Lane Kenworthy
That’s the topic of a New America Foundation forum, with contributions by Robert Atkinson, Josh Bivens and Heidi Shierholz, Heather Boushey, James Galbraith, Joel Kotkin, Thomas Kochan, Katherine Newman, Paul Osterman, and yours truly.
Mine is titled “Low-wage jobs and no wage growth: Is there a way out?”
The problem is especially pronounced in the social sciences. I’d guesstimate perhaps 2% of the pages in the leading economics, policy, political science, and sociology journals feature replication and/or reanalysis. We ought to aim for something in the neighborhood of 33%.
Why the continued preference for new theory and analysis? My guess is simple path dependence. Until recently data were relatively scarce. Replication and reanalysis was no less valuable than today, but it was more difficult, time-consuming, and expensive. That’s changed profoundly. Yet the norm at most journals hasn’t. It should.
Where the journals go, tenure committees will follow.
Then again, perhaps the shift can happen without the journals.
“After a recession, this economy usually gets people back to work quickly. Not this time.” That’s Clive Crook in a recent FT column.
Actually, the lack of employment recovery since the official end of the great recession looks quite similar to what happened in the early 1990s and early 2000s (more data here, here, here).
If the degree of employment recovery is our signal, at this point we might indeed forecast a lost business cycle (“decade”), as in the 2000s, with the employment rate never even reaching its previous peak. But we might just as reasonably hope for a very healthy upturn, as in the 1990s.
We should be worried not just because employment growth so far has been sluggish, but also because:
1. This recession was much more severe than the previous two. We have a good bit more ground to make up.
2. Output may recover more slowly than in the past. Heavy household debt and the collapse of home values are likely to hamper consumption growth, and the Fed and Congress seem unwilling to further stimulate the economy.
3. The pattern of the 2000-07 business cycle may indicate a fundamental shift in employer practices, with greater reluctance to hire and eagerness to fire.
Update: More here from Josh Bivens and Isaac Shapiro at the Economic Policy Institute.
Many researchers and policy makers favor a “relative” measure of poverty. That’s because our notion of what constitutes a minimally acceptable standard of living tends to be shaped by what’s typical in our own society. This approach dates back at least to Adam Smith. Its contemporary popularity owes much to Peter Townsend and Amartya Sen, and to the fact that the European Union’s official poverty measure is a relative one.
A relative poverty measure is essentially a measure of inequality within the bottom half of the income distribution. As long as this is made clear, such a measure can serve a purpose. But it’s fairly common for commentators to refer to a relative poverty rate as though it’s an indicator of absolute well-being. A journalist or opinion writer or researcher may say something like “We’ve failed to make things better for the poor; the poverty rate is the same now as a decade ago.” When hearing this, some (many?) of us will assume it reflects stagnant incomes for households at the bottom. But low-end incomes may actually have increased; that can happen without yielding any reduction in the relative poverty rate if incomes in the middle rise too.
Consider the experiences of six rich nations from the late 1970s to the mid-2000s. The following charts show trends in relative poverty and in absolute inflation-adjusted household income at the tenth percentile (a good proxy for “the poor”) in three of the six: the United States, Canada, and Germany. In each of these three countries the relative poverty rate increased slightly over the period. And in each of them absolute incomes of low-end households were flat or increased only minimally. Relative poverty rates and absolute incomes tell a consistent tale.
The story is quite different in Sweden, Norway, and Ireland. Here too relative poverty rates were flat or slightly rising. But in these three countries the absolute incomes of low-end households increased by a substantial amount. The reason the relative poverty rate in these countries didn’t fall is that household incomes at the median increased just as rapidly.
Relative poverty is of some interest, to be sure. But to avoid confusion, we might do well to sometimes refer to it as “lower-half inequality.”
A review and dissection by Wolfgang Streeck. Well worth reading.
Reducing poverty is widely viewed as a key objective of a good society. The U.K.’s Labour government set a formal poverty reduction target in the late 1990s, and the European Union recently did so as well. In the United States, public opinion surveys consistently find a solid majority saying government spends too little money on assistance to the poor.
The standard poverty measure in comparisons of rich nations is a “relative” one. The poverty line for each country is set at a percentage, usually 60% or 50%, of that country’s median household income.
Which countries have been most successful in reducing relative poverty in recent decades? And how have they done it? Here’s what the picture looks like for twenty affluent nontiny longstanding democracies. The data are from three sources: the Luxembourg Income Study (LIS), considered the most reliable for comparative purposes; the European Union’s Statistics on Income and Living Conditions (EU SILC), which covers recent years for EU countries; and the OECD.
