Share

U.S-Iran War

The U.S.-Iran War has now been ongoing for the past five months, with numerous starts and stops along the way. There was a graphic within the July 30 daily morning newsletter from the New York Times that I thought successfully captured the cycle of “conflict and calm” between the two nations. The dataset, which comes from ACLED (Armed Conflict Location & Event Data Project), helps to illustrate that a) agreements haven’t led to sustained periods of peace and b) July marked an uptick in escalation.

From an economic perspective, one of the clear impacts of the war has been through elevated energy costs and prices, which have led to increases in the nation’s inflation measures. Looking at regular gasoline (87-octane) price data from GasBuddy for Colorado and the U.S. shows that it still costs substantially more to fill up your vehicle (about $4/gallon) compared to the year and a half that preceded the war. Based on this data, regular gasoline in Colorado peaked at just under $4.80/gallon in mid-May, while prices ranged between $2.25-$3.50/gallon from August 2024 to February 2026.

Federal Open Market Committee

On Wednesday, the Federal Open Market Committee (FOMC) met and determined to hold the federal funds rate unchanged for a fifth consecutive meeting. Nine individuals within the 12-person committee voted to keep the rate constant, while three preferred a one-quarter of a percentage point increase to the rate. While the aforementioned war and impacts to higher rates of inflation should continue to serve as factors for FOMC consideration on rate direction, there does appear to be the risk of shifting to a more opaque structure within the FOMC, led by new Chair, Kevin Warsh. First, the FOMC has moved away from forward guidance under Warsh, which represents a notable change in communication strategy compared to the past quarter century. Second, there seems to be uncertainty in how Warsh will approach data interpretation/utilization going forward, particularly with the Personal Consumption Expenditures (PCE) Price Index, the Federal Reserve’s preferred measure of inflation (bond markets also reacted strongly to Warsh’s press conference last week). Third, the FOMC has met eight times a year since the early 1980s (here’s the posted calendar since 2021); however, it was reported last week that Warsh is considering reducing those number of annual meetings and subsequent votes on the federal funds rate.

2^nd Quarter GDP

The Bureau of Economic Analysis (BEA) released the advanced GDP estimates for 2^nd quarter of 2026 on Thursday. This data showed that real (inflation adjusted) GDP last quarter increased at an annual rate of 1.5%, a slowdown compared to the 2.1% annualized growth of the U.S. economy during the first three months of the year. The Federal Reserve Bank of Atlanta’s GDPNow forecasting model, which is routinely published based on updates to key data inputs, was spot on with the officially published 2^nd quarter rate of 1.5%, while expert forecasted consensus averaged closer to an annualized increase of 2%. However, it’s worth noting that the GDPNow forecast for 2^nd quarter growth was over 3% through late June and only started to stabilize around 1.5% in mid-July. Something to keep in mind considering GDPNow for the third quarter currently forecasts a growth rate of 5%!

Subscribe now

Tariffs

Announced a few weeks ago, a new phase of global tariffs has come into existence, which should add more layers of complexity to the U.S. and world economy (for a colorful take, check out this Substack article from economist, Justin Wolfers). Related to tariffs, I recently came across an analysis from the Peterson Institute for International Economics (PIIE) that did a great job of breaking out U.S. tariff revenue over the past year and a half – so sharing that chart below.

Noise Infusion Update

In June, I wrote about an order on noise infusion and how that could impact a lot of economic data series. Last week I learned that the U.S. Census Bureau will publish the latest release of Postsecondary Employment Outcomes (PSEO) sometime in late August. That’s good news because that implies that existing datasets from PSEO will not be discontinued. However, it’s unknown at this time how future PSEO products could be impacted, as well as other data series I discussed in that June article. Hopefully more updates regarding these data impacts are released in the coming months.

New Level of AI Creepiness

This goes well above my head, but in July there were some unexpected events that occurred during benchmark testing of OpenAI models that were alarming and unprecedented from a cybersecurity and AI capability perspective. For those interested, here’s a July 24 article from the Ringer that serves as a nice (and humorous) explainer for this incident, along with a technical article from Anthropic (owners of Claude) on their own lessons learned. Not to take Anthropic off the hook, though, Claude Mythos definitely has a creep factor.

Data Centers

Finally, everyone’s favorite AI-adjacent topic – data centers. Within the July 30 daily morning newsletter I linked at the top of this article, there was a map gif that showed data on current and planned data centers across the nation. This data, which comes from Cleanview, states that there are presently 1,214 operating data centers in the U.S. (operating capacity of 57,341MW), with that number expected to expand to an additional 1,764 data centers (planned operating capacity of 369,538MW) after 2026. While Cleanview does publish a map showing those current and planned data centers, I thought the NY Times graphic did a better job of conveying the scale in change between the two periods, which you can see below.

When thinking about this likely explosion in data centers following this year (150% increase in the number of centers; 660% surge in operating capacity), I can’t help but to consider the construction labor that will be required for these projects. This need for workers is especially prevalent when also contemplating the concurrent national necessity to build homes.

Using national estimates from the Bureau of Labor Statistics’ Occupational Employment and Wage Statistics (OEWS) program, we can look at occupational-level staffing data for the U.S. residential and nonresidential building construction industries. The following table shows the top 15 occupations that engage in construction work within those two industries. For residential building construction, the list of 15 occupations accounts for about 53% of total employment in that U.S. industry, while for nonresidential construction the listed occupations represent around 47% of total employment. 11 of the 15 occupations in this table are identified in both industries, signaling there is significant overlap in the type of construction workers needed. These common occupations (highlighted in green) include: construction laborers; carpenters and carpenter helpers; construction supervisors; cement masons and concrete finishers; electricians; painters; construction equipment operators; plumbers and pipefitters; drywall and ceiling tile installers; and building inspectors. While nonresidential building construction includes more than just data centers, I think this helps to illustrate how the broad construction industry is competing with itself for labor. This also coincides at a time when the national immigrant labor pool is being severely constricted.

While there’s a wealth of research, analysis, and opinions available on the reasons for the long-term decline of manufacturing employment in the U.S., the following chart displays the respective U.S. and Colorado proportions of total nonfarm employment allocated to the manufacturing industry over time. The U.S. shares (which are available going back to 1939) are shown since 1976, while the Colorado data begins in 1990. In the late 1970s, approximately 22% of