Search the site
Recently, both Presidents Trump and Obama have publicly taken credit for the strong US economy. In this post, Christopher Knittel turns to data to evaluate their claims.
Recently, both Presidents Trump and Obama have publicly taken credit for the strong US economy. President Obama was quick to remind Americans that the economy began to grow well before the 2016 election when he said recently, “When you hear how great the economy’s doing right now, let’s just remember when this recovery started...When you hear about this economic miracle that’s been going on, when the job numbers come out, monthly job numbers, suddenly Republicans are saying it’s a miracle. I have to kind of remind them, actually, those job numbers are the same as they were in 2015 and 2016.” President Trump, however, is eager to point to the 2017 tax cuts and a trend toward less regulation as the source of the strong economy, saying “He was trying to take credit for this incredible thing that’s happening...It wasn’t him.”
When I first heard of each president’s claims, I found myself wondering who was right. As an academic economist that usually works with data to answer questions, I turned to data to be the objective arbiter in this disagreement.
I collected data on a variety of key economic indicators commonly used among economists and policy makers measuring both labour and business growth. The data for each of the economic indicators are collected from official governmental websites. The sources and a detailed description of the data are described at the end of this note. Specifically, I collected data on growth rates of:
All of the economic growth measures paint a similar picture: there is no data-driven evidence that the trends in economic growth under President Trump are any different from a continuation of the trends in economic growth established under President Obama. This, of course, is not a negative reflection on President Trump; the current administration has been able to sustain the growth trends that began under President Obama.
More specifically, the trends we observe under President Trump are within the 95% confidence interval of a continuation of the economic growth observed since the end of the Great Recession during President Obama’s tenure. In cases where a monthly or quarterly data point does fall outside of the confidence interval, roughly half of these reflect underperforming growth, while half reflect outperforming growth. Furthermore, a formal statistical test finds no evidence of a break in economic growth trends.
A weaker test is to ask whether a given data point under President Trump is outperforming or underperforming the trend established under President Obama, regardless of statistical significance. This is a weaker test because it ignores the simple statistical chance a measure of economic growth will exceed or be below the trend, but may be of interest to some. If we apply this metric, growth rates under President Trump outperformed the trend established under President Obama 48.5% of the time, and underperformed the trend 51.50% of the time.
Finally, I statistically test for a break in the trends of economic measure using the Wald test for structural breaks with a known break date. None of the variables exhibit statistical evidence of a structural break.
For each economic measure, I plot the data for that measure beginning at the official end of the Great Recession, as measured by the NBER (June 2009), to the most recent value available. I include a trendline depicting the pre-Trump administration trend in the economic measure. The trendline is the average growth rate for that economic measure observed during the Obama administration, again beginning from the end of the Great Recession to the end of President Obama’s second term. For ease of interpretation, I also include a vertical line indicating when President Trump took office. I also include the 95% confidence intervals for the trendline. The 95% confidence interval reflects the range of values for the growth measure that we might expect to observe if there is no change in economic growth trends. Therefore, if the economic growth measures under President Trump fall within this confidence interval, we can conclude that there is no statistical evidence of a change in economic trends.
Before turning to the evidence with respect to the claims made by the two presidents, let me provide a picture of what we would expect to see if the economy did undergo a structural break in growth rates. To do this, I make use of a recent event that we are certain changed economic activity, namely, the Great Recession. I analyse for both GDP growth rates and unemployment growth rates.
Figure 1 plots the growth rate of real GDP from the official start of the Great Recession (December 2007) to the most recent time period. GDP growth rates are measured at the quarterly level. For this illustrative exercise, I include a trendline for GDP growth rates depicting the trend only during the Great Recession (December 2007 to June 2009) as well as the 95% confidence interval for this trendline. Therefore, the trendline measures what we would have expected GDP growth to be if the Great Recession had continued. The vertical line is the official of the Great Recession.
The figure makes clear that there was a break in the trend of GDP growth rates after the end of the Great Recession. Indeed, the official end of the Great Recession corresponds perfectly with the first instance where the observed GDP growth rate is outside of the 95% confidence interval of the GSP growth rate prediction based on the time period during the Great Recession. Therefore, we can statistically say that GDP growth exceeded predicted growth rates.
Evidence that the economy structural changed after President Trump took office would be similar; GDP growth rates under the Trump administration would exceed predicted growth rates based on the period from the recovery to the end of the Obama administration.
Figure 1 Data-driven evidence of a structural shift in real GDP growth rates after recovery
Source: Database on economic outcomes of the Federal Reserve Bank of St. Louis.
Notes: The thick line represents actual economic outcomes. The linear trendline is fitted using the data during the Great Recession. This trendline is then extended after the recovery (delineated by the vertical red line). The confidence intervals represent the 95% confidence interval associated with the trendline prediction.
Figure 2 repeats this exercise for the growth rate in unemployment. Here, a growing economy implies a negative growth rate for unemployment. The trendline again is based on unemployment growth during the Great Recession. As expected, if the Great Recession had continued we would have expected unemployment growth to remain positive. In contrast, we observe that after the official end of the Great Recession unemployment growth rates were more negative than what would be expected had the Great Recession continued, so much so, they fell outside of the 95% confidence interval. Again, an objective measure of whether the Trump economy is growing faster than the Obama economy would imply a similar finding; unemployment growth is more negative that would be predicted if the Obama economy continued after January 2017.
