Philly Works from Home?

Not long after the onset of the COVID-19 pandemic Kastle Systems — a company that manages office workplace security for a number of businesses across the U.S. — begun releasing a weekly newsletter with data about workplace attendance. The Back to Work Barometer uses daily office key card swipe data from a sample of 300,000 users across 10 major metro areas: New York, Los Angeles, Chicago, Dallas, Houston, Washington DC, Philadelphia, Boston, San Francisco, Austin and San Jose. Together these include 8 of the nation’s 11 largest metro areas with Austin (26th) and San Jose (36th) somewhat smaller.

I’ve been subscribed to Kastle’s weekly email blasts of data for a few years now and have been consistently intrigued by the notably low reported office attendance rates for one metro area in particular: Philadelphia. According to Kastle’s data office attendance rates in the Philly metro area have been hovering around 40% for the last two years, one of the consistent lowest in the recorded set of cities.

In fact, since the start of 2022 Philly has had an office attendance rate around 4-8 percentage points lower than both the average of the 10 city group and the nearby metro areas of New York and Washington, DC which have all clustered fairly consistently just below 50%. The only cities measured in Kastle data that have office attendance rates as low as Philadelphia have been the Bay Area metros of San Francisco and San Jose.

As of writing, data from Kastle’s Occupancy by Day of Week measurement on 9/2/24 has the office attendance rate in Philly ranging from 26.6% on the Friday low to 47.1% on the Tuesday high. This is the one of the lowest peak date in the set of measured cities and far below the average spread of 31.5% to 58.4% over the same time period.

Unfortunately, data from the Survey of Working Arrangements and Attitudes — which has produced detailed data on both work-from-home rates as well as manager plans and employee attitudes about returning to the office — does not publicly release any data about the Philadelphia Metro Area specifically. However, the Census Bureau’s American Communities Survey (ACS) does ask about workplace location and data from its 2022 survey show a bit of a different story.

According to data from the ACS, an estimated 18.6% of all workers 16 or older in the Philly MSA answered that they worked from home when asked what method of transportation they used to usually get to work in the previous week. The work-from-home rate for those in the Philadelphia area was slightly above the national rate of about 15% and in line with other metros in the Kastle set of Dallas, Chicago, Los Angeles, and New York.

Most of the Kastle metros that had the highest work-from-home rates in Census ACS data like San Francisco and San Jose also had relatively low office attendance rates in Kastle data. Notably however Austin had the highest estimated work-from-home percentage in ‘22 ACS data at 27% but also the second-highest in-office rate in Kastle data with the latest four-week rolling average at 60.3% and most of the non-holiday 2022 period in the high 50% to low 60% range. A bit paradoxically, Houston — which was the other metro area Kastle had with high 50% to low 60% office attendance rates — had the lowest work-from home rate of the metro group in ACS ‘22 data at about 13%, the only below the national rate.

Of course these datasets aren’t going to line up one-to-one. An ACS respondent could have marked that they “usually” worked from home in the previous week while still going into the office 2 days of the week, or more if they work more than five days. Additionally the Census ACS is covering an entire year rather than weekly trends. That said, some of the wide variances between work-from-home rates and office attendance rates in various metros Kastle tracks are odd. They also don’t even track with monthly time-series data from the BLS/Census Current Population Survey (CPS).

According to CPS data telework/work-from-home rates in the Philly metro area have remained relatively consistent. Between 25% and 30% of Philadelphia area workers have responded that they teleworked at any time in the previous week in the monthly survey since October ‘22 when the CPS started surveying teleworking. This puts Philadelphia at relatively similar rates as the New York, Los Angeles, Dallas, and Chicago metro areas, all of which have been consistently between one to nine percentage points more than the country as a whole. In the CPS data, Washington, DC, Austin, San Jose, and San Francisco metro have had higher teleworking rates mainly between 30% and 40% over the past 18 months while Houston is the only metro area from the Kastle set that has actually had lower teleworking rates than the entire nation. This monthly CPS data aligns fairly closely with the annual ACS data.

