The economy is growing strongly, the unemployment rate has been at or below 4.5 percent for 16 straight months, but wage growth remains disappointingly low.

Wages are growing much more slowly than the last time we had sustained low unemployment rates, the late 1990s. This is the notorious “wage puzzle,” a topic that has been the subject of much speculation by everyone from Federal Reserve Chair Jay Powell to pretty much every economist I follow on Twitter.

A variety of explanations have been proposed: Perhaps wages lag because there’s more labor market “slack” than there appears to be (the result of low labor force participation), or employers may be wielding new market clout against workers, or the trend could be a byproduct of the fact that wages are still effectively adjusting from the recession. Or perhaps inequality itself produces a dynamic that drives down wages.

My interpretation of the wage puzzle produces a different answer: This is simply what a high-pressure economy looks like when productivity growth rates and inflation are both relatively low. Some of the other explanations — including inequality, an especially popular answer these days — appear to point in the wrong direction. For instance, despite what you may have heard, economists broadly accept that in the past few years, wage growth has actually been stronger at the bottom than at the top.

My preferred explanation for the wage puzzle involves a reinterpretation of the rapid increase in average wages in the late 1990s, a period that is often held out as a model for what we should be trying to achieve now.

Tight labor markets contributed to the wage growth then — just as tight labor markets contribute to wage growth today. But in the late 1990s, average wages also were particularly strong because of both high productivity growth and the large increase in inequality that led to very rapid increases in wages at the very top. In contrast, today inequality is actually falling in some dimensions and rising more moderately in other dimensions.

The moderation of the pace of increase in inequality in no way takes away from the urgency of combating inequality, which remains very high by almost any standard. But it does make it hard to argue that recent wage stagnation is driven by inequality.

It is a mistake to think that the deck is so stacked against workers that any gains in efficiency will just go to increased corporate profits. In fact, the biggest problem for wages in recent years has not been that corporations are getting all the gains but that the gains are not nearly large enough overall.

What is the wage puzzle?

To explain the wage puzzle, we first need to define it. The central question is: Why are we not seeing the pace of wage growth we saw in the late 1990s? Specifically, why is the 3.4 percentage point wage growth in hourly earnings for prime-age wage and salary workers over the past three years lower than the 4.8 percentage point increase in wage growth for the same group in what I will call “the late 1990s”? (For my purposes, this is from the first quarter of 1998 through the first quarter of 2001.)

The growth rate over time is shown in the figure below, smoothed over three-year periods to highlight some of the underlying trends. I am using one particular wage measure, but other measures tell the same story.

(Note: I am limiting my focus to prime-age workers because it reduces but does not eliminate some demographic issues that affect comparisons of overall wage numbers. Older workers tend to be paid more, so when they retire and are replaced by younger workers, that artificially lowers reported average wage growth. A lot more of that shift from older to younger workers is happening now than in the late 1990s, which creates issues in comparing the two periods.)

We need to adjust for inflation, which is lower now than in the late 1990s. When we do that, it changes the picture considerably.

In the late 1990s, inflation grew at a 2.7 percent annual rate, compared with the much more subdued 2 percent annual rate over the past three years. (In the past year, a spike in energy prices has turned real wage growth negative: Gasoline and heating bills took a larger bite out of paychecks. But this is just the flip side of the large fall in energy prices that led to unusually high real wage growth a few years ago. The three-year window I am using averages out much of this volatility.)

While inflation varies a lot from year to year, especially because of those volatile energy prices, the underlying inflation trend and expectations for the future were higher in the late 1990s than they are today. This slowdown in price inflation is, itself, somewhat puzzling. It may in part be caused by a general anchoring of expectations as price setters expect the Federal Reserve to reliably keep price inflation low, a view that can become self-fulfilling. This anchoring effect may be stronger today after 30 years of low inflation than in the late 1990s when memories of high inflation in the late 1970s and early ’80s were still fresh in people’s minds.

Whatever the explanation for lower price inflation, it is clear that workers ultimately care about how much their purchasing power is going up. And with lower inflation now than in the late 1990s, any increase at all in paychecks goes a bit further. When we take lower inflation into account, instead of a 1.4 percentage point slowdown in nominal wage growth, we only need to explain a 0.7 percentage point slowdown in real wage growth (that is, wage growth adjusted for inflation).