Figure 2 Data-driven evidence of a structural shift in unemployment growth rates after recovery
Source: Database on economic outcomes of the Federal Reserve Bank of St. Louis.
Notes: The thick line represents actual economic outcomes. The linear trendline is fitted using the data during the Great Recession. This trendline is then extended after the recovery (delineated by the vertical red line). The confidence intervals represent the 95% confidence interval associated with the trendline prediction.
Figure 3 plots the growth rate of real GDP, the pre-Trump trendline, and the 95% confidence interval for the trendline. Growth rates under the Trump administration fall within the 95% confidence interval generated from the trendline. These data suggest that recent growth rates are in-line with what we would have expected if the Obama economy continued.
The preceding discussion compares GDP growth to the 95% confidence interval as predicted from the Obama-era growth rates. A weaker test is to compare growth rates during the Trump administration to the trendline, counting growth rates that exceed the trendline as evidence that the Trump economy is growing at a more rapid pace. This is a weaker test because it ignores the natural variability in economic growth, but I include it anyways. I calculate the percentage of quarters during the Trump administration that exceeded the trendline and the percentage of quarters that underperformed the trendline. Two-thirds, or 66.7%, of the GDP growth rates under President Trump exceeded the trendline constructed from the post-Great Recession growth rates under President Obama, while one-third of the growth rates were below the trendline.
Figure 3 Real GDP growth rates from June 2009 to June 2018
Source: Database on economic outcomes of the Federal Reserve Bank of St. Louis.
Notes: The thick line represents actual economic outcomes. The linear trendline is fitted using the data from the end of the Great Recession as determined by the National Bureau of Economic Research (June 2009) to the beginning of the Trump administration. This trendline is then extended during the Trump administration (delineated by the vertical red line). The confidence intervals represent the 95% confidence interval associated with the trendline prediction.
Figure 4 plots the analogous graph to the previous one for the growth/decline rate of unemployment. These data are available monthly. To be clear, negative trends indicate a growing economy as they indicate that the unemployment rate is shrinking. Only one of the observations under the Trump administration falls outside of the 95% confidence interval. This data point reflects a weaker economy in the sense that unemployment did not fall as quickly as the trendline.
If we apply the weaker test of where the growth/decline rate outperformed or underperformed the trendline itself, thus ignoring statistical significance, we find that 47.4% of the time the unemployment growth rate outperformed under President Trump and 52.6% of the time it underperformed.
Figure 4 Unemployment decline rates from June 2009 to August 2018
Source: Database on economic outcomes of the Federal Reserve Bank of St. Louis.
Notes: The thick line represents actual economic outcomes. The linear trendline is fitted using the data from the end of the Great Recession as determined by the National Bureau of Economic Research (June 2009) to the beginning of the Trump administration. This trendline is then extended during the Trump administration (delineated by the vertical red line). The confidence intervals represent the 95% confidence interval associated with the trendline prediction.
I repeat this same exercise for the economic measure listed above. They all paint a similar picture: there is no data-driven evidence that the trends in economic growth under President Trump are any different from the trends under President Obama. To be clear, this finding is not a negative reflection on President Trump; the current administration has been able to sustain the growth trends that began under President Obama. Just as there is no data-driven evidence that the economy under President Trump is growing at a faster rate than had the economy continued down the path forged under President Obama, there is no evidence that it is underperforming this path.
The following table summarises all of the figures in this post:
Table 1 Comparison of growth rates to Obama administration trends
Notes: For each economic measure, ‘Percent above 95% CI” is the percentage of data points during the Trump administration that are above the 95% confidence interval of the Obama-era trendline. ‘Percent above 95% CI’ is the percentage of data points during the Trump administration that are below the 95-percent confidence interval of the Obama-era trendline. ‘Percent above Trendline’ and ‘Percent below trendline’ correspond to the weaker test of whether a Trump-era data point is simply above or below the trendline, respectively. This these ignore whether a given data point is statisticallydifferent from the trendline. It is important to note that whether the data point is above the trendline reflects outperformance varies across variables. Being above the trendline signifies outperformance for: GDP, Labour Participation, Real Wages, and the DJIA. Being above the trendline signifies underperformance for: Unemployment, Debt-to-GDP, and the CPI. P-value of Chi-Square Test is the p-value from a statistical test of whether there was a trend break in the economic measure after President Trump took office. The null hypothesis of this statistical test is that there is no trend break. Therefore p-values above 0.05 suggest that there is no statistical evidence of a trend break. This is implemented using the estat sbknowncommand in Stata 15.
Besides the unemployment growth rate discussed above, there are only three other measures of economic growth where the observed growth rate under President Trump falls outside of the 95% confidence interval. These measures are the unemployment growth rate among African Americans, the growth rate for labour participation, and the growth rate in the CPI. African American unemployment growth (again, decline) underperformed and outperformed the trendline 10.5% of the time. The growth in labour participation underperformed the trendline 5% of the time (and never outperformed), while CPI growth exceeded the trendline 5.6% of the time.
