I don’t spend a lot of time thinking about macroeconomics or monetary policy. I did spend part of my career in public market investing, but I realized that the operation of businesses and the power of technological innovation were more compelling to me than interest rates and macroeconomic indicators. Every so often, however, I have a train of thought that spans both the world of tech startups and the world of macroeconomics. Listening to Mohammed El-Erian discussing Fed policy on the radio yesterday resurfaced some thoughts I periodically turn over in my head about the experimentation occurring at the Federal Reserve, and how this compares with the experimentation engaged in by the technology innovation ecosystem. The two types of experimentation have markedly different risk profiles, and make me wonder how we can implement low-risk experimentation into government policymaking.

It is widely acknowledged that the policy actions taken by the Fed and other central banks since the financial crisis have been unprecedented and experimental. Since 2009 the Fed, through its quantitative easing program, has purchased about $3 trillion in mortgage-backed securities, federal agency debt, and long-term Treasuries. The sheer scale of this program is difficult to comprehend. It is true that history holds examples of central bank balance sheet expansions, in the US and elsewhere, that were comparable in magnitude to the recent post-crisis expansion, but these were usually to finance wartime spending as in WWI and WWII. Today’s massive balance sheet expansion is intended to stimulate demand and employment, which has little or no precedent in history. In addition, the Fed has stated its intention to eventually unwind this massive expansion, which has no precedent because the Fed did not shrink its balance sheet after WWII.

No one knows exactly what consequences the Fed’s actions will have. There are many who argue strongly that they’re certain of what the consequences will be (rampant inflation, etc.), but these commentators are armchair quarterbacking. The only thing one can say for sure is that a $3 trillion experiment is likely to have some unforeseen results. The scale of the experiment means the consequences, where they occur, are unlikely to be trivial. The full impact may not be observed for many years to come. Fed policy doesn’t just impact the US economy, either. Given the US dollar’s status as the world’s reserve currency, the impacts are likely to be global and possibly even greater in developing countries than at home in the US.

Startups do lots of experimentation as well. Like the Fed in recent years, startups make big, bold bets in the face of tremendous uncertainty, and there is little precedent for what they are doing. From this experimentation emerges lots of failure and lots of brilliance. Yet the innovation ecosystem is markedly different from the world of monetary policy, because the costs of failure are well-contained (without constraining the upside of successful experiments). A failed startup has zero systemic impact on the technology industry as a whole. Venture capital portfolios are set up to accommodate failure; it is expected that many startups won’t succeed. A loss on one investment is made up for by gains on another. A failed startup is more difficult for the entrepreneurs involved, because they lack the diversification enjoyed by their VC backers. They pour themselves wholeheartedly into a single venture. Yet thankfully, our tech industry is so dynamic that a failed entrepreneur can typically rebound quickly and either move on to a new startup or work for a bigger company. In today’s environment of acqui-hires, failed startups can even make a return for their founders and investors, though this feels more like a cyclical rather than structural feature of the industry.

Within this environment that is so highly robust to failures, aggressive experimentation is encouraged and rewarded. The Lean Startup philosophy has provided a framework for conducting these experiments as efficiently as possible, and technologies ranging from infrastructure-as-a-service to 3d printing help accelerate the process by shortening the feedback cycle.

By contrast, central banks are not robust to failure. A failed monetary policy experiment can have a detrimental impact on the entire global economy and billions of people in some form or another. Experimentation at the Fed lacks the characteristics that make experimentation by startups such a powerful force for improving standards of living, namely that the costs of failure are borne by a few but the potential benefits to society as a whole are unbounded.

This raises to me an important question: how can governments and central banks conduct policy experiments where the costs of failure are contained? Policies must evolve over time, and there are surely beneficial but untested policies out there. It would seem that a logical method is to try out policies on a local level and adopt the most successful policies on a larger scale. The legalization of marijuana in Colorado and Washington is a good example of local policy experimentation that could pave the way for a national policy shift. Unfortunately, I’m not aware of any ways the Fed can conduct legitimate policy experiments where the cost of failure is contained, but perhaps they exist or will exist in the future.

At the risk of sounding like I’m trumpeting the tech industry as a model for the way everything should work, it seems like governments and maybe even central banks could benefit by adopting systems that permit low-risk experimentation. I am sure there are others who have thought about and studied this more than I have. I look forward to a day when policymaking is both more innovative and less systemically dangerous.