Across industries, workers in the Philly area are working more than in the country overall. But this is particularly true in a few NAICS “supersector” industries which the BLS produces workforce statistics for, specifically manufacturing, professional and business services, and financial activities. In the combined monthly CPS surveys from October 2022 to May 2024 an average of 62.1% of Philly area workers in the financial activities sector stated they worked from home at some point in the previous week, compared to 48.5% for workers in the same sector in the U.S. overall.

So what gives with the Kastle data and Philly? I think there might be multiple factors at play. First off, it does seem like workers in the generally high telework sectors of information, professional and business services and financial activities in Philadelphia are working from home at a higher rate than many of their counterparts in other metros tracked in the Kastle data. In fact, other than major tech hubs like Austin, San Francisco and San Jose, from Oct. ‘22 to May ‘24 Philly has had the second highest teleworking rate for high work-from-home industries at 55.6% according to CPS data.

That said, the gap between Philly and cities that had similar work-from-home rates in the overall ACS and CPS data probably can’t be explained solely by workers in Philadelphia’s high-telework industries working from home five to ten percentage points more. Especially since the Kastle data had in-office rates in Philly at just as low as the Bay Area metros while even the teleworking rates in the CPS data for high work-from-home industries in Philadelphia were still between five and seven percentage points lower than the Bay.

The composition of Philly’s workforce doesn’t seem to be that substantively different from other metros that had very similar overall teleworking rates in the CPS and ACS data. Namely New York, Dallas, and Chicago all had very similar proportions of their total workforce made up by high-teleworking industries like Information, Professional and Business Services, and Financial Activities. New York only slightly more overrepresented by the Information sector and Chicago slightly underrepresented in Financial Activities.

If Philadelphia’s workforce both had a much higher proportion of workers who are more likely to telework and those workers were somewhat more prone to working from home compared to workers in the same industry in other metros, then that would amplify the low office attendance rates for the Philly area.

It doesn’t quite square that Philadelphia — with a 56% work-from-home rate for workers in high telework industries — would have one of the lowest office attendance rates while Dallas office workers have a 47% telework rate but have some of the highest office attendance rates. Much less sensible is the fact that somehow Austin has both some of the highest telework rates among office workers at almost 65% but also some of highest office attendance rates at just over 60%!

Although the Kastle data seems to be able to pick up the general trend office attendance rate for high-teleworking industries, some of the peculiarities in specific cities like Philadelphia and especially Austin are more likely attributable to composition effects. The Kastle data by its very nature is restricted only to offices that they manage and thus can collect key card entrance usage for. Although they might manage a large number of offices in the metro areas that they report data for, to my knowledge there is nothing that Kastle is doing to account for the fact that the offices they manage might not cover a representative sample of even the population of workers in high teleworking industries, much less the entire workforce of a metro area. If this is the case, this isn’t really a deficiency that Kastle could or should fix. They are a private company that’s sells security management software and infrastructure to offices. Doing complex, representative samples of large populations is something that organizations like the Census Bureau and Bureau of Labor Statistics were literally made to do and are much easier to accomplish effectively with the resources at their disposal.

To this end, I would assume that the population of offices that Kastle manages in the Austin metro just happen to be made up of more companies in lower work-from-home industries compared to the population overall. Or Kastle’s offices could be overrepresenting companies in high-telework industries that happen to have stricter office policies. Or some combination of both. Likewise for Philadelphia I think that Kastle’s offices are either with companies that have even higher teleworking rates than even Philly area office workers overall, or that they happen to be overrepresenting locations where more office workers in Philadelphia are teleworking. One last peek into the ‘22 ACS data can show where those office are most likely to be located.

It’s likely that the Kastle offices in the Philly metro area are more likely to be including office workers who life in Montgomery, Chester, and Bucks counties in Pennsylvania, rather than the New Jersey counties of Gloucester, Camden, and Burlington across the river, based on the work-from-home rates of workers in those areas.

Data and code underlying this post can be found here.