Garden Grove, California. Methyl methacrylate is a volatile, flammable liquid used to make plastics and resins. When it overheats it enters thermal runaway, a self-accelerating reaction that can end in an explosion. The Orange County Fire Authority identified two scenarios: a toxic spill of 7,000 gallons into a residential neighborhood, or a full detonation that could destroy the site and surrounding homes [1].
By Saturday, firefighters found a crack in the tank that might have been relieving pressure. Even so, 50,000 people were evacuated across six cities [2]. Governor Newsom asked for a federal emergency declaration [3]. The Orange County District Attorney opened a criminal investigation into GKN Aerospace, and a class-action lawsuit was filed.
Environmental remediation is already a $368 billion global industry growing at nearly 7% per year [4]. Companies like Clean Harbors, Tetra Tech and AECOM are the ones who show up after a disaster like this with hazmat teams and cleanup contracts. At least 15 US states are passing new chemical safety laws in 2026 [5].
So the investor question is not whether this is bad. It obviously is. The question is whether you can build a profitable screen around the companies that either survive events like this or get hired to clean them up.
How to Think About This
The first instinct is straightforward. Two kinds of company should benefit from an industrial disaster.
The survivors are large industrial companies with enough cash, profit margin and low debt to absorb a lawsuit, a cleanup bill or a regulatory crackdown without it denting the business. Picture a $50 billion company paying a $200 million fine. It hurts, but it does not break them.
The fixers are environmental services and engineering firms whose revenue rises every time something goes wrong. These companies bill by the hour, and the disaster is their demand signal.
Both sound sensible. Here is where the thinking has to get more honest.
The survivor screen has a problem. The traits you would use to find companies that shrug off regulatory shock, meaning high margins, strong cash flow, low debt and good returns on capital, are the same traits that describe almost every well-run large company in America. Apple survives regulatory shock. So does Johnson and Johnson. So does Costco. There is nothing specific about the screen. You are really asking Stax to find good companies and calling it a safety thesis.
That is a useful thing to notice, because it changes the test. The question is not whether safe companies beat unsafe ones. It is the harder question of whether a safety-specific screen beats a generic quality screen. If it does not, the thesis is not wrong so much as redundant.
The fixer screen has a different problem. Environmental remediation firms are mid-sized specialists. They are profitable, but their revenue spikes after disasters and flattens in between. Across a five-year test you are averaging over years when disasters happened and years when they did not, and the screen cannot capture a surge in demand that began in the last week of May 2026.
So here is what was tested. Not the vague question of whether safety stocks are good, but three specific ones.
| What I'm really asking | How the Stax screen tests it |
|---|---|
| Can companies with fat margins, low debt, and high capital efficiency outperform because they can absorb regulatory shock? | operating_profit_margin ≥ 15, free_cash_flow_yield ≥ 2, debt_to_equity ≤ 1.5, current_ratio ≥ 1.5, ROIC ≥ 12, market_cap ≥ $10B |
| Can profitable companies in the cleanup/remediation space outperform? | operating_profit_margin ≥ 10, free_cash_flow_yield ≥ 3, ROE ≥ 15, debt_to_equity ≤ 2, market_cap ≥ $2B |
| Or does none of this matter , and you'd be better off just buying big profitable companies ranked by momentum? | net_profit_margin ≥ 5, market_cap ≥ $50B, 100% momentum ranking |
What the Data Revealed
All three strategies ran from January 2020 through December 2024 with $100,000, equal weighting, monthly rebalancing, quarterly reconstitution, realistic trading costs of $0.005 per share commission and 0.1% slippage, and risk controls of a 20% hard stop, a 15% trailing stop and a 35% maximum drawdown limit that halts the plan.