Inequality does not explain the slowdown in wage growth relative to the late 1990s

Over the past 35 years, we have seen a substantial increase in inequality for a variety of reasons, which include a slowdown in increased educational attainment, the erosion of labor unions, the erosion of the minimum wage, and increasing monopsony power by employers — that is, the market power to set wages.

Does this help explain the slower wage growth in recent years? The answer is no. In fact, inequality has been complicated lately but, on balance, has ameliorated the wage slowdown, not made it worse.

We are well-trained to recognize the fallacy in the statement “workers should be happy that average incomes grew strongly.” Most people would quickly respond: “That is misleading because you are looking at average growth, which might tell you more about how the very high earners are doing than how middle-income workers are doing.” But we seldom ask whether the reverse might be true — if the disappointments many observers have been expressing in the average wage numbers may reflect the opposite phenomenon.

Put another way: Could recent wage growth be stronger at the low end of the wage scale than in the middle or upper middle of the distribution? In fact, that’s exactly what appears to be the case, at least partially.

In the past three years, workers in the bottom two-fifths of the income distribution have seen comparable or faster real wage growth than they did in the late 1990s. Higher earners, meanwhile, have seen much slower wage growth than they did in the 1990s.

This trend is clear in the data but little understood among the broader public. In the three-year period I am focusing on, we have seen some narrowing of inequality, measured as wages at the top relative to the bottom (the 90-10 ratio), although there has been a continued widening of inequality relative to the middle (the 90-50 ratio).

And note that the slowdown in the increase in inequality is not just for this particular measure for this particular period — again, contrary to conventional wisdom. CEO pay skyrocketed in the 1990s, for instance, but since then it’s been stable (at a very high level, to be sure) relative to the wages of average workers. Similarly, the share of income going to the top 1 percent rose much more sharply in the two decades before 2000 than the nearly two decades since.

So a substantial part of the wage puzzle involves upper-income workers. The large majority of the difference in the slowdown in average wage growth relative to the late 1990s is due to slower wage growth for workers at the top.

So with that in mind, which would we rather have: the 3.4 percentage point annual wage growth in the past three years, or the 4.8 percentage point annual wage growth in the late 1990s? It depends who “we” are. Roughly the bottom third of the income distribution would be better off under today’s lower growth rate because with inequality rising by less (and decreasing in some respects) they are getting a greater share of the subpar wage growth.

The majority of workers, however, would still be better off with the faster productivity growth of the 1990s, even though it was accompanied by faster increases in inequality. The median worker, in particular, is experiencing real wage growth that is 0.5 percentage points lower than it was in the late 1990s. That is less of a slowdown than for the mean worker that we usually look at in the data (0.7 percentage points), but it is still meaningful. What could explain it?

With productivity growth slowing, it follows that wage growth would slow too

Lower productivity provides much of the answer. The standard economic theory is that wages are determined by productivity. When workers in one country can build, say, cars better and more quickly than workers in another country, their average wages will be higher. This helps explain why average wages are higher in the United States than in India. Like most standard economic theories, this one is missing a lot, but it is the best place to start in understanding wages.

Productivity does a good job of explaining the evolution of average wage growth in the United States as well, especially prior to 2000. (But it does a decent job even since then, even with the wage slowdown.) From the end of World War II until around 2000 average wages grew almost in lockstep with productivity — if you use the same measure of inflation for both concepts, so the comparison is apples to apples. Often, presentations of the comparison between wages and productivity, as in this much-reproduced graphic, use a higher inflation measure to adjust wages than productivity and thus produce a misleading impression.

Productivity growth has slowed dramatically in recent years in the United States and almost all other advanced economies as well. Productivity growth was spectacular in the late 1990s, rising at 3 percent annually. It has been dismal in recent years, rising at only a 0.7 percent annual rate. That’s part of the explanation for slow wage growth: Based on the productivity numbers alone, one would predict that average wages would be growing about 2.3 percent more slowly than they did in the late 1990s.

