View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

Search

MACROBLOG

November 28, 2016

Does Lower Pay Mean Smaller Raises?

I've been asked a few questions about the relative wage growth of low-wage versus high-wage individuals that are measured by the
Atlanta Fed's Wage Growth Tracker. Do individuals who were relatively lower (or higher) paid also tend to experience lower (or
higher) wage growth? If they do, then wage inequality would increase pretty rapidly as low-wage earners get left further and further
behind.
The short answer is no. As chart 1 shows, median wage growth is highest for the workers whose pay was relatively low (in the
bottom 25 percent of the wage distribution), and lowest for those who were the highest-paid (in the top 25 percent of the wage
distribution). Median wage growth is reasonably similar for those whose pay was in the middle 50 percent of the wage distribution.

To understand what's going on, let's look at the construction of a Wage Growth Tracker sample. In simple terms, a person's wage is
observed in one month, and then again 12 months later. But relatively low-wage workers are less likely to remain employed (and
hence more likely not to have a wage when observed a second time) than other workers. Almost half of workers who are not
employed 12 months later come from the lowest 25 percent of the wage distribution. For workers in a relatively low-wage job, a
greater share who might otherwise have experienced a declining wage left their employment, resulting in a larger share of wage
increases among those who remained employed.
In contrast, relatively high wage earners in the Wage Growth Tracker sample have a remarkably low median wage growth—zero in
recent years. They also have a much greater chance of experiencing a wage decline than other workers (see chart 2).

A

A

A

However, getting a complete picture for high-wage individuals in the Current Population Survey is limited by the fact that
observations are top-coded (or censored to preserve identifiable individuals' anonymity). For example, weekly earnings higher than
$2,885 are currently simply recorded as $2,885. If a person in this circumstance gets a wage increase, it will still be reported as just
$2,885, which would make it seem as if wages didn't increase, even if they did.
Top-coding itself has only a relatively small effect on the median wage growth for the whole sample because top-coded earnings
aren't that common. But they are a reasonably large share of the upper part of the wage distribution, which makes the median wage
growth pretty unrepresentative for people who were relatively high wage earners. In principle, one could try to surmount this problem
by estimating the earnings for top-coded workers, but my experience has been that doing so is likely to add more noise than insight.
What about examining a worker's current wage instead of their prior wage? Is the median wage growth also higher for workers who
are currently in the lowest part of the wage distribution? No. In fact, they are more likely than others not to have received a pay raise
or even to have had the rate of pay reduced. Conversely, someone who is currently in the upper part of the wage distribution is more
likely to have received a larger pay raise than other workers. Some workers move up the wage distribution—but not all.
The bottom line is that the point of reference matters a lot when looking at the tails of the wage distribution, and top-coding limits the
ability to learn much about the wage growth of high wage earners. But for the middle part of the wage distribution, it doesn't matter
so much. The median wage growth of the overall sample is pretty representative of the typical wage growth experience of workers in
the heart the wage distribution.
By John Robertson, a senior policy adviser in the Atlanta Fed's research department

November 28, 2016 in Employment, Labor Markets, Wage Growth | Permalink