Macro Roundup

Research attempts to gauge the impact of competitiveness as a character trait, paper finds it is predictive of educational and labor market outcomes though they note the relationships"... are (conditional) associations and cannot be interpreted causally..."

Their takeaway, “…In this paper we introduce such measures of competitiveness. The first is incentivized and is an online adaptation of the original Niederle-Vesterlund measure. The second is a single and general unincentivized survey question which can be quickly elicited in any setting. The two measures arestrongly correlated at the individual level and both measures are strong and consistent predictors of completed level of education, field of study in college, occupation and income. The predictive power of the unincentivized measure for these outcomes is robust to controlling for other traits, including risk seeking, confidence and the big five personality traits. For most outcomes, the predictive power of competitiveness exceeds that of the other traitsWe also contribute to the literature on gender differences in the labor market. We find that gender differences in competitiveness can explain 5-10% of the observed gender differences in education and labor market outcomes and that gender differences in competitiveness, risk seeking, confidence and the big five personality traits combined can explain up to 25% of gender differences. Our results are also relevant to the ongoing discussion in the experimental economics literature on whether gender differences in willingness to compete in the lab are due to individual differences in preferences for competition or rather due to differences in beliefs and risk preferences. We show that competitiveness and risk attitudes predict a separate set of outcomes. Risk attitudes are only weakly related to career outcomes. On the other hand, risk attitudes, but not competitiveness, predict choices that are intrinsically risky but not competitive (investing in stock, self-employment and smoking..”

Thomas Buser, Muriel Niederle and Hessel Oosterbeek, "Can Competitiveness predict Education and Labor Market Outcomes? Evidence from Incentivized Choice and Survey Measures," National Bureau Of Economic Research, June 2021, https://www.nber.org/papers/w28916

Gender Differences“…While 34 percent of working-age men choose to compete in the experiment, only 22 percent of working-age women do so (p=0.000; chi-squared test).24 65 percent of men and 47 of women rate their competitiveness as at or above the median (p=0.000).25 In the right-most panel, we show a histogram of competitiveness levels by gender which reveals large gender differences in the tails. 11 percent of men but only 5 percent of women rate themselves as extremely competitive (9 or 10 on a 0-10 scale).On the other hand, 9 percent of women but only 5 percent of men rate their competitiveness as very low(0-2)...In Figure 9, we ask whether our unincentivized measure of competitiveness, as well as the other trait measures, can explain a significant part of these gender gaps. To do so, we estimate the percentage reduction in the gender gap upon controlling for each of the traits separately and jointly.26 Controlling for the unincentivized competitiveness measure reduces the gender gap in income by 4.3 percent. Competitiveness is among the traits which explain most of the gender gap in income, together with stability (5.3 percent) and general confidence (3.6 percent). In the rightmost graph in each panel, we estimate the proportion of the gender gap that is explained by all traits together. We estimate that around 12 percent of the gender gap in income in our sample can be explained by the psychological traits we consider. Following previous results that the potential of competitiveness for explaining differences in educational career choices is highest at the top of the ability distribution (Buser, Peter, and Wolter, 2018), we repeat this exercise for collegeeducated individuals. Here, competitiveness explains 7.6 percent of the gender gap, while stability explains 9.9 percent and confidence 6.2 percent. All traits jointly explain roughly 18 percent of the gender difference in income for college-educated individuals. Competitiveness also has the most explanatory power for the gender gap in holding a high-level professional position conditional on education level (8.7 percent) followed by stability (7.3 percent), risk seeking (6.3 percent), confidence (5.9 percent), and intellectual openness (5.8 percent). All psychological traits jointly explain around 25 percent of the gender gap. For those who graduated from college, competitiveness (6.1 percent), agreeableness (7.5 percent) and risk seeking (3.4 percent) are the only significant explanatory factors for the gender gap in major choice. All traits jointly explain 18 percent of the gender gap..."