This one fact more than explains the 0.7 percentage point slowdown in real wages relative to the late 1990s or the slightly smaller 0.5 percentage point slowdown in median wages since then.

What about “slack”?

One of the most widely cited explanations of the wage puzzle is that there is more “slack” in labor markets today than in the late 1990s. It’s important to note, however, that this is not true according to the standard measures economists use to assess how tight labor markets are. The official unemployment rate is similar to what it was in the late 1990s, and the short-term unemployment rate (those unemployed 26 weeks or less), a potentially better measure of slack, has been even lower.

The “quit rate,” the number of individuals voluntarily leaving their jobs (and not for retirement) as a percentage of total employment, is another measure of the confidence workers have in the job market. It’s also around the same as its highs in 2000 and early 2001.

Nonetheless, some economists have emphasized that the fraction of people between ages 25 and 54 working is still 2.3 percentage points below where it was in 2000, and have argued this is an indication that there is still substantial slack — that overall labor force participation has fallen.

I am skeptical of this argument (as is Paul Krugman). The biggest reason is that the labor market is not behaving as if it has 2.3 percentage points more slack than it did in the late 1990s. If we had that much slack, you would expect much slower wage growth. You’d also expect the labor share of income to continue to fall, yet we’re seeing modest increases on this score.

If anything, after adjusting for inflation and productivity, the labor market is behaving as if it has less slack than it did in the 1990s. (I will not bore you with a number of the other considerations about using broader measures of slack, but, if you are interested, see some of my Twitter threads.)

The implications for policy

Given my explanation for the wage puzzle, what can we do to boost wages? The most immediate answer has to do with the most straightforward dial we have in economic policy: the Federal Reserve’s setting of interest rates.

Labor markets are much tighter now than they were just a few years ago and are roughly as tight as they were in the late 1990s. But that does not mean we could not do even better. We have no idea where full employment actually is. I see little downside to continuing to push the envelope — in fact, more people have reentered the workforce in the past two or three years than I had been expecting.

Interest remains at remarkably low levels, especially given the low unemployment rate. The Fed’s gradual rate hikes do not amount to hitting the brakes; the Fed is just easing off the accelerator a little. So I think the Federal Reserve has it just about right.

The more durable efforts to raise wages should involve combating inequality. There is clearly scope to raise wages for workers at the bottom and, hopefully, the middle, without harming productivity very much and in some cases even helping it. The progress we have made in decreasing some measures of inequality in recent years is an argument for more policy in this area, not less.

The strong wage growth for workers at the bottom of the distribution is partly the result of efforts that have raised the minimum wage in 21 states and DC since the beginning of 2014. This success should encourage us to do more.

We should increase competition through more vigorous antitrust enforcement, by strengthening workers’ bargaining powers, and by reducing the ability of companies to use measures like no-poaching contracts to get leverage over their employees. There is still a lot of room for workers to get the type of larger share of the economic pie that they did in the past.

Greater productivity growth holds the potential of being the most powerful source of sustained wage growth across the income spectrum, but we likely do not have policy tools that will quickly and sustainably increase growth. I’d take a chance on every tool we have (provided it does not have negative side effects). That means investing more in infrastructure, research, and education, along with a genuine tax reform. We can improve economic efficiency while making revenue levels more sustainable by raising top individual and corporate tax rates and improving the tax base.

The wage puzzle is not particularly puzzling when you think about what determines wage growth. The policy solutions are even less of a mystery. We may not have the solution for magically improving productivity. But there’s a lot we can do to continue to increase wages for all workers.

Jason Furman is a professor of practice at Harvard Kennedy School and a senior fellow at the Peterson Institute for International Economics. He was chair of the White House Council of Economic Advisers from 2013 to 2017. Find him on Twitter @jasonfurman. 

The Big Idea is Vox’s home for smart discussion of the most important issues and ideas in politics, science, and culture — typically by outside contributors. If you have an idea for a piece, pitch us at thebigidea@vox.com.

How fast will the economy grow over the next 10 years? The question is not purely academic. For those of us who think about the trade-offs inherent in spending and tax policies, this is a question with great significance. With a faster growth rate the budget deficit would be smaller — taking pressure off deficit reduction and possibly even opening the door to tax cuts or spending increases.