As it turns out, there is hardly any success to explain. In almost every one of these nations the relative poverty rate was no lower in 2007, the peak year of the pre-crash business cycle, than at the end of the 1970s. The only clear exception is Ireland. Denmark also reduced poverty according to the LIS data, but the OECD data suggest little or no change. Portugal is another possibility. A few countries succeeded in reducing relative poverty during certain portions of this period, such as the U.K. in the early 2000s.
What accounts for this near-universal failure?
A relative measure of poverty is essentially a measure of inequality in the lower half of the income distribution. A nation’s relative poverty rate is determined largely by three things: wage inequality among individuals in the bottom half of the distribution, employment inequality among households in the bottom half, and the generosity of the public safety net. The wage distribution has become more unequal in many countries, though by no means all. This owes to a host of developments, including globalization, deregulation of product and labor markets, manufacturing decline, weakening of collective bargaining, and increased immigration of people with language barriers and/or limited job skills. The trend in employment likewise has tended to be inegalitarian, depending on the magnitude and character of the rise in single-adult households, the movement of women into jobs, and government efforts to promote employment. Government transfers have increased in a number of countries, but often only enough to offset the rise in market inequality. And in a few nations transfers have stagnated or decreased. (More discussion here, here, here, here, and here.)
I prefer a focus on absolute incomes and living standards rather than on relative poverty, and that approach yields a very different conclusion about progress in recent decades. Still, the widespread failure of rich countries to make any headway in reducing relative poverty rates is striking.
Taxes reduce the payoff to entrepreneurship, investment, and work effort. If taxation is too heavy, these disincentives will weaken a nation’s economy. But at what point does the harmful impact kick in? And how large is it?
A puzzle
Half a century ago, in 1960, taxes totaled about a quarter of GDP in Denmark, Sweden, and the United States. The tax take then began to rise in Denmark and Sweden, reaching half of GDP by the mid-1980s, where it has remained. In America it has barely budged, hovering between 25% and 30% of GDP throughout the past five decades.
Has heavy taxation hurt the Danish and Swedish economies? If so, how much?
Begin with GDP per capita. America’s is higher than Denmark’s or Sweden’s. But that’s a legacy of the distant past. Growth of per capita GDP in the three countries has been virtually identical, both in the five decades since 1960 when the divergence in tax levels began and in the three decades since the 1970s (shown in the chart) when the tax difference has been most pronounced.
(Here and throughout I use 2007, the peak year of the pre-crash business cycle, as the end point. Adding the crash and its aftermath would improve the standing of Denmark and Sweden relative to the U.S.)
Each year since 2001 the World Economic Forum has scored most of the world’s countries on a “competitiveness” index. The index aims to assess the quality of twelve components of a nation’s economy: institutions, infrastructure, macroeconomic stability, health and primary education, higher education and training, goods market efficiency, labor market efficiency, financial market sophistication, technological readiness, market size, business sophistication, and innovation. In 2007 Denmark and Sweden were judged to be nearly identical to the United States in competitiveness. That was true throughout the decade. It also was true for the “innovation” components of the index in particular.
Employment, measured as average hours of paid work per working-age person, is a little lower in Denmark and Sweden (more here ). A larger share of working-age Danes and Swedes are employed — around 76%, compared to 72% in the U.S. But employed Danes and Swedes tend to work fewer hours than employed Americans — about 1,600 per year versus 1,800. This is due in large part to the fact that Danes and Swedes have more than five weeks of legally-mandated paid vacations and holidays, whereas Americans have none. This gap, in turn, is a function of historical differences in the strength of unions.
Employment hours increased between 1979 and 2007 in all three countries. The rate of growth was fastest in Denmark, followed by the U.S. and then Sweden.
Household income (after taxes and transfers) is higher in the United States at the ninetieth percentile (p90) of the distribution and at the median (p50). This owes to differences in per capita GDP, in income inequality, and in the degree to which citizens receive their income in the form of (tax-financed) public services. Here too the U.S. has not gained ground in recent decades. Household incomes in the middle of the distribution have grown more rapidly in Denmark and Sweden than in the U.S. (shown in the chart), and at the ninetieth percentile they’ve increased at about the same pace.
At the tenth percentile (p10), incomes are higher in Denmark and Sweden. And they’ve increased more. (See here and here.)
Denmark and Sweden have done better than the United States at keeping government debt in check.
