Blogging on Employment and Job Markets in US Metro Areas

2016 Metro Area Employment Growth and Housing Permits

Update: a more recent post (dated March 24) covers metro area employment growth and housing permits for the 2011 to 2016 period.

The Census Bureau recently released 2016 housing permit numbers for metro areas (click here), so I decided to compare employment growth to housing permits. The idea is to see how planned increases in housing supply match up (or don’t) with increases in employment and (implicitly) demand for housing.

The housing permits figures are more properly known as “new housing units authorized by building permits”, meaning that those housing units have been approved for construction, and not necessarily that construction has started. Though we do not have data for housing starts or completions by metro area (if you’re reading this and know of a source for metro area starts and completions, please let me know), at the national level housing permits and starts trend together very closely, followed with some lag by completions (see this FRED graph).

Looking at the 51 largest metro areas in the US (those that had populations of 1 million or more as of the 2010 census), we can compare nonfarm payroll (NFP) employment growth to the increase in housing units as gauged by permits. At the end of this post is a table of all 51 metro areas that can be sorted by any of the table’s columns. The right-most column is the ratio of housing units per increase in employment, referred to as “H/E ratio” for the rest of this post. For 2016 the average H/E ratio for the 51 largest metro areas was 0.48.

Caveat: the H/E ratio tells us only about the incremental supply of housing units to be added, and incremental demand as gauged by increase in employment, and does not tell us anything about the existing housing stock or the existing employment base. That is, it does not give us the entire picture of supply and demand for housing.

Among the large metro areas with the top 10 highest rates of employment growth in 2016, the H/E ratio ranges from a bit higher than the average (0.55 for Las Vegas and 0.53 for Denver), to a low of 0.17 for San Jose, California (which is also the lowest of all 51 large metro areas). Orlando, with the highest rate of employment growth among large metro areas in 2016, has an H/E ratio just below the average at 0.46.

Top 10 Large Metro Areas – 2016 Nonfarm Payroll Employment Growth Rate

Top 10 large metro areas NFP employment growth
Next, looking at the large metro areas with the 10 highest H/E ratios, we can see that all but one the top 10 have H/E ratio of 1.0 or higher (and Kansas City, in tenth place, is very close at 0.96). This list also illustrates the importance of the ‘E’ (employment growth) in the H/E ratio: The top two metro areas in this top 10 list, Oklahoma City and Houston, with H/E ratios of 13.54 and 3.02 respectively, saw 2016 employment growth significantly below the national average of 1.4%. If they had experienced employment growth at the same rate as the national average, Houston’s H/E ratio would be around 0.95, and Oklahoma City’s would be around 1.31. Having stated that, this list of metros with the top 10 H/E ratios are building housing (or at least planning to) at a rate roughly in line with employment growth, and thus should be alleviating the pressures causing housing crunches in many parts of the country.

Top 10 Large Metro Areas – 2016 H/E Ratio

Top 10 large metro areas housing units (permits) per NFP employment increase
If we look at the large metro areas with the 10 lowest H/E ratios we can see that all except one had employment growth that was close to or greater than the national average of 1.4% (that one exception is Cleveland, Ohio, which saw employment growth of 1.0% in 2016). A few of the metro areas on this list are most likely dealing with constraints on available land on which new housing can be built, such as Boston and San Jose. Detroit is known to have lots of housing stock that is vacant, so its low H/E ratio is not a surprise.

Bottom 10 Large Metro Areas – 2016 H/E Ratio

Bottom 10 large metro areas housing units (permits) per NFP employment increase
The housing permits data from the Census Bureau shows us single unit and multi-unit figures, and from that I calculated for each metro area the percentage of units that are multifamily. Sorting by this percentage reveals that California cities San Diego and Los Angeles are building a higher percentage of multifamily housing than most cities known for their density. San Diego tops the list with multifamily units at 78.1%, higher than that of New York City. We should keep in mind, though, that these figures are for entire metro areas, and that the greater NYC metro area includes many residential suburbs. Interestingly, two other California metro area have close the lowest multifamily percentages: Sacramento (14.8%) and Inland Empire (20.9%).