In short, regardless of your political persuasion, your chances of implementing the policies of your choice depend a great deal on economic growth. And yet, recent history offers a number of reasons to expect relatively low growth rates in the future. Moreover, it is harder to push growth rates up than members of either party like to admit.

On the campaign trail, President Donald Trump promised we would see growth of 4 percent or more. Most economists were skeptical, given that the average annual growth rate since 2001 has been 1.8 percent. The budget proposal that President Trump’s team released Tuesday dials things back a little, but still assumes growth of 3 percent annually by 2021.

But how plausible is growth of even 3 percent a year at a time when certain demographic trends — most notably, the aging of the US population — have been weighing down the economy? And how big a difference could the types of policies promised by the Trump administration make — both the ones they say would boost growth like deregulation of the financial sector, repealing the Affordable Care Act, and cutting taxes, or the ones that would harm it, like restrictions on immigration and trade? What lessons does all of this have for attempts by policy to promote inclusive growth as we move forward?

The sources of growth for the US economy

Economic growth can be broken down into two sources: the underlying potential growth rate of the economy and the cyclical boost or hit it gets as the unemployment rate falls or rises. A few years ago, both factors were important, as we were recovering from a severe recession and the unemployment rate was falling by more than 1 percentage point per year. But now, with the unemployment rate hovering just below 4.5 percent — at or near “full employment,” by some measures — there is no longer much of a cyclical boost, as shown in Figure 1. The unemployment rate is now declining by only 0.2 percentage points per year.

There may be some capacity for additional short-run cyclical growth — through additional government spending or tax cuts, for example — but any future cyclical boost is limited by the fact that the Federal Reserve would likely try to offset any additional demand from fiscal policy, given its mandate to maintain price stability. (That is, the Fed will raise interest rates if it believes inflation is a threat.) As a result, looking over the next decade, overall economic growth will have to come almost entirely from the economy’s underlying potential growth.

And where does that growth come from? The economy’s potential growth is the sum of the growth rate of two factors: total labor hours and output per hour. Total labor hours are influenced by population growth, by changes in the labor force participation rate (the share of the population working or looking for work), and by changes in the average number of hours per worker. Output per hour, meanwhile, is also called labor productivity. Below, I delve into both factors to develop a range of plausible outcomes for growth over the next decade.

How much do we expect the labor force to grow over the next decade?

Of the sources of potential economic growth, population growth is the easiest to predict: It depends on fertility choices, which in most cases were made decades ago; on the trajectory of mortality, which is relatively predictable; and on immigration, which can be affected by policy.

What is particularly striking is the changing age structure of the US population. The population ages 25 to 54 — a cohort in its prime earning working years — grew 2.2 percent a year in the 1980s, but by just 0.1 percent in the past decade, as shown in Figure 2.

The slowdown in US population growth, combined with the aging of the US population, indicates that the pace of immigration will play a much bigger role in promoting economic growth — or in retarding it, if there are major restrictions — in coming decades. In fact, according to analysis by the Pew Research Center, absent immigration, the working age population (25 to 64) would decrease over the next decade and beyond.

For any given population, and any given age cohort, the total quantity of labor supplied could also rise if people work longer hours. But the long-term trend has been toward shorter workweeks (although that has stabilized in recent decades, as shown in Figure 3).

For the analysis below, I assume that hours are unchanging over the next decade, which is consistent with assumptions underlying the forecasts produced by the Congressional Budget Office and the forecasts that I helped oversee in the Obama administration. If anything, this could be a little optimistic, because as society gets richer and more productive, decreases in average hours worked become more likely than increases.

As we look ahead, the bigger unknown is what fraction of the population will be working. The labor force participation rate rose from 1950 to 2000, as the large number of women entering the workforce outweighed the men that were slowly but steadily leaving over that entire period. (For more on the decline of male labor force participation see this report that I supervised at the Council of Economic Advisers.) We interpreted the phenomenon as a combination of declining demand due to factors like job-displacing technology and globalization and the fact that the US government does much less than governments in other countries to help workers find jobs.