Have high taxes required a sacrifice of liberty? Not according to the Freedom House measure of civil liberties or the Heritage Foundation-Wall St. Journal measure of economic freedom.
Finally, consider two social indicators of well-being: life expectancy and life satisfaction. On both counts, Danes and Swedes fare, on average, just as well as or better than their American counterparts.
If heavy taxation has harmful economic effects, why have Denmark and Sweden performed similarly to the United States during a period of several decades in which their taxes were much higher than America’s?
Three explanations that sidestep the puzzle
One common explanation is that small size facilitates administrative efficiency. The Danish and Swedish governments can function effectively because their scale is manageable. They are “big” governments, but in small countries. This might be true, but to say that heavy taxation isn’t a problem if government works well is to say that heavy taxation isn’t in and of itself a problem.
A second explanation looks to the mix of taxes countries use. The Nordic countries rely disproportionately on consumption taxes; in 2007 consumption taxes totaled 16% of GDP in Denmark and 13% in Sweden, compared to just 5% in the U.S. These are said to create less in the way of investment and work disincentives than do taxes on individual and corporate income.
Yet there is a sizeable difference in income taxation too. In the U.S. income taxes were 14% of GDP in 2007, versus 19% in Sweden and a whopping 29% in Denmark. More important, to suggest that heavy taxation isn’t harmful given an effective tax mix is to suggest that a high level of taxation per se is not necessarily harmful.
A third explanation points to tax compliance. Each April most Swedes receive a pre-prepared tax form. The relevant information about income, deductions, and the amount still owed or to be refunded has already been filled in by the Swedish Tax Agency. If the information is correct, the taxpayer simply confirms that by mail, telephone, or text message. Pre-prepared tax returns not only are more convenient for taxpayers; they also reduce cheating. Greater compliance, in turn, is likely to make heavy taxation more workable. If cheating is extensive, tax rates need to be higher in order to raise a given quantity of revenue, which increases the likelihood of disincentive effects on entrepreneurship, investment, and work effort. In a tax system with minimal cheating, more revenue can be raised at moderate tax rates.
This can’t be done in the United States, so the argument goes, because the American tax code (unlike its Swedish counterpart) has too many available deductions and rebates. But the U.S. could simplify its tax code to enable pre-preparation. Moreover, even with this advantage, income tax rates in Denmark and Sweden are a good bit higher than in the U.S. And a large portion of Danish and Swedish tax revenues come via payroll and/or consumption taxes, which are less vulnerable to evasion, in those countries and in the U.S. as well.
Two explanations that attempt to address the puzzle
Here are two accounts of Danish and Swedish economic performance that don’t sidestep the question of tax levels’ impact.
The first is hypothetical; I don’t know of anyone who’s offered this argument explicitly. It says that the adverse effect of taxation kicks in once a country passes 15% or 20% or 25% of GDP, and it doesn’t worsen the farther beyond that you go. Denmark, Sweden, and the United States each exceeded 25% already by 1960, so in this story we would expect the three countries to have experienced similar (poor) economic performance in subsequent years.
This hypothesis doesn’t strike me as especially compelling. None of the world’s rich nontiny democracies have had tax levels below 25% of GDP since the 1970s, and only a few have been below that level since 1960. Yet a number of these countries have had relatively good economic outcomes during this period.
A second explanation says the Danish and Swedish economies have performed similarly to America’s despite heavier taxes because they have some advantage(s) that I haven’t adjusted for. This certainly would be true if I had chosen Norway as one of the comparison countries. Norway’s economy has been boosted by extensive oil resources. Has Denmark or Sweden had any such advantage?
One possibility is catch-up. Laggard countries can get an economic growth boost by borrowing technology from the leaders. But this has become less relevant for Denmark and Sweden in recent decades, as they’ve invested heavily in education and R&D and become technological leaders in their own right (more here).
Ethnic and cultural homogeneity is sometimes mentioned as a key economic asset of the Nordic countries. This might help, though in rich nations diversity may have some benefits as well.
Corporatist policy making, which features institutionalized participation by business and labor representatives, is associated with faster economic growth in affluent countries in recent decades. This may have helped Denmark and Sweden. Yet both countries have made their share of policy mistakes.
Of course, the United States has some important advantages of its own, including a huge domestic market, excellent universities, a culture that prizes innovation and entrepreneurship, a well-developed venture capital system, bankruptcy laws that facilitate risk-taking, a tradition of regional mobility, and an attractiveness to talented immigrants. The question is: If taxation at Danish and Swedish levels has a significant negative economic effect, do Denmark and Sweden have advantages relative to the U.S. that are large enough to have fully offset that effect in recent decades? It’s a difficult question to answer with any certainty, but I think probably not.