Top 10 Large Metro Areas – 2016 Multifamily Housing Unit Percentage (Permits)

Top 10 large metro areas percentage of multifamily housing units

As mentioned above, here is the sortable table of all 51 large metro areas:

2016 Employment Growth and Housing Permits for Large Metro Areas

Metro Area2016 NFP
2016 NFP
% of units
housing units
per NFP
Oklahoma City, OK0.50.1%6,77025.6%13.54
Houston, TX14.80.5%44,64320.7%3.02
Raleigh, NC10.91.8%13,50730.1%1.24
Austin, TX18.81.9%22,24238.8%1.18
Providence, RI2.30.4%2,54734.7%1.11
Memphis, TN4.00.6%4,35531.0%1.09
Richmond, VA4.50.7%4,83718.2%1.07
Phoenix, AZ27.51.4%28,54234.9%1.04
Charlotte, NC18.91.7%19,35332.9%1.02
Kansas City, MO10.51.0%10,06347.8%0.96
Birmingham, AL4.20.8%3,46322.0%0.82
New Orleans, LA3.80.7%2,94916.8%0.78
Nashville, TN24.32.6%18,55735.3%0.76
Hartford, CT3.20.6%2,05459.1%0.64
Tampa, FL28.42.2%17,18037.8%0.60
Chicago, IL32.20.7%19,46958.3%0.60
Portland, OR25.12.2%14,72350.1%0.59
Las Vegas, NV24.82.7%13,57735.1%0.55
Jacksonville, FL22.13.3%11,69827.3%0.53
Minneapolis-St. Paul, MN27.41.4%14,13345.1%0.52
Atlanta, GA70.82.7%36,12136.5%0.51
Dallas-Fort Worth, TX113.53.3%55,61846.3%0.49
Denver, CO44.83.2%21,32252.1%0.48
Orlando, FL50.34.2%23,25138.9%0.46
Rochester, NY4.40.8%2,00035.9%0.45
San Antonio, TX21.82.2%9,78534.2%0.45
Columbus, OH19.21.8%8,24951.0%0.43
Louisville, KY12.11.8%5,00139.5%0.41
Seattle-Tacoma, WA64.63.4%25,51663.2%0.39
Indianapolis, IN19.71.9%7,55425.3%0.38
Salt Lake City, UT23.73.4%8,80051.0%0.37
San Diego, CA28.92.0%10,66978.1%0.37
Washington, DC66.62.1%24,94448.0%0.37
Baltimore, MD22.21.6%8,04041.6%0.36
Los Angeles, CA90.01.5%32,00870.9%0.36
New York, NY120.61.3%42,46676.5%0.35
Pittsburgh, PA4.40.4%1,50933.8%0.34
San Francisco, CA46.62.0%14,98967.1%0.32
Miami, FL60.82.4%18,69464.2%0.31
Philadelphia, PA41.31.4%12,11543.7%0.29
Buffalo, NY7.11.3%1,99253.1%0.28
Cleveland, OH10.91.0%2,95510.2%0.27
Cincinnati, OH21.72.0%5,85932.9%0.27
Sacramento, CA29.33.2%7,21714.8%0.25
Inland Empire, CA40.82.9%10,01920.9%0.25
St. Louis, MO33.62.5%7,94332.8%0.24
Boston, MA53.22.0%13,01759.5%0.24
Detroit, MI33.31.7%7,53624.0%0.23
San Jose, CA36.03.4%6,12766.4%0.17
Milwaukee, WI-4.7-0.5%3,82957.3%
Hampton Roads, VA-2.8-0.4%6,37537.1%
data source: Bureau of Labor Statistics, Census Bureau

Initial Unemployment Claims Update

The DOL has released unemployment claims data for the week ending February 25 (see the DOL news release here).