Since 2000, participation rates for both men and women have been falling. The key question is whether this downward trend will continue. Since no one knows for certain, I consider growth under three different assumptions shown in Figures 4a and 4b:

• The base case, which is consistent with CBO’s assumptions, is that participation rates are unchanging for each age group. Under this assumption, even though participation rates within each age group remain unchanged, the aging of the population will still drag down the participation rate by 0.2 percentage points per year over the next decade (because older Americans tend to participate in the workforce less than younger Americans.)
• In the pessimistic case, I assume instead that the same trends that prime-age workers experienced from 2000 to 2016 will continue at the same pace over the next decade. That assumption would translate to an annual decline of 0.3 percentage points in the overall labor force participation rate over the next decade.
• In the optimistic case, I assume the opposite — that prime-age workers see their labor force participation rate undo all of its losses from 2000 to 2016 over the next decade. Even under this very rosy scenario, the labor force participation rate still declines because of the powerful force of aging — but only by 0.1 percentage points a year.

Even the base case in Figures 4a and 4b, which is consistent with CBO’s projections, represents the triumph of hope over experience, in my view: It is a considerably better trajectory than men have experienced in more than a half century, or that women have experienced in more than 15 years.

In fact, Harvard’s Lawrence Summers has argued that even the assumptions underlying the pessimistic case may not be sufficiently pessimistic, as a result of a combination of technological development, declining marriage rates, slowing educational attainment, and the “contagiousness” of not working.

On the other hand, the optimistic case would represent an effectively unprecedented reversal of recent trends. Nonetheless, I will consider it as I map several possible paths for growth.

How much will productivity growth rebound — if at all?

The other major factor underlying economic growth is productivity growth — a subject that remains something of a mystery for academic economists and policymakers. Productivity growth has slowed in nearly all of the advanced economies, including the United States — although the United States has enjoyed the fastest productivity growth of the G-7 economies over the last decade (see Figure 5).

While this global slowdown has triggered a major debate among economists as to its causes and the outlook for productivity growth in the future, that debate is beyond the scope of this piece. You can find my perspective here.

As with future participation in the workforce, I will explore what happens under three possible scenarios for productivity growth. CBO’s assumption, which I will again adopt as the base case, is that labor productivity will grow 1.7 percent annually over the next decade — essentially the same as productivity growth in the 1980s or since 2001.

Given the history of variability of this data — see Figure 6, which shows trailing 10-year averages for productivity growth (meaning the data point for 1980 shows average productivity growth from 1970 to 1980) — to create a plausible range I will make the optimistic and pessimistic case. This time, I will do so using the 90th and 10th percentiles of 10-year productivity growth averages since 1958, when the dataset begins. The full set of assumptions is shown below. 

The base case is for productivity growth of 1.7 percent, the pessimistic case 1.4 percent, and the optimistic case 2.8 percent. Notably, all three of these assumptions reflect an increase of annual productivity growth from its 1.2-percent rate over the past 10 years — and the even more dismal 0.6-percent rate over the past five years.

Note, too, that the leading techno-pessimist Robert Gordon, of Northwestern University, also assumes that productivity growth will rebound. In his case, he assumes the equivalent of 1.6 percent growth in the economy going forward, placing him between my pessimistic case and base case. In the optimistic case, productivity growth over the next decade would be similar to growth from 1995 to 2005, when the internet transformed the economy for both businesses and consumers.

Bringing these assumptions together: what’s a plausible growth rate over the next decade?

Having now specified a range of scenarios for the most important variables, we can put the pieces together to consider a range of plausible outcomes for annual potential real GDP growth over the next decade. These are shown in Table 1. 

Under the base case scenarios for both labor-force participation and productivity growth, GDP growth is projected to average 1.8-percent growth a year — which is intentionally done to match CBO’s estimate, and also happens to be the median estimate by members of the Federal Open Market Committee (FOMC), a group that expects long-run growth rates somewhere from 1.6 percent to 2.2 percent. The Blue Chip Economic Indicators, a survey of private forecasters, is somewhat more optimistic, with a consensus expectation of 2.2 percent long-run growth.

