NOTE: The Growth Economics Blog has moved sites. Click here to find this post at the new site.
Over at the Cato Institute, they hosted an online forum about reviving economic growth. There are lots of smart people involved. The web page has lots of big pictures of their heads, I guess to indicate that their brains are like, totally huge.
Anyway, each one wrote up some proposed policy reform that would help boost long-run growth prospects. Brad DeLong responded to many of the proposals here before his head exploded reading Doug Holtz-Eakin’s essay.
I’m not going to quibble with any of the minutiae of the proposals. My point is going to be a general one on the possible growth effects of [insert policy here]. Short answer, there won’t be any.
There are two ways to boost GDP growth. Either
- Actively raise current GDP through increased spending by some sector of the economy.
- Raise potential GDP and let transitional growth speed up.
The second one perhaps deserves a little explanation. Transitional growth is an extra boost to growth that occurs when current GDP is below potential GDP. Why does this occur? Bob Solow is why. In an economy with accumulable factors of production (physical capital, human capital, knowledge capital) being below potential GDP means that the return to these factors is relatively high, and hence more investment in those factors is done, boosting GDP growth. The wider is the gap between current and potential GDP, the stronger this transitional growth.
The issue is that [insert policy here] is a policy to raise potential GDP, not current GDP. But the transitional effects this encourages are inherently small. So even if [insert policy here] opens up a big gap between potential and actual GDP, this doesn’t translate into much extra growth. In fact, the effects are likely so small that they would be unnoticeable against the general noise in growth rates year by year.
To give you an idea of how little an effect [insert policy here] will have on growth, let’s play with math. Output in period can be written in terms of output in period this way
This says that output in is equal to times current output. That is “regular” growth. The term with the is the additional boost in growth we get from being below potential. is potential GDP in period , and is the gap in GDP. tells us how much of that gap we make up from period to . If , then we are stuck below potential (secular stagnation). If , then immediately next period our GDP will be at potential again.
Let’s think about this in terms of growth rates, so
The growth rate from to depends on the ratio of potential to actual GDP today, period . If that ratio were equal to one – meaning that we were at potential – then the growth rate just becomes , the trend growth rate. The larger is – meaning the farther we are from potential – the higher is the actual growth rate.
Now we can go back to thinking about the possible growth impact of [insert policy here]. GDP today () is about 16 trillion. Potential GDP today () is probably about 17 trillion. You can get a lower estimate from the CBO, Robert Gordon, or John Fernald, or a higher estimate from older CBO forecasts. I’m going to err on the high side for potential because this will inflate the growth effect of [insert policy here].
We also need to know the value of , the percent of the GDP gap that is closed in a year. We’ve got lots of evidence that this value is about , or 2% of the gap closes every year. This estimate goes back to the original cross-country convergence literature starting with Barro (1991), but consistently across samples (countries, US states, Japanese prefectures, Canadian provinces, etc..) economies converge to potential GDP at about 2% of the gap per year.
You get higher values of if you assume that economies pursue optimal savings plans, like in the Ramsey model, meaning that they save at a higher rate when they are farther below steady state. But if there is an economy that saves according to the predictions of the Ramsey model, it is populated by unicorns.
Back to the calculation. The last thing we need is a value for , trend growth. Let’s call that , or trend growth in GDP is about 2% per year. Again, we can argue about whether that is higher or lower, but that’s not going to be the important factor here.
Okay, so based on the fact that we are currently 1 trillion below trend, the growth rate today should be
or growth should be 2.13%. Growth will be about 0.13 percentage points higher than normal – that’s a little over one-tenth of one percent – because we are below potential. The value of is really irrelevant. All the action is inside the brackets. Because is small, there isn’t much bite from transitional growth, even though we are $1 trillion below trend.
But what about [insert policy here]? That will *raise* potential GDP, and therefore will induce faster transitional growth to the new, higher potential GDP. Okay. Let’s say that [insert policy here] has an astonishingly positive impact on potential GDP. I mean massive. [insert policy here] adds a full $1 trillion to potential GDP, which is now $18 trillion. Now, growth under the [insert policy here] regime is
Uh, wow? Growth will be an additional 0.12 percentage points higher thanks to [insert policy here]. This is not a massive change in growth. And the growth boost will *decline* over time as we get closer to potential.
Fine, but what if [insert policy here] is truly revolutionary, and raises potential GDP by $2 trillion? Then growth will be 0.0238. This could be generously rounded to 0.025, meaning you added a half-point to the growth rate of GDP. But let’s not kid ourselves that [insert policy here] is going to have that big of an effect on growth. $2 trillion implies that [insert policy here] is raising potential GDP by about 12%. That would be an anomaly of historic proportions.
[insert policy here] will not generate any appreciable extra economic growth, even though in the very long-run [insert policy here] may be a net positive for the level of economic activity. The problem is that it takes a very, very, very long time for those positive effects to manifest themselves, and thus [insert policy here] won’t do anything to fundamentally change GDP growth.
What about the exceptions I mentioned? Among the proposals, there are a few that could boost current GDP (and thus growth) directly and immediately by encouraging spending.
- Scott Sumner’s NGDP targeting. The proposal speaks directly to raising current GDP, as opposed to raising potential GDP. I think of this as solving the balance sheet problems of households. Boost nominal spending and nominal incomes rise, while nominal debts like mortgages remain fixed, leading to extra spending.