Impact on risk taking, “…. Our results indicate that competitiveness, risk and confidence are three separate traits which predict outcomes independently or even predict different outcomes. For example, Figure 5 shows that our measure of competitiveness significantly predicts education and labor market variables even when controlling for risk preferences, confidence, self-esteem and personality. Neither the effects of risk preferences nor of confidence are consistently robust to controlling for the other traits…. In Figure 7 shows the difference in outcomes between individuals in the top 30% and individuals in the bottom 30% on each trait, controlling for age and gender. We replicate the Dohmen et al. (2011) result that risk seeking is a significant predictor of all four outcomes. As expected, competitiveness does not predict investing in stocks, being self-employed or smoking. On the other hand, competitiveness is the strongest predictor of practicing sports. This analysis further underlines that competitiveness and risk attitudes are both relevant individual traits which predict different economically important outcomes and should therefore be accounted for separately…”

Incentivized measure, “…Figure 3 shows how respondents who chose the tournament option and respondents who chose the piece-rate option are distributed across different levels of each of the four education and labor market variables. In the two upper-most panels, we divide the sample into income quintiles. Because the income distributions of men and women are very different, we do this separately by gender. For each gender, the income distributions of individuals who chose to compete and individuals who chose not to compete differ strongly. Individuals who compete are more than twice as likely to be in the highest quintile and much less likely to be in lowest two quintiles…”

The evidence:“…Our results confirm that both the incentivized and the unincentivized measures are strong predictors of the same education and labor market outcomes. Conditional on performance in the task, gender, age, education level, risk preferences, confidence and the big five personality traits, individuals who compete in the experiment earn roughly 13 percent more, and are around 50 percent more likely to hold a high-level managerial or professional position. Individuals who compete are also 13 percentage points more likely to go to college and, conditional on going to college, choose more prestigious study majors that lead to higher salaries. We find the same effects for people who score themselves higher on competitiveness in our survey question. Comparing the predictive power of different traits, competitiveness is consistently among the strongest predictors. The fact that the incentivized and unincentivized measures predict the same outcomes justifies the use of the unincentivized survey question as a valid measure of competitiveness. This makes it possible to elicit competitiveness in large-scale representative national or cross-country surveys and survey panels in the future….”

Core results: First, “…willingness to compete as measure by an incentivized choice experimentstrongly predicts earnings and occupation conditional on education level, as well as education level and college major choice in a representative sample….”Second, “… unincentivized survey question predicts the same outcomes in the same sample with comparative predictive power….”Third, “…our survey measure of competitiveness predicts these outcomes over and above other unincentivized trait measures that the education economics literature has typically focused on, including risk seeking, confidence and the big 5 personality traits, indicating that competitiveness is a separate influential trait that is not captured by these other, more traditionally used traits…”

Ruy Teixeira, "Ten Things We Now Know About the White Working Class Vote in 2020," The Liberal Patriot, June 24, 2021, https://theliberalpatriot.substack.com/p/ten-things-we-now-know-about-the-1a6

Ten Things We Now Know About the White Working Class Vote in 2020

Trump’s not-so-secret weapon going into the 2020 election was the white working class (noncollege) vote. That weapon didn’t work or didn’t work well enough to save him. The release of Catalist and Census data, as well as other data sources, now allow us to sketch a portrait of that demographic in the 2020 election and how their voting patterns fell short of what Trump needed.

So, here are ten things we now know about white working class voters in the 2020 election (margin shift and support figures all based on the two-party vote from Catalist):

1. White working class voters did indeed shift against Trump in 2020 relative to 2016, albeit not as much as pre-election polls suggested would happen. That 3 point shift against Trump was exactly what Trump didn’t need; what he needed was a 3 point shift toward him to replicate his 2020 success. It’s a popular, if unenlightening, exercise to claim that such-and-such a demographic group “won” the election for Biden, given the small vote margins in a handful of states. I won’t do that here but it’s fair to say that the white working class vote was “the dog that didn’t bark” in 2020. Trump needed more of their support, not less, in 2020 and he just didn’t get it.

2. Trump lost white working class support across most states, including key swing states. This includes swing states in both the Rustbelt and the Sunbelt.

3. That said, Trump still carried white noncollege voters nationally by a wide margin—26 points according to Catalist. Other data sources have Trump’s margin even higher. With that kind of margin, the GOP will continue to dream on the political bonus from reversing the 2020 trend and shifting the group further in their direction.

4. This is underscored by the raw fact of this group’s continued large numbers. Despite a roughly 2 point drop in their share of voters from 2016, they are still 44 percent of all voters in the Catalist data (the Census data have that share at a still very large 40 percent). Moreover, the white working class share of the vote in swing Rustbelt states is much larger than the national figure (by 7-19 points).