In the upper left corner of the table is the “most pessimistic” scenario of 1.4-percent annual growth — which reflects both low productivity growth and the most pessimistic labor-force participation trend. It’s worth pointing out that this scenario is not at all implausible: It assumes that trends in prime-age labor force participation since 2000 continue and that productivity growth rebounds somewhat from its recent pace.

The lower right corner of the table shows that if everything goes right — an absolutely unprecedented reversal of labor force participation trends and an unusual but not unprecedented burst of productivity growth — economic growth would average 3.2 percent a year over the next decade. (See here for more details of how these figures are computed.)

The difference between the lowest projected growth rate, 1.4 percent, and the highest, 3.2 percent, is substantial, but so is our uncertainty about the future. Notably, however, all of these are below the 3.5 percent average growth rate of the economy from 1950 to 2000 — reflecting the demographic fact that an aging population will lead to slower workforce growth going forward.

Can policy boost growth?

How much can policy affect the growth rates shown in Table 1? A little, but not a huge amount.

It is worth noting that the fact that nearly all the advanced economies have seen a slowdown in productivity growth makes it unlikely that the US productivity slowdown has been caused primarily by US-specific factors like the Dodd-Frank reforms or the Affordable Care Act. Correspondingly, reversing these policies would be unlikely to undo the productivity slowdown.

That said, policy can certainly make a difference for growth. To get a sense of how much of a difference, I compiled growth estimates for several policies, relying in each case on credible sources.

To be sure, these are illustrative estimates, and one could argue with many of them. The CBO’s dynamic score of the repeal of the Affordable Care Act, for example, is based on the removal of provisions that may create a disincentive work due to subsidy phase-outs or increased capital investment due to lower taxes on capital income. But it does not count the benefits that ensue when healthier workers take fewer sick days, or when workers feel free to pursue new opportunities without fear of losing insurance.

Not included in this table are estimates of the growth effects of deregulation, which many have cited as the primary policy basis for much faster economic growth. Unfortunately, it is very hard to study and predict the macroeconomic consequences of the accumulation of regulatory changes in a wide range of areas, and I am not aware of credible estimates that would let us assess specific policies going forward.

One estimate of the effects of deregulation comes from Douglas Holtz-Eakin — a serious economist, albeit one who comes at these issues with strong anti-regulatory views. He believes that the annual growth rate would be 0.06 percentage point higher in the absence of the Dodd-Frank Act and the Basel III rules, an international framework for bank regulation. Even on the terms of this estimate, which ignores the benefits of regulations that make a financial crisis less likely, it is notable that the partial repeal of Dodd-Frank that the Trump Administration is likely to pursue would have even smaller growth effects.

Likewise, the Mercatus Center at George Mason University, which advocates for deregulatory intervention, has estimated that all of the regulations since 1980 have lowered the growth rate by 0.8 percentage point annually. Even granting this estimate — and I have doubts about the methodology that produced it — Mercatus’s figure still implies that partially rolling back the Obama administration’s regulations would have an annual impact that was a small fraction of that. After all, Obama-era regulations amount to only a fraction of the total new regulations since 1980 — and are comparable, in fact, to the pace of regulation during the 12 years of Presidents Reagan and George H.W. Bush.

Those are all policies that could increase growth. On the other hand, increased deficits, reduced public investment, curtailed immigration and new trade barriers — all of which have been proposed at various points by President Trump — would all reduce growth. In most cases the downside is symmetric with the estimates in Table 2: If $4 trillion of deficit reduction would add 0.1 percent to the annual growth rate over the next decade then $4 trillion of deficit increases would subtract a corresponding amount from growth. In some cases the policy risks are asymmetric. For example, the Trans-Pacific Partnership (TPP) took America’s already low tariffs and lowered them a little bit. The President’s campaign promise of a 45 percent tariff on Chinese goods and a 35 percent tariff on Mexican goods would effectively undo more than a half century of tariff reductions — hurting the economy far more than TPP would have helped it.

What’s the upshot?

