Skip to main content
The FRED® Blog

Shopping lines: The evolution of retail in the U.S.

When it comes to shopping, Americans have many options: the corner store, the supermarket, the specialty store, the “big box,” online, and more. Above, we committed a graphing sin by displaying 12 different series in a single graph to show how retail trade has evolved across various categories.

Signs can be seen over the past two and a half decades: First, while food and beverage stores (supermarkets, convenience stores) were major destinations in the past, they have been joined by general merchandise stores, typically large suburban big box stores. A very volatile bunch are the gasoline stations, which sell mostly one commodity with a very variable price. Finally, one category seems to be steadily overtaking the others: mail-order and online shops (the line in black).

How this graph was created: Search for the Advance Monthly Sales for Retail and Food Services release and select the series you want to display. The release offers more series than the 12 we chose, but 12 is the maximum that can be shown on one FRED graph. In fact, with so many series, you need to make the graph larger by dragging the red marker in the bottom right corner of the graph. We chose the seasonally adjusted series and made the nonstore retail series black.

Suggested by Christian Zimmermann

View on FRED, series used in this post: RSBMGESD, RSCCAS, RSDBS, RSEAS, RSFHFS, RSFSDP, RSGASS, RSGMS, RSHPCS, RSMSR, RSNSR, RSSGHBMS

The Canadian dollar and the price of oil

Canada’s oil sector amounts to about 10% of its GDP and 25% of its exports, almost all of which go to the U.S. It’s not too surprising, then, that the U.S./Canada exchange rate mirrors the price of oil. Of course, trade between the countries is much more than oil, but many of Canada’s other commodity exports have a price that is well correlated with the price of oil. And the financial linkages between the countries are also disproportionately tied to the mining and extractive industries.

That said, this relationship hasn’t always existed. See the graph: If you expand the time sample to more than the 10 years shown above, the correlation becomes gradually less clear. But the reason is clear: Canada has continuously expanded its oil production, and oil simply did not matter that much a few decades ago when it was not nearly the dominant revenue source it is today.

How this graph was created: Look for the Canadian/U.S. dollar foreign exchange rate and select the monthly series. Then use the Add a Series option to search for and select “WTI” (again, the monthly series). Modify this second series as follows: Switch the y-axis to the right side and create you own data transformation with formula 1/a. Finally, restrict the graph’s sample period to the past 10 years.

Suggested by Christian Zimmermann and inspired by a tweet from Paul Storer, who recently passed away

View on FRED, series used in this post: EXCAUS, MCOILWTICO

Have earnings kept up with growth?

Recent policy discussions have focused on wage growth and whether it’s been “too sluggish.” In this post we argue this is a feature of the trend and not the cycle.

Before looking at the actual statistics, it’s worth noting that it’s not totally obvious how we should define wages—because wage dynamics change so much over the distribution. Low, medium, and high wages have grown at different rates and at different times. From a macroeconomic perspective, however, it makes some sense to measure the average wage. The effect of so doing is that we put more weight on the higher earners than the average person, a result of a positively skewed wage distribution. (Recall the definition of skewness: Here, it means the top tail can pull up the mean past the average person’s wage, the median wage.) Still, with macroeconomic aggregates, using the average makes sense. We often look at GDP per capita, and average wages are equivalent to wages per capita. Now that we’ve decided what moment of the distribution to study, we have to choose what constitutes wages: total compensation including benefits or strictly labor income. Here, we focus on strictly wages and salaries rather than on other benefits: Labor income is more directly linked to economic motivations, whereas other side benefits are often the result of tax distortions. There are two good sources for these data: The BLS uses survey data to provide an estimate of wages and salaries, and the BEA creates a measure of wage and salary income while putting together the national income and product accounts.

In the top graph, we plot the BLS and BEA measures of labor income as the red and blue lines, respectively, and look at this relative to GDP (the green line). We normalize all of these series to be 100 in 1982, at the trough of the recession and when the BLS data become available. It’s immediately apparent that the GDP figure is now higher than wages, meaning that it has grown faster since the 1980s. This observation, which isn’t new, is related to a large literature about how the labor share of output has (or has not) diminished. We see the separation in 2015, but this is not a result of developments during the Great Recession. In the bottom graph, we plot these series in growth rates: year-over-year percentages. Notice that through the first decade of the 2000s, GDP growth was almost always faster than wage income growth. Both plummeted in the Great Recession, but since then have been growing at about the same pace. The decline in wages as a fraction of GDP is not a result of a sluggish recovery from the Great Recession, but rather from effects predating it.

How these graphs were created: For the top graph, search for “compensation of employees: wage and salary,” “total wages and salaries,” and GDP. Add each series as a separate line. Then choose the units to be “Index (scale value to 100 for chosen period)” and choose the observation date of 1982-11-01 (the 1982 recession trough when all three series are available). For the bottom graph, follow the same process to show all three series but, instead of choosing an index scale, make the units “Percent Change from Year Ago.”

Suggested by David Wiczer.

View on FRED, series used in this post: A576RC1, BA06RC1A027NBEA, GDP


Subscribe to the FRED newsletter


Follow us

Back to Top