5. White noncollege turnout performance in 2020 was quite good. Turnout among these voters went up 6 points, according to Census data, a larger turnout increase than among white college and black voters and about the same as Latinos.

6. The pattern of strong white working class turnout increases can also be seen across states, including swing states. Only six states saw declining turnout among white noncollege voters, while 15 states showed declining white college turnout. In all but 11 states, white working class turnout increases were larger than white college increases.

7. Even with their declining population share and Trump’s dominance of their group, white noncollege voters still made up a larger share of Biden’s coalition (32 percent) than white college voters (29 percent), according to Catalist.

8. The shift against Trump by white noncollege voters was strong in the suburbs where most of these voters actually live (contrary to popular perceptions). White noncollege voters are almost three-fifths of white voters in suburban areas and they shifted against Trump by over 5 points.

9. Even white working class rural voters shifted against Trump, albeit by a modest 2 points. Still, that was Trump’s wheelhouse.

10. The shift against Trump among white noncollege voters was almost entirely driven by men. They shifted 7 points against him, while women in this group barely moved at all (.2 points). This is a not a widely-appreciated fact, to say the least.

The Democrats will hope that dog doesn’t bark—in fact, wags its tail--in 2022 and 2024. That is a challenge the party needs to keep its eye on especially since nonwhite working class voters no longer seem like such a lock.

Ulrike Malmendier and Stefan Nagel look at inflationary expectations across generations and find expectations are anchored by experience, “…Our empirical analysis shows that individuals’ inflation expectations differ depending on the inflation process experienced during their life times. Differences in the experienced mean inflation rate and the persistence of inflation shocks generate (time-varying) differences in inflation expectations between cohorts. The experience of younger individuals is dominated by recent observations, while older individuals draw on a more extended historical data set in forming their expectations.Such learning from experience can explain, for example, why young individuals forecasted much higher inflation than older individuals following the high inflation years of the late 1970s and early 1980s. Both the mean rate of inflation and inflation persistence were particularly high in the short data set experienced by young individuals at the time. Learning-from experience complements the sticky information explanation of expectations heterogeneity put forward in Mankiw and Reis (2002) and Carroll (2003) for the same time period. For the more recent time period towards the end of our sample in 2010, our estimates imply that the perception of the persistence of inflation shocks is close to zero, particularly for young individuals. This suggests that unexpected movements in the inflation rate are currently unlikely to move inflation expectations much. As argued in Roberts (1997), Orphanides and Williams (2005a), and Milani (2007), these changes in individuals’ perceptions of persistence are likely to influence, in turn, the persistence of inflation rates….”

The basic dynamic, “…When forming macroeconomic expectations, individuals put a higher weight on realizations of macroeconomic data experienced during their life-times compared with other available historical data. As a result, learning dynamics are perpetual. Beliefs keep fluctuating and do not converge in the long-run, as weights on historical data diminish when old generations disappear and new generations emerge. Such learning from experience carries two central implications. First, expectations are history-dependent. Cohorts that have lived through periods of high inflation for a substantial amount of time have higher inflation expectations than individuals who have mostly experienced low inflation. Second, beliefs are heterogeneous. Young individuals place more weigh on recent data than older individuals since recent experiences make up a larger part of their life-times so far. As a result, different generations tend to disagree about the future. Both effects have been noticed in practice. During the high-inflation period of the late 1970s, the Chairman of the Federal Reserve Paul Volcker remarked:“An entire generation of young adults has grown up since the mid-1960’s knowing only inflation, indeed an inflation that has seemed to accelerate inexorably. In the circumstances, it is hardly surprising that many citizens have begun to wonder whether it is realistic to anticipate a return to general price stability, and have begun to change their behavior accordingly.”Both effects are also visible in Figure 1: Following years of high inflation, young people expect much higher inflation going forward than older people…”

Ulrike Malmendier and Stefan Nagel, "Learning from Inflation Experiences," University Of California, 2013, https://eml.berkeley.edu/~ulrike/Papers/InflExp_44.pdf

According to the draft bipartisan infrastructure deal (as of June 24) < than 19% of new spending will go to the category described as “Roads, bridges, major projects.”..." (=110/579)

“Bipartisan Infrastructure New Spending,” Office of Senator Sinema, June 24, 2021, https://www.sinema.senate.gov/sites/default/files/2021-06/bipartisan%20infrastructure_Investment_Finance.WH_.06-24-21.Final__0.pdf