Initial claims show us a labor market that continues to improve, with the 4-week moving average as of February 25 at 234,250 down 10.3% from a year earlier (here’s a link to the FRED chart below):

FRED chart: year-on-year percent change of 4-wk moving avg of initial unemployment claims

That figure of 245,250 is also well below the low point for the previous expansion, 286,500 in February 2006 (here’s a link to the FRED chart below):

FRED chart: 4-wk moving avg of initial unemployment claims

Weekly Initial Unemployment Claims down 12.1% from a year ago

The Department of Labor reported that for the week ending February 4, seasonally adjusted initial claims were 234,000, down from the previous week’s unrevised level of 246,000. The 4-week moving average was 244,250, a decrease of 3,750 from the previous week’s unrevised average of 248,000. Compared to a year earlier, the 4-week moving average is down 12.1%. Here’s a graph from FRED:

4-week moving average of unemployment cliams down 12.1% from a year earlier

December 2016 Metro Area Unemployment Update

Now that metro area unemployment rates for December have been released, it’s time for an update. For metro areas with populations of 1 million or more as of the 2010 census, here are the 10 lowest unemployment rates:

Dec 2016 - 10 lowest unemployment rates for large metro areas
Topping the list, is Boston with an unemployment rate of 2.5%. Only two other large metros had an unemployment rate below 3%: Denver (2.6%), and Salt Lake City (2.7%). For comparison, the national unemployment rate for December was 4.5%.

Among large metro areas, Boston also tops the list for largest decrease in the unemployment rate over the course of 2016, dropping 1.5% from an already low rate of 4% in December, 2015:

arge metro areas with decreases in employment in 2016
Other metro areas to have seen a decrease in unemployment of 1% or more were Las Vegas (1.2%), Detroit (1.1%), Hartford, CT (1.1%), Seattle-Tacoma (1.0%), and Providence, RI (1.0%). The national unemployment rate fell 0.3% over the course of 2016. The outlier in this list is Detroit with an unemployment rate of 10.4% as of December, more than twice the national rate.

Of the 51 metro areas with population of at least 1 million, only twelve saw unemployment increase over the course of 2016:
arge metro areas with increases in employment in 2016

Most of these metros had low unemployment rates as of December, 2015, and for most of them the increase has only been a fraction of a percent. However, Cleveland not only saw the largest increase in unemployment over 2016, it also has the highest unemployment rate of 10.7% as of December, 2016. Only Detroit comes close with its 10.4% unemployment rate (though Detroit’s unemployment rate decreased in 2016). All other larger metro areas posted unemployment rates less than 5.5% in December.

Metro Area Job Growth in 2016

With the release of metro area nonfarm payroll employment figures for December, we can compare the rate of job growth over the course of 2016. The table below shows the top 10 rates of 2016 job growth for metro areas with populations greater than 1 million (as of the 2010 census), and the change in employment since December, 2007 (the table also includes corresponding figures for national employment growth, for comparison). The sparkline chart on the right side of the table includes a bar for each year (2008 through 2016) showing the percent change in employment since 2007.

2016 employment growth top 10 - metros with population over 1 million

At the top of the list, with employment growth of 4.2% (three times the national growth rate) is Orlando, Florida. The next three top growth rates are for the Salt Lake City, San Jose, and Seattle-Tacoma metro areas, tied at 3.4%. Next in this ranking are Jacksonville, Florida and Dallas-Fort Worth at 3.3%, followed by Denver and Sacramento at 3.2%. Atlanta, tenth in this top 10 list at 2.7%, saw employment growth almost double the national rate.

When looking at the change in employment growth since 2007, seven of these top 10 are at 10% or more, with San Jose, California posting the highest rate of 18.1%. Atlanta and Jacksonville, though they’ve seen less than 10% employment growth since 2007, are both significantly above the national rate of 4.9%. Only Sacramento, California at 3.7% has seen a lower rate of growth since 2007 than the national rate.

Moving on to metro areas with populations between 500,000 and 1 million:

2016 employment growth top 10 - metros with population between 500,000 and 1 million

At the top of this list are Sarasota, Florida and Provo, Utah both with 3.8% employment growth in 2016. Next are Spokane, Washington (3.6%), Boise, Idaho and Deltona, Florida (both at 3.5%).