- Brad DeLong’s raising K-12 teacher salaries. If you could do it *now*, then it would raise incomes for these folks, and boost spending. The second part of the proposal, to tie this to teacher tenure changes, is more of a potential GDP changer. Question, how big of an impact would this really have on spending?
- A number of people mention infrastructure spending. Yes, if we would spend that money *now*, then it would materially boost GDP growth *now*, and as a bonus have long-run benefits for potential GDP.
Ultimately, the issue in the U.S. right now is not with potential GDP. We do not need policies to raise this potential GDP so much as we need policies to get us back to potential. That requires actively boosting immediate spending.
Thanks for the link! They broadcast some events online. Maybe you know some other economic conferences broadcasted online?
I’m sorry, but I don’t agree with this neoclassical growth model…no doubt it looks fancy, but it is incapable of explaining the complexity of the growth process!
Two things on this. First, the convergence speed stuff (lambda) is an empirical finding, and thus doesn’t necessarily depend on the neoclassical growth model. So use whatever model you want.
Second, the neoclassical growth model isn’t intended to capture complexity in growth. It’s specifically designed to simplify this horrifically complex process down to something manageable. The fact that it still has some explanatory power for growth rates and levels of income per capita is pretty amazing, actually, given how simplistic it is.
It does not simplify anything, it simply captures the identity, and what you call ‘some explanatory’ power comes from this.
You do not need to make neoclassical assumptions to arrive at identical results, all you need is to assume stability of the parameters in the identity! Eventually, what you get as total factor productivity is just a weighted average of the growth of capital and labour, again, derived directly frim the identity. See Khan (1974); Solow (1974) on his unsuccessful defense and Khan (1980) in exposing it…
I was glib, for sure. I just meant that production cannot instantly jump to a new potential level. I think that is what you are commenting on, but correct me if I missed your point.
Are all the proposals really about increasing the level of potential GDP? What about something addressing technology spillovers or the like? Where does endogenous growth theory fit in to all this?
Or, looking at it another way, you’re arguing that they’ll all only add marginal changes on top of the trend 2% growth rate – but what is the cause of the trend? Can’t we influence that?
Adam – in general, no, it is really hard to influence the trend growth of 2%. See my reply to Luis. If you look at the US, through 140 years of policies and history the trend growth of GDP per capita is about 1.8% per year. Nothing causes that to vary, so I don’t believe that any of the Cato proposals would either.
The fact that the long-run growth *rate* is constant and unchanging is perfectly compatible with endogenous growth theory. We can influence the *level* of economic activity (the potential GDP) through changing policies, but would have a hard time changing the growth rate itself.
Why? The problem is that growth in GDP depends on growth in the “stock” of innovations. And the more innovations we have, the bigger absolute increase in innovation is necessary to keep the growth rate higher. So it takes more and more effort just to keep the growth rate constant.
“If you look at the US, through 140 years of policies and history the trend growth of GDP per capita is about 1.8% per year. Nothing causes that to vary, so I don’t believe that any of the…”
Do you agree that, within the next 30 years, computers costing under $50,000 will be equal to–or superior to–human beings in measures of general intelligence (such as standard IQ tests)?
If so, don’t you think that will increase the trend growth rate of GDP per capita above 1.8% per year?
If you tell me that we’ll get AI that can innovate at 100x, 1000x, or 10000x human speed, then sure, growth may accelerate.
But policy reforms of the type suggested by the Cato meeting are not going to make that happen.
But where does g come from? Are there no policies that operate directly upon g?
In general, there isn’t any evidence that policies can have big effects on g itself. The evidence is that long-run growth rates of GDP are remarkably stable within developed countries. GDP per capita growth is even more stable – this grows at 1.8% per year, roughly, for the US over the course of 140 years. So income taxes, the New Deal, Depression, WWII, industrialization, electrification, computerization, changes to R&D, changes in college education, changes in secondary education, etc. etc. etc. were not sufficient to materially alter that growth rate. So I’m skeptical that any of the Cato growth items will do it.
also (and this is predictable) but what about China? I think that country has averaged 10% growth over a 30 year period. Maybe if you policy change is: start from Maoism, end Maoism, sort your agricultural sector out, initiative manufacturing export led urban growth, then [insert policy here] can boost growth by a very large amount indeed
Luis. Yes, China’s [insert policy here] was broadly the set of things you describe, and these made China’s potential GDP get much, much, much bigger. Say that Y(POT) in China in 1978 after the reforms was something like 40 times bigger than Y(ACTUAL) in 1978. That generates a lot of growth even with a low lambda.
The US doesn’t have policy reforms available on the scale of “end Maoism” that can raise potential GDP to 40 times actual GDP. So we can’t generate growth at 10%.
sorry. typo. Initiative should be initiate
What if lambda is partly a function GDP Pot? or, the degree of divergence matters as driver of divergence, not as a divergence itself. This means that (lack of) reforms causing limits actual GDP to reverse to GDP Potential.
You could play with that a little, I think. This is the old Gerschenkron argument about catch-up growth. It’s easier to be behind then to lead, because you can just copy.
But – the size of lambda is more about the slow accumulation of productive assets, not really about adopting technology. And empirically we don’t see lambda being very big. If lambda were really big, then China would be growing at like 30% a year. We don’t see that, so it seems that lambda isn’t huge.
We only need more money spent on AI, AI is the future for everlasting growth
I for one welcome our new robot overlords.
Pingback: Quel sera l’impact de la loi Macron sur la croissance française? | Classe éco | Francetv info
Pingback: You Can’t Reform Your Way to Rapid Growth | The Growth Economics Blog