Subway costs per km, in millions of USD

Eric Goldwyn, Alon Levy, Elif Ensari, Marco Chitti, Abdirashid Dahir, Rachel Daniell, Yinan Yao and Anan Maalouf, "Transit Costs Project," New York University Marron Institute, Accessed June 28, 2021, https://transitcosts.com/projects/

Alon Levy on American infrastructure spending versus international comps, “…The United State has a severe construction cost problem for infrastructure. Together with Eric Goldwyn and Elif Ensari, with the assistance of Marco Chitti, Abdirashid Dahir, Yinan Yao, and Anan Maalouf, I have constructed a database of subway projects around the world with their costs.3 Subways are amenable to analysis from a 30,000’ altitude, since they are such big projects that they are amply covered in mass media and trade media, and costs across countries tend to be comparable. Out of several hundred subway lines in our database, the five that are in New York are the five most expensive — all well above $1 billion per kilometer where the global median is about $250 million;other American subways range between $400 and $800 million per kilometer. This is not about our wealth: there is no correlation between a country’s GDP per capita and its subway construction costs. Nor is it about geological factors: the biggest factor behind a project’s cost is what country it is in, and costs are fairly consistent even across different geologies. Some specific cases have unusually difficult local geography, like Shanghai and Amsterdam’s alluvial soil, but the American cases are not among those. This is purely institutional. Worse, this is not just a problem for public transportation. To the extent we have any data for road tunnels, the same problem is seen there, too: the Alaskan Way Viaduct Replacement in Seattle cost around $1 billion per kilometer, for example, while the A86 beltway tunnel in Paris and the M-30 tunnel in Madrid cost about $300 and $130 million per kilometer. respectively….”

Alon Levy, “So You Want To Do An Infrastructure Package,” Niskanen Center, March 2021, https://www.niskanencenter.org/wp-content/uploads/2021/03/levy-infastructure.pdf

Hassett/Mathur concluded that a 1% rise in the corporate-tax rate is associated with a 0.5% drop in wages which implies that more than 100% of the burden of corporate tax lands on workers.

Core of paper, "...To summarize, our results indicate that corporate taxes are significantly related to wage rates across countries. Our coefficient estimates suggest that a 1 percent increase in corporate tax rates leads to a 0.5 percent decrease in wage rates. These results also hold for effective marginal and average tax rates. The coefficient estimate is (on average) close to 0.5. This suggests that wages are as likely to be influenced by the top statutory corporate tax rate, as by the effective marginal and average tax rates. Hence corporate tax cuts in the form of large allowances for depreciation of equipment and structures which reduce effective marginal rates could effectively influence wage levels as well. We find evidence that international corporate tax rates affect domestic wages. Capital formation and therefore wage rates are affected not only by domestic tax rates, but also tax rates in competing economies. The coefficient estimates for the spatial tax variables range from 0.6 to 1.4 suggesting significant quantitative impacts. Comparing different weighting schemes, the effects are largest when “neighbors” are defined as countries within the same income group, rather than within the same region. This suggests that tax competition is most intense among, say, high income countries such as Canada, France and Italy, rather than between geographic neighbors. This makes sense intuitively since there do not appear to be large transport costs associated with moving capital across large distances, so capital can easily flow to the most remunerative locations….”
Evidence, "…Table 1 provides summary statistics for our data. The average dollar wage over our sample period for all countries was approximately $5.2 per hour, while the average headline corporate tax rate was 34 percent. The effective average rate was lower at 30 percent, and the effective marginal rate was nearly 10 percentage points lower at 26.5 percent. In general average wages have increased between 1981 and 2005 from $3.5 to $9. per hour, while average tax rates have declined substantially. In 1981, the average top rate was 42 percent. In 2005, the average rate was 25 percent....Table 5 presents various tests of our hypothesis. In specification (1), we use a five year (log) average of the capital-labor ratio as the dependent variable. We find that all measures of corporate taxation, such as the top national corporate tax rate, the effective average and the effective marginal tax rate negatively affect capital formation, though only the first two measures show up as significant. The coefficient on (Log) Top Corporate Tax Rate implies a value of the elasticity of close to -0.14. Clearly, higher top rates discourage capital formation. This result is even stronger for effective average tax rates which take into account depreciation allowances, inflation and interest rates and other factors that affect capital formation through the user cost of capital (specification 2). The estimated elasticity in this case is close to (negative) 0.16. Other studies, using micro data and the actual user cost (not only the tax rate) estimate elasticities that are higher than this. Balistreri, McDaniel and Wong (2002) using industry data from the Bureau of Economic Analysis estimate elasticities in the range of 1-1.22, using different weighting schemes. Leung and Yuen (2005) using industry-level data on Canadian manufacturing estimate an elasticity of 0.33. While our coefficient estimate is likely to be heavily biased due to aggregation, measurement issues and data constraints (the capital-labor ratio is not specific to the manufacturing sector), we present these results simply to show that different measures of corporate taxation can significantly and negatively affect capital-labor ratios. In column (4), we test to see if spatial tax rates have any effect on domestic capital labor ratios. Our results indicate that both domestic and neighbor country tax rates are important in explaining the formation of domestic capital labor ratios. Higher tax rates in neighboring countries have a positive and significant effect on capital formation in the domestic country, which supports our hypothesis that capital flows out of high tax jurisdictions and to low tax jurisdiction. The final column, Column (5), studies the link between high capital-labor ratios and average wages. Since capital-labor ratios are a direct proxy for worker productivity, it is not surprising that higher capital-labor ratios are associated with significantly higher wages. In general, a 1 percent increase in the capital-labor ratios is associated w..ith a 0.45 percent increase in wages….”