When looking at the change in employment growth since 2007, half of this group has seen 10% or more, with Provo, UT posting a very healthy 22.9% (over four times the national rate). Not so far behind is McAllen, Texas with 18.4%. Only two of these top ten, Spokane, Washington and Deltona, Florida are showing lower growth since 2007 than the national rate of 4.9%.

Below is the top 10 list for metro areas with populations between 200,000 and 500,000:

2016 employment growth top 10 - metros with population between 200,000 and 500,000
Prescott, Arizona tops this list with 50%, followed closely by Ann Arbor, Michigan at 4.8%.
As we saw above with, half of this group has seen employment growth of 10% or more since 2007, with College Station, Texas posting 22.2%. However, Reno, Nevada, despite healthy employment growth of 3.4% in 2016 is still 0.5% below its 2007 level of employment. Prescott, Arizona, with 5% employment growth in 2016, has seen only 1.2% employment growth since 2007.

When looking these three tables, showing where metro area employment growth was highest in 2016, we can see that roughly half (and more than half for the top ten of metros with populations of 1 million or more) have seen relatively strong growth since 2007. The other half have seen employment growth recover more slowly and/or more recently over the course of the great recession. The red and blue bars in the sparkline charts convey this visually.

Also apparent from these three top then lists is that metro employment growth in 2016 was strongest in the West, particularly California, Washington, and Utah, and warm weather states, mostly Texas and Florida. Notably absent are metro areas in the Northeast and most of Midwest (two smaller midwestern cities, Ann Arbor, Michigan and Sioux Falls, South Dakota are in the top ten for metro areas with population between 200,000 and 500,000).

December 2016 Nonfarm Payroll Update

With the BLS release of December nonfarm payroll (NFP) numbers we can compare 2016 job growth to that of 2015. First, though let’s take a look at year-on-year (YoY) changes since 1990:

Nonfarm Payroll December 2016 update

While the economy continues to add jobs, the rate is slowing: 2.03 million in 2016 compared to 2.79 million in 2015. In percentage terms, NFP employment grew by 1.4% in 2016 compared to 2.0% in 2015. Here’s table showing employment growth (in millions) and the YoY percent change for each year of the current expansion:

Table: NFP Employment Growth 2010 to 2016

2016 has posted the smallest job growth since 2010, and slightly less than 2011. Job growth for the current expansion may have peaked in 2014.

Here’s a table showing NFP employment growth in 2016 and 2015 at the industry level:

NFP employment contribution by industry 2015 and 2016

A few observations:

  • In 2016, as in 2015, growth was largely being driven by these four industries:
    • Education and health services
    • Professional and business services
    • Trade, transportation, and utilities
    • Leisure and hospitality
  • Construction’s contribution for 2016 was less than half that of 2015 in percentage terms.
  • As noted above, manufacturing and state governments switched from positive contributions (adding jobs) in 2015 to a negative (losing jobs) in 2016.
  • Compared to 2015, local governments were a significant source of job growth.

This chart shows the contribution to NFP employment by industry for both 2015 and 2016:

graph of contribution by industry to NFP employment growth 2015 and 2016

Mean vs Median

Mean or median, which should we pay more attention to? First, for some background, the difference between the mean and median:

The mean is the average, the result of dividing the sum of two or more values by the number of values. So for three values, X, Y, and Z, the mean is (X+Y+Z)/3.

The median is the middle value in a set of values sorted in ascending or descending order. If the sample contains an even number of values, the median is defined as the mean of the middle two. To use X, Y, and Z again, if X > Y and Y > Z, then the median will be Y. No matter how many values we have, the median will be middle point of the dataset, with half of the remaining values above the median and the other half of the remaining values below the median.

As an example, lets look at a hypothetical sample of seven salaries for statisticians in a given city. From lowest to highest the salaries are:

$72,000  $75,000  $78,000  $82,000  $85,000  $88,000  $96,000

The mean, or average, of these salaries is $82,286, while the median is the middle value, $82,000. In this case the mean and median differ by only a small amount, so one may be tempted to conclude that we can use them interchangeably. However, the mean and median will not always be so close.