A challenge
At what point does the harmful impact of taxes on the economy kick in? And how large is it? The Danish and Swedish experiences over the past generation pose a challenge for those who believe the answers to these two questions are “somewhere below 50% of GDP” and “large.” It’s a challenge that in my view has yet to be met.
Working-age Belgians, French, and Germans spend, on average, about 1,000 hours a year in paid work. In the United States, Switzerland, and New Zealand, by contrast, the average is around 1,300. This is a pretty big difference.
These averages are determined by the share that have a paying job and the number of hours worked over the course of a year by those with a job. In the United States, for instance, the employment rate in 2007 was 72% and those employed worked an average of 1,800 hours (.72 x 1,800 = 1,296). In France, the employment rate was 64% and the average number of hours worked by those with a job was 1,550.
In a paper published in 2004, Edward Prescott concluded that taxes are the principal cause of the cross-country variation in working time. Prescott’s conclusion was based on the association between tax levels and work hours in the early 1970s and the mid-1990s in Canada, France, Germany, Italy, Japan, the United Kingdom, and the United States.
The hypothesis is sensible. Taxes reduce the (direct) financial reward to paid work. This encourages people not to work at all or to work fewer hours.
But how large is the effect? After all, some people will work more when taxes are higher, in order to reach their desired after-tax income. More important, lots of other things affect people’s calculations about whether and how much to work, including wage levels, employment and working time regulations, paid vacation time and holidays, availability and generosity of government income transfers, access to health insurance and retirement benefits, the cost of services such as child care, and preferences for work versus leisure. A good recent study of work hours among those who have a job concludes that taxes seem to have an effect for women but not for men, and that taxes account for a limited portion of the cross-country variation. In own my research (here and here), I’ve found pretty strong indication that the tax mix matters; heavy reliance on payroll taxes is associated with slower increase in the employment rate over the past three decades. But that doesn’t necessarily tell us anything about the impact of overall tax levels.
Here is the association between annual work hours per working-age person in 2007 (before the crash) and tax revenues as a share of GDP over the years 1979 to 2007. The pattern looks supportive of the notion that high taxes reduce work hours.
But knowledgeable comparativists will notice a familiar clustering of countries. Here’s the same chart with three groups highlighted.
One group, in the lower-right corner, includes Germany, Italy, the Netherlands, France, and Belgium. These countries, along with Austria, have several features that might contribute to low work hours. One is strong unions. Organized labor has been the principal force pushing for a shorter work week, more holiday and vacation time, and earlier retirement. These nations also have been characterized by a preference for traditional family roles: breadwinner husband, homemaker wife. This preference, often associated with Catholicism and “Christian Democratic” political parties, is likely to influence women’s employment and work hours. It is manifested in lengthy paid maternity leaves, lack of government support for child care, income tax structures that discourage second earners within households, and practices such as German school days ending at lunch time and French schools being closed on Wednesday afternoons. These countries also fund their social insurance programs via heavy payroll taxes, the kind most likely to discourage employment growth.
A second group consists of the four Nordic nations: Denmark, Sweden, Finland, and Norway. These countries too have strong unions. But they also have had electorally successful social democratic parties, which have tended to promote high employment. Denmark and Sweden, in particular, have been at the forefront in use of active labor market programs to help get young or displaced persons into jobs, public employment to fill gaps in the private labor market, and government support for child care and preschool to facilitate women’s employment.
A third group of countries, in the upper-left corner, includes the United States, Japan, Australia, New Zealand, and Canada. These nations have relatively weak labor movements and limited influence of social democratic parties and Catholic traditional-family orientations.
The other five countries — Ireland, Portugal, Spain, Switzerland, and the United Kingdom — are a hodgepodge. (Some would include Ireland and the U.K. in the “weak labor” group and Spain and Portugal in the “traditional family roles” group. Doing so doesn’t alter the conclusion here.)
Based on their institutional-political makeup, we would expect the weak-labor countries to have comparatively high work hours, the social democratic countries to be intermediate, and the traditional-family-roles countries to have low hours. As the following chart indicates, that’s exactly what we observe.
So is it really heavy taxation that produces comparatively low work hours? Or is it strong unions and preferences for traditional family roles? If we adjust for institutional-political group membership, the negative association between tax levels and work hours disappears.