Kevin Hassett and Aparna Mathur, "A spatial model of corporate tax incidence," Applied Economics, 2015, https://www.aei.org/research-products/working-paper/a-spatial-model-of-corporate-tax-incidence/

Kimberly Clausing argues there isn’t evidence to support the idea that the corporate tax falls on workers, her takeaway, ".... workers have thus far remained insulated from their countries’ corporate tax policies. While prior work had identified burdens on workers from high corporate tax rates, a careful literature review reveals that this work suffers from essential drawbacks. Together with new evidence from the present analysis and also in Clausing (2012), one is left with a discrepancy between the theoretical expectation that workers will bear a large share of the corporate tax burden in a global economy and the empirical reality that there is very little robust evidence linking corporate tax rates and wages. There are several potential solutions to this puzzle. One can conclude that the data are simply too coarse to robustly pick up these theoretical mechanisms, or one can conclude that the theory itself is not an adequate depiction of reality and misses features of corporate taxation and competition that are important for understanding corporate tax incidence. It is also important to note two ways in which globalization itself may undermine the open-economy general equilibrium tax incidence result. First, if corporations are mere intermediaries in global capital markets in which a wide assortment of investors with different tax treatments invest, tax policy changes could affect the ownership and financing patterns of assets more than they affect the aggregate level of investment in different countries. Second, since multinational firms have become increasingly adept at separating the reporting of income from the true location of the underlying economic activities, international tax avoidance itself comes with a silver lining. Mobile firms move profits without needing to substantially alter the underlying investments, whereas immobile firms do not respond like the open-economy actors of modern corporate tax incidence models. In both cases, workers in high-tax countries are relatively insulated from adverse wage effects due to capital reallocation toward low-tax countries. The insights of this paper generate several policy-relevant considerations. First, there is already a strong case for rate-lowering, base-broadening corporate tax reforms. Any possible adverse effects on workers from high corporate tax rates just make that case stronger. Second, international tax reforms need to pay particular attention to tradeoffs that come from competing policy goals. Territorial systems of taxation would both heighten tax sensitivity of real investments to tax differences among countries and also (likely) widen the escape valve of income shifting out of high-tax country jurisdictions; a truly “tough” territorial system would predominantly have the former effect. On the other hand, limits on deferral, when combined with a lower rate, reduce incentives to move both real investments and income abroad. Finally, a formulary apportionment approach, while curbing the tax sensitivity of income reporting, would need to be designed carefully in order to avoid heightening real responses to tax differences among countries, since real responses generate greater concern about adverse effects on labor..."