In many real-world situations, such as with salaries or house prices, the mean and the median often differ substantially. This is due to outliers, abnormally low or high values, which have a greater effect on the mean than the median. To illustrate this, let’s add a substantially higher salary to list above, an eighth statistician who has a salary of $145,000. Maybe this statistician is a manager, or for some other reason receives a much higher salary than the other statisticians in our sample. In any case, the average of these eight salaries is $90,125. The median of these eight salaries is $83,500. So with the addition of one outlier, the average has increased dramatically (by $7,839, or 9.5%) while the median’s increase is much smaller (by $1,500, or 1.8%). And this why the median is more often cited than the mean or average when comparing salaries or house prices among cities or over time: outliers or abnormal values have much less impact on the median.

Hourly and Weekly Earnings in the July Jobs Report

The July jobs report, released by the BLS last Friday, shows that wages (specifically average hourly earnings) are now growing at their fastest rate since the end of the great recession. The number most often cited was 2.6% year-on-year growth rate for both June and July. That figure is based on the seasonally adjusted series supplied by the BLS. I prefer to use the unadjusted series, and to smooth out its monthly volatility with a three-month moving average, and then calculate year-on-year growth (labeled as “3moMA %ch YoY” in the graph below). The results are very similar, with June and July showing post-recession high growth rates of 2.8%.

private sector hourly earnings 2016 July
When we look at average weekly earnings, however, while the growth rates are the same for June and July at 2.8% (as above, that’s the year-on-year growth rate of the three-month moving average), this is not a post-recession high. Weekly earnings growth is stuck in the same range it’s been in since 2011:

private sector weekly earnings 2016 July
Here’s another graph showing the both hourly and weekly earnings growth (again, the year-on-year growth rates of the three-month moving average):

private sector hourly and weekly earnings 2016 July
The difference, of course, is explained by average weekly hours worked, also included in the BLS’s jobs report. While average weekly hours has been between 34 and 35 since mid-2011, the year-on-year growth rate has been negative for most of 2016 as the below graph shows.

private sector average weekly hours 2016 July

June 2016 Nonfarm Payroll Update

The release of June nonfarm payroll (NFP) numbers was greeted with headlines such as “Job Growth Surged in June.” Of course, that surge was in comparison to May. The month-on-month changes can be quite volatile, and for that I reason I prefer to look at year-on-year (YoY) changes.

Nonfarm Payroll June 2016 update
June’s NFP employment was up 2.5 million from a year earlier. While that shows that job growth is continuing, it also means the rate of job growth has been slowing since early 2015, as can be seen in the above chart. In the current expansion, the highest year-on-year job growth was 3.1 million in February 2015. In percentage terms, June’s YoY growth was 1.8% compared to 2.3% in February 2015 (see the red series in the bottom section of the above chart). For 2016 YTD (January thru June), average YoY growth is 1.9%, only slightly higher than the average since January 2012 of 1.8%. The slow, but steady job growth of this expansion continues.

NFP employment growth for the first half of 2016 was 1.12 million, compared to 1.39 million for the first half of 2015 (a drop of 19%).

Using the seasonally adjusted industry-level NFP data I’ve prepared a comparison of the first half of 2016 to that of 2015 (2016H1 and 2015H1 in the below graph), showing us how job growth is shifting.

Employment contribution by industry 2015H1 and 2016H1
A few industry-level observations:

  • Education and health services continues to be largest source of job growth, contributing 29.8% in 2016H1, up from 25.7% in 2015H1.
  • Professional and business services remains the second largest source of job growth, but its contribution dropped to 17.2% for 2016H1 from 22.8% for 2015H1.
  • Construction’s contribution for 2016H1 was 4.5%, less than half that of 2015H1 (9.5%).
  • Manufacturing was drag on employment in 2016H1 (-2.3%), compared to a 2.3% contribution in 2015H1.
  • Local government contribution more than tripled to 6.6% in 2016H1 from 2.0% for 2015H1.