One thing astute readers of my missives may have noticed is that I almost never write about theoretical models. The reason is simple, most models in economics and finance are useless.
John von Neumann famously said:
“With four parameters I can fit an elephant, and with five I can make him wiggle his trunk.”
Now consider that in many firms I have encountered discounted cash flow models and similar valuation tools that use sometimes dozens of parameters, many of which like the equity risk premium and the liquidity premium, highly volatile and unobservable.
But predicting stock market valuations is one thing. Where it gets really dangerous is when economists think they can predict optimal policy decisions that impact entire countries and the people living in it.
This is why I am so angry at economists who thought they know how to manage the Covid-19 crisis and model optimal policy measures to fight the pandemic. Have they looked at the performance of their previous model predictions on things they have studied for years and decades? And now they think they know how to handle a pandemic?
I am an empiricist and rely on observations rather than models, but just for fun, let’s have a look at economic models and their predictions on the effectiveness of infrastructure investments. As we try to climb out of the hole the pandemic has pushed our economy in, fiscal stimulus is a necessity. And we know empirically that government spending on infrastructure is one of the most efficient ways to stimulate the economy. Fiscal multipliers for infrastructure spending are generally above one, meaning that every dollar spent on infrastructure creates more than one Dollar in additional economic growth.
But how much exactly?
The chart below shows the estimated fiscal multiplier for infrastructure investments in the short run (i.e. in the first year or so after government investments). A neoclassical model (the ones typically dominating the policy debate into the 1960s and again in the 1980s and 1990s) would say that the short-run multiplier of infrastructure spending is 0.4. That is not only less than one (thus indicating that infrastructure spending is wasteful spending) but somewhat similar to the fiscal multipliers found for tax cuts (which are typically in the area of 0.3 to 0.4). So, from a policy perspective, it is just as effective to invest in infrastructure as it is to cut taxes. That is why conservatives in the 80s and 90s became so enamoured with tax cuts and privatisation. The models they used told them that it works.
Meanwhile, politicians on the left of the political spectrum trusted Keynesian models more (they were the dominant models in the 1930s and again in the 1970s and over the last decade). These models predicted a fiscal multiplier of 1.1 in the short run, indicating that not only is infrastructure spending more effective than tax cuts, but it is a veritable growth engine, which explains why politicians on the left tend to favour government spending. Your policy recommendations and preferences depend on the model you use.
But wait. Infrastructure projects typically are long-term projects that take a year or more to ramp up. If we introduce a 1.5 year ramp-up phase to infrastructure spending, the fiscal multipliers in the short run drop significantly in the Keynesian model because job creation and actual spending simply don’t happen that much in the first year of the project. As a result, the short-term fiscal multiplier in the Keynesian model drops to almost zero.
Short-run fiscal multipliers for infrastructure investments
Source: Ramey (2020).
At which point it becomes clear that for long-term projects one has to look at the long-run fiscal multipliers (i.e. fiscal multipliers for growth over three to five years or so). The chart below shows that the long-run fiscal multipliers for both the neoclassical models and the Keynesian models are above 1 and the delays in ramping up the projects don’t matter much.
However, while neoclassical and Keynesian economists would now agree that infrastructure investments are a growth engine, they disagree about how much growth you can create with that. Neoclassical models, typically used on the conservative side of the political spectrum would say that the fiscal multiplier is around 1.3 giving you a 30% boost for every Dollar spent on infrastructure while Keynesian models indicate a fiscal multiplier of 1.8, thus providing a boost of 80% for every dollar spent. No wonder government investment and spending have different priorities for politicians on the left and the right. They look at different models and thus come to different conclusions.
Long-run fiscal multipliers for infrastructure investments
Source: Ramey (2020).
But, with a simple trick, we can make the estimated fiscal impact from infrastructure spending much bigger and push it above 2, indicating that every dollar spent on infrastructure more than doubles its value in terms of economic growth. Who could say no to that?
Well, all we have to do is play around with something as sexy and easily understandable to politicians and the public like the “exponent for government investment in the aggregate Cobb-Douglas production function”.
A short explanation for the uninitiated: Economic models depend on the assumption of how investment and consumption translate into economic growth. A Cobb-Douglas production function is the most commonly used function that describes this relationship. However, nobody has ever directly observed a production function because it cannot be done. Instead one has to make several assumptions that plausibly connect inputs with outputs. The Cobb-Douglas production function makes intuitive sense and at the same time is mathematically easy to manipulate. So, we just took that and stuck with it. Not because it is an accurate description of reality, but because it was simpler math and seemed to make sense at the time.
Within the Cobb-Douglas production function, one can model economies of scale which are expressed by the exponent for government investment (or any other investment). If this exponent is positive, then there are economies of scale and the government has a distinctive advantage in building large scale projects. If it is negative, then smaller, independent players are better at building things.
Throughout the first part of this post, we have used an exponent of 0.05, indicating some economies of scale. In the previous chart, we have simply increased the exponent from 0.05 to 0.1 and lo and behold, government infrastructure investing becomes so much more effective. By the way, nobody knows how big these economies of scale really are. Most economists just use values between 0.05 and 0.1 for convenience because the results seem to make sense, but I have seen models use 0.0 and up to 0.4. You can imagine what the policy recommendations are for these models…
So now you know a little bit more about how the sausage is made, do you really want to trust economists’ policy recommendations on taxes, government spending, or how to fight a pandemic? Depending on your political leaning, you will always be able to find an economist armed with a model that provides the outcome you like.
Of course, there are circumstances when I do trust economic models. There are situations when the large majority of models agree on the outcome of certain policies. For example, free trade is universally positive for economic growth, which is why I am a big proponent of free trade. But as the last five years in US and UK politics have shown, even in these situations that have been settled a long time ago, you will always find some loony fringe economists who pretend that trade wars are good for the economy and Brexit will boost growth. Which way the elephant wiggles his trunk all depends on the parameters in the model…
PS: Empirical studies indicate that the long-run fiscal multiplier of infrastructure investments is somewhere around 1.5, so roughly in the middle between neoclassical and Keynesian model predictions.
This post reminds me of one of my favourite jokes: And economist, mathematician and logician are on a train that crosses over the border into Scotland and they see a black cow. The economist says: All the cows in Scotland are black. The mathematician says: There is at least one cow in Scotland that is black. The logician says: There is at least one side of one cow in Scotland that is black.
Thanks for the thoughtful blog. I always enjoy reading it.