For the weaker test that ignores statistical significance, whether the economy outperformed or underperformed the trendline varies considerably across the measures. Across all measures, the data find that growth rates under President Trump outperformed the trendline 48.5% of the time and underperformed their respective trendlines 51.5% of the time.1
Finally, I statistically test for a break in the trends of economic measure using the Wald test for structural breaks with a known break date.2 The ‘P-value of Chi-square test’ column is the p-value from a statistical test of whether there was a trend break in the economic measure after President Trump took office. The null hypothesis of this statistical test is that there is no trend break. Therefore p-values above 0.05 suggest that there is no statistical evidence of a trend break. Intuitively, this is a direct test of whether the trendline changed under President Trump (in either direction). The null hypothesis of this test is that there is no break. Therefore, a small p-value would be statistical evidence that this null hypothesis can be rejected. The typical critical value for the p-value used in economics is 0.05. Therefore, any p-value above this would suggest that there is no statistical evidence of a break in trends.
Each of the p-values is well above 0.05 with the exception of the debt-to-GDP growth rate suggesting no evidence of a break in the trend. Here, the p-value is 0.11. I note that if one were to take this p-value as evidence of a structural break, it would imply a higher growth in debt-to-GDP under President Trump.
Figure 5 Unemployment decline rates for African Americans from June 2009 to August 2018
Source: Database on economic outcomes of the Federal Reserve Bank of St. Louis.
Notes: The thick line represents actual economic outcomes. The linear trendline is fitted using the data from the end of the Great Recession as determined by the National Bureau of Economic Research (June 2009) to the beginning of the Trump administration. This trendline is then extended during the Trump administration (delineated by the vertical red line). The confidence intervals represent the 95% confidence interval associated with the trendline prediction.
Figure 6 Labour participation growth rates from June 2009 to August 2018
Source: Database on economic outcomes of the Federal Reserve Bank of St. Louis.
Notes: The thick line represents actual economic outcomes. The linear trendline is fitted using the data from the end of the Great Recession as determined by the National Bureau of Economic Research (June 2009) to the beginning of the Trump administration. This trendline is then extended during the Trump administration (delineated by the vertical red line). The confidence intervals represent the 95% confidence interval associated with the trendline prediction.
Figure 7 Labour participation growth rates for African Americans from June 2009 to August 2018
Source: Database on economic outcomes of the Federal Reserve Bank of St. Louis.
Notes: The thick line represents actual economic outcomes. The linear trendline is fitted using the data from the end of the Great Recession as determined by the National Bureau of Economic Research (June 2009) to the beginning of the Trump administration. This trendline is then extended during the Trump administration (delineated by the vertical red line). The confidence intervals represent the 95% confidence interval associated with the trendline prediction.
Figure 8 Average real wage growth rates from June 2009 to August 2018
Source: Database on economic outcomes of the Federal Reserve Bank of St. Louis.
Notes: The thick line represents actual economic outcomes. The linear trendline is fitted using the data from the end of the Great Recession as determined by the National Bureau of Economic Research (June 2009) to the beginning of the Trump administration. This trendline is then extended during the Trump administration (delineated by the vertical red line). The confidence intervals represent the 95% confidence interval associated with the trendline prediction.
Figure 9 Debt-to-GDP growth rates from June 2009 to June 2018
Source: Database on economic outcomes of the Federal Reserve Bank of St. Louis.
Notes: The thick line represents actual economic outcomes. The linear trendline is fitted using the data from the end of the Great Recession as determined by the National Bureau of Economic Research (June 2009) to the beginning of the Trump administration. This trendline is then extended during the Trump administration (delineated by the vertical red line). The confidence intervals represent the 95% confidence interval associated with the trendline prediction.
Figure 10 Consumer Price Index growth rates from June 2009 to August 2018
Source: Database on economic outcomes of the Federal Reserve Bank of St. Louis.
Notes: The thick line represents actual economic outcomes. The linear trendline is fitted using the data from the end of the Great Recession as determined by the National Bureau of Economic Research (June 2009) to the beginning of the Trump administration. This trendline is then extended during the Trump administration (delineated by the vertical red line). The confidence intervals represent the 95% confidence interval associated with the trendline prediction.
Figure 11 Dow Jones Industrial Average growth rates from June 2009 to August 2018
Source: Database on economic outcomes of the Federal Reserve Bank of St. Louis.
Notes: The thick line represents actual economic outcomes. The linear trendline is fitted using the data from the end of the Great Recession as determined by the National Bureau of Economic Research (June 2009) to the beginning of the Trump administration. This trendline is then extended during the Trump administration (delineated by the vertical red line). The confidence intervals represent the 95% confidence interval associated with the trendline prediction.
[1] It is important to note that whether the data point is above the trendline reflects outperformance varies across variables. Being above the trendline signifies outperformance for: GDP, Labour Participation, Real Wages, and the DJIA. Being above the trendline signifies underperformance for: Unemployment, Debt-to-GDP, and the CPI.
[2] Implemented using the estat sbknowncommand in Stata 15.