Kimberly Clausing, "Who Pays The Corporate Tax In A Global Economy,"National Tax Journal, March 2013, https://www.ntanet.org/NTJ/66/1/ntj-v66n01p151-84-who-pays-corporate-tax.pdf

How labor avoids tax, “…The first possibility to recognize is that the open-economy general equilibrium corporate tax mechanism that lies at the heart of the prediction that labor will bear a large share of the corporate tax burden is difficult to identify with real world data. We are limited to a universe of about 30 comparably affluent countries with about 30 years of data of sufficient quality and comparability. Even with the large variation in corporate tax policies by countries over this time period, there remain many confounding influences on the data. It is indeed possible that corporate taxes have negative effects on wages that our data are simply too coarse to detect…As noted above, the labor share would be lower if trade substitution elasticities were lower, or if other parameters are taken at their median values, according to Gravelle (2013). But even beyond these modifications, the Harberger-style model neglects several features of real-world corporate taxation that are likely to have consequences for incidence. First, as Gravelle and Hungerford (2011) note, since the current corporate tax has residence elements, that would cause it to fall more heavily on capital than the above models imply. Second, if the corporate tax in fact subsidizes debt-financed investments (due to interest payment deductibility and accelerated depreciation), then raising the corporate tax could actually cause capital inflows of debt-financed investments. This would also reduce negative impacts on workers.. Considering the corporate tax as a tax on pure economic profit, or rents, alters the efficiency implications of the tax, as the tax has smaller effects on factor use choices if it falls primarily on pure profits rather than on corporate capital. This consideration also affects the policy implications of the tax. For example, if the rent-sharing mechanism is key, then cuts to the corporate tax will allow more rents to fall into the hands of shareholders and workers, who will share the excess returns. However, since labor markets are integrated across firms and industries, workers will eventually move from job to job and industry-to-industry, eroding wage differences. Thus, the economy-wide wage effects from such rent-sharing may be smaller than the rent-sharing specifications alone would lead one to believe…One possibility is that clientele effects may be important. Desai and Dharmapala (2009) find evidence of substitution between foreign portfolio investment and foreign direct investment in response to tax incentives. Also, as noted above, corporate taxation may actually subsidize debt-financed investments…. This divergence could reduce the wage effects of relative corporate tax rates,since internationally agile firms can move income without commensurate movements of investment and jobs. Indeed, many of the most global companies have become increasingly adept at the creation of stateless income, as discussed in Kleinbard (2011). If firms can respond to tax differences among countries through financial or organizational decisions, this will lower the tax sensitivity of real activity, thus reducing adverse effects on labor associated from tax-induced reductions in the capital stock…”

More evidence, “…Table 5 shows results for the same VAR analysis, using relative tax rates instead of level tax rates. Again, these relative tax variables were calculated as the country tax variable minus the OECD average for the same tax variable. Results were little changed; in all but one case the tax variables were jointly statistically insignificant. In the statistically significant case, the impulse response function shows that the six-year effect of taxes on wages was small and negative. Given the known sensitivity of VAR analysis to the number of lags and variables included, I also experimented with other specifications. For example, I considered a model with ten lags for both level and relative tax rates. Table 6 shows the results for the relative tax term specifications; the level tax rate results were nearly identical. Of the four VAR systems, in three cases there was no evidence of a jointly statistically significant relationship between lagged values of the corporate tax variables and wages. In the other case, the impulse response function indicates an approximately zero net relationship…”

Evidence, “…Figures 3 and 4 show scatter plots examining simple correlations between the BLS data hourly wage growth over three decades (the 1980s, 1990s, and 2000s) and the corresponding relative corporate income tax rates. Figure 3 shows the average relative statutory tax rate of each country in comparison to the rest of the countries in the sample over the same decade. Figure 4 is identical except it considers decade-long averages of the relative effective tax rate. In both cases, the figures do not show clear empirical relationships between wage growth and average relative tax rates over the three most recent decades. Figures 5 and 6 show the same types of simple scatter plots; the only difference is that the use of the OECD annual wage data, which constrains the sample to two decade averages, the 1990s and the 2000s. While there seems to be some evidence of a negative relationship between annual wage growth over these two decades and decade-average relative statutory tax rates in Figure 5, that relationship is less apparent in Figure 6, which shows relative effective tax rates. The correlations between the variables in Figures 3, 4, 5, and 6 are -0.07, +0.07, -0.47, and -0.09. The largest coefficient implies that about 22 percent of the variation in wage growth is explained by variation in relative tax rate; the other three coefficients imply that the variables are not well correlated…”