Ultimately, policymakers should be concerned with both increasing growth and helping to make sure that growth is shared. This analysis contains a few broad lessons to guide these efforts:

Lesson 1: Advocates overstate the benefits of the policies they like and the costs of the policies they oppose. Consider: Over the past century, Argentina has done just about everything wrong with its economic policies and institutions. It has seen waves of populism, abrogation of the rule of law, massive deficits, multiple defaults, misguided industrial policies, political upheaval, large-scale price controls, corruption, and much more. The result of all of this? A per capita growth rate that was only 0.7 percentage points lower than that enjoyed by the United States.

Taking a less extreme example, France scores well below the United States on just about every measure of regulation and economic freedom, but its workers produce just as much in an hour as American workers. Any difference that President Trump could make for US policy, for better or for worse, is much smaller than the differences between the United States and Argentina, or even between the United States and France. (Although the large-scale trade war he promised in the campaign could test that proposition.)

Lesson 2: All else equal, do everything you can for growth, because over time a few tenths of a percentage point really do matter. At the beginning of the 20th century, the United States had GDP per capita that was roughly 50 percent higher than Argentina’s. Today it is about three times as high — all because of a 0.7 percentage point difference in annual growth rates.

Small differences in growth rates can cumulate to a lot over time — for example, a 0.2-percent increase in the growth rate over a decade is worth about $1,000 in additional income for the typical family. In fact, the largest factor in the slowdown in the growth of median household income since 1973 in the United States has been the concurrent slowdown in productivity growth.

All else equal, we should be doing everything we can to push growth up by even a few tenths a year (like investing in infrastructure or sensibly reforming the business tax system) — and to oppose anything that that would lower it even by a few tenths (like drastically restricting immigration or increasing the deficit).

Lesson 3: But it is generally not worth making large distributional or other sacrifices for a little additional growth. As I have argued at length elsewhere, most of the standard estimates of the effects of tax cuts find that their distributional impact is much larger than their effect on incomes as a result of growth. Consider the $1,000 windfall I mentioned that a family gets over a decade when the growth rate increases by 0.2 percentage points a year — a growth increase more than twice as large as what CBO estimates would result from ACA repeal. That is effectively meaningless if they lose government benefits and services that are worth more than $1,000 — and for many households the premium tax credit reductions or Medicaid cuts in the House’s American Health Care Act are more than 10 times as large as that.

Also, a focus on growth, to the exclusion of other factors, overstates the improvements in people’s well-being to the degree higher growth comes from longer hours for workers, or reductions in consumer consumption. Similarly, environmental or safety deregulation may increase GDP growth by a little bit, but still might not pass a cost-benefit test when the lives lost are considered.

This is not to say that there are no policies that could boost growth without these side effects, or that sometimes a trade-off of growth for higher inequality would not be worthwhile. But large sacrifices in well-being or equity are typically not worth small increases in growth rates.

Lesson 4: It is best to be realistic and conservative in formulating plans for the future. You need very optimistic assumptions about the future to support Secretary Mnuchin’s claim that growth of 3 percent or more is possible. Such a growth rate is not impossible, especially in the short run, but sustaining it would require everything to go right for participation and productivity in ways that are either historically unparalleled or toward the upper end of the historical range. Policies can make a small difference, but nothing like the difference between the 1.8 percent growth rate base case and a 3 percent or more projection.

Assuming growth rates much higher than recent experience — or even most of past experience — is unwarrantedly optimistic. To the degree that such optimism is the basis for excessive tax cuts or increases in defense spending, the result would be high deficits that would themselves hurt growth. While these growth effects may be small on an annual basis, they would compound over time in ways that would increasingly matter to people. Optimism that is detached from reality could carry a growing economic price, one that would fall heavily on average American families.

Jason Furman is a senior fellow at the Peterson Institute for International Economics. He served as a top economic adviser to President Barack Obama during the previous eight years, including as the 28th chair of the Council of Economic Advisers from August 2013 to January 2017, acting as both Obama’s chief economist and a member of the Cabinet.

The Big Idea is Vox’s home for smart, often scholarly discussions of the most important issues and ideas in politics, science, and culture — typically written by outside contributors. If you have an idea for a piece, pitch us at thebigidea@vox.com.