The Safety Resilience screen found 17 stocks. That number is the story. When you stack six fundamental filters on the entire market, requiring high margins and high returns on capital and strong liquidity and low debt and a large market value and positive free cash flow all at once, you are drawing a diagram with almost nothing in the middle. The strategy made only 23 trades in five years. It barely moved. The companies that passed were mature and capital-efficient by design, so they did not fall much, which gave the screen a low 7.0% worst drop. They also did not rise much during a period when the market was rewarding risk-taking rather than risk-avoidance. Total return: negative 1.9%. A savings account would have paid more.
That makes sense given what was happening between 2020 and 2024. The Federal Reserve cut rates to zero, flooded the system with stimulus, then raised rates faster than at any time since the 1980s. That environment rewarded speculative growth companies during the cheap-money phase from 2020 to 2021, then companies with pricing power during the inflation phase from 2022 to 2024. Mature, boring, safety-first industrial companies got neither tailwind. They did not crash, but they went nowhere.
The Remediation Cash Flow screen found 117 stocks and lost 2.8%. More companies and more trades, 150 across five years, but the same outcome: underwater. Worse, it fell 25.1% at its lowest point. Your $100,000 would have dropped to $74,900 at the worst moment and still finished down. The profit factor was 0.95, meaning that for every dollar the winning trades made, the losing ones gave back about a dollar and five cents. This screen found real companies in a real industry, but across a five-year average that industry does not generate consistent enough demand to overcome the fact that many of these mid-sized firms are volatile, capital-hungry and cyclical.
The momentum control returned positive 24.7%. The simplest strategy won. It looked for companies worth $50 billion or more with at least a 5% net profit margin, sorted them purely by how much their share price had risen over six months, and rotated monthly. It found 212 companies, made 193 trades and returned 24.7% over five years with a Sharpe ratio of 0.15. That Sharpe is not impressive, and it came with a real 30.6% worst drop. But it is positive, which is more than either thesis strategy managed.
Why did it win? Because this window included the post-COVID rally from 2020 to 2021 when rising prices kept rising, the inflation trade in 2022 when energy and commodities surged, and the AI boom across 2023 and 2024 when Nvidia alone went from $150 to $500. A screen that follows rising prices catches all three waves. A safety screen, by design, avoids them.
| Strategy | Total Return | Sharpe | Max DD | Win Rate | Trades |
|---|---|---|---|---|---|
| Chemical Safety Resilience | -1.9% | -1.07 | 7.0% | 52.2% | 23 |
| Remediation Cash Flow | -2.8% | -0.34 | 25.1% | 46.0% | 150 |
| Large-Cap Momentum Control | +24.7% | 0.15 | 30.6% | 53.4% | 193 |
What You Can Do With This
The lesson is not that safety screening cannot work. It is that this particular version was too restrictive to trade and not specific enough to differentiate. Both of those are fixable.
Drop the current ratio filter from the safety screen. A company worth $10 billion or more does not need current assets at 1.5 times current liabilities to survive a crisis, because it can issue commercial paper or draw a credit line overnight. That filter probably excluded aerospace and defense companies that are exactly the kind of survivor the thesis wanted, and removing it might triple the universe.
Run 2022 to 2024 on its own. The full window includes the post-COVID rush when everything risky did well. If your thesis is that safety screens earn their keep during stress, test them during the stress: rate rises, the Silicon Valley Bank collapse, inflation. The safety screen shallow 7.0% drop might read very differently in a window where the momentum screen is taking damage.
Blend the two approaches. The safety screen ranked candidates 40% on price behaviour and 60% on fundamentals, while the control was 100% price behaviour. Try the safety filters with the ranking weighted 70% toward price. You keep the quality floor but let the ordering follow what drove returns in this window.
Watch the real-world catalyst. If Garden Grove leads to stricter chemical storage rules, meaning new inspection mandates, higher insurance requirements and larger cleanup bonds, that demand shift shows up in remediation company earnings over quarters, not days. A backtest cannot capture a regulatory change that has not happened yet, but you can save the screen and run it again in six months.