Given that the institutional-political groupings account for much of the cross-country variation in levels of work, we might be better able to detect the true impact of taxes by examining changes. The following chart shows change in work hours from 1989 to 2007 by change in taxation from the 1980s to the 2000s. There is no association to speak of; the regression line is negatively sloped, but it is nearly flat and the countries are widely dispersed around it. Perhaps most revealing is the pattern among the twelve countries bunched around zero on the horizontal axis; despite little or no change in tax levels over this period, these nations varied sharply in the degree to which average work hours changed.
Is it levels of taxation, rather than changes, that cause changes in work hours? No; here too we find no association.
While heavy taxation surely creates some work disincentives, the overall tax level doesn’t seem to be an important determinant of differences in employment hours across the world’s rich countries.
A high level of government spending doesn’t necessarily produce heavy government debt. Nor does low spending guarantee low debt. Debt levels are a function of government expenditures and revenues and economic growth.
Kip Hagopian says no. He considers various arguments in favor of progressivity and isn’t persuaded. I appreciate Hagopian’s attempt to engage these arguments. Unfortunately, he says little or nothing about the three I find most compelling.
1. Luck. Many of the things that determine our incomes — intelligence, creativity, physical and social skills, motivation, persistence, confidence, connections, discrimination, occupation, employer, spouse, inherited wealth — are in significant measure a product of chance. They are heavily influenced by genes, our parents, our childhood neighborhood and schools, timing, and various fortuitous occurrences. Opponents of progressive taxation often emphasize the role of effort, but much of the variation in effort is itself a product of luck. (Progressive tax proponents sometimes fall into the trap of accepting the distinction between effort and luck; they’re then forced to argue that the latter matters more than the former.)
2. Ability to pay. Higher-income households tend to be able to pay not only more dollars but also a larger share of their income without suffering. One sign that this is true is that the savings rate increases with income; those with higher income tend to save a larger percentage. This may owe partly to a stronger future-orientation, but it’s mainly because they can afford to.
3. Income tax progressivity helps to offset the regressivity of other taxes. Some taxes are regressive, with higher-income households paying a smaller share of their income than lower-income households. Payroll (Social Security) and consumption (sales) taxes are the most prominent. If income taxes weren’t progressive, the tax system as a whole would be regressive.
Fairness is not the only criterion by which a tax system should be judged. We also need to consider how much revenue we want to raise and taxes’ impact on the economy. For my thoughts, see here and here.
Income inequality in America has soared over the past generation. But some see little cause for concern. One reason is that our inequality statistics — Gini coefficient, share of income going to the top 1%, and so on — are calculated based on households’ income in a single year. This misses the fact that people move up and down over time. Our incomes in any given year may be more dispersed now than several decades ago, but if many of us are switching places from year to year, why the fuss?
Two claims need to be distinguished here. One says there’s enough movement up and down in the income distribution over time (in technical lingo, relative intragenerational income mobility) that we needn’t worry about single-year inequality at all. It doesn’t matter whether inequality is high or low; it doesn’t matter whether it’s rising or falling. Single-year income inequality is simply irrelevant, on this view, because there is a lot of mobility. Since “a lot” and “enough” are in the eye of the beholder, evidence can’t confirm or refute this claim.
A second claim says that the rise in income inequality has been offset by a rise in mobility. Here we can look to the data for a verdict. Has income mobility increased?
For the bulk of the population — everyone but the richest — we have multiple sources of mobility data. One is the Panel Study of Income Dynamics (PSID); another is earnings records from the Social Security Administration. Both indicate that there has been no increase in income mobility in recent decades (see also here).
What about at the top? A good bit of the past generation’s rise in inequality consists of growing separation between the rich, especially the top 1%, and the rest of America. But has this been accompanied by increased churn among those at the top? In 2007 the Treasury Department released a study based on analysis of tax records. It included data on movement out of the top 1% over two nine-year periods: 1987-1996 and 1996-2005.
Single-year income inequality rose sharply during these two periods. The share of income going to the top 1% of households jumped from 11% in 1987 to 14% in 1996 to 18% in 2005. The Treasury study found that mobility, by contrast, was essentially unchanged.
The large increase in income inequality has not been offset by a rise in mobility at the top.
Paul Krugman rightly notes a potential problem in comparing the post-1973 trend in GDP with the trend in median income: the price indexes used to adjust for inflation differ. But that’s not an issue in this “decoupling” chart. It uses the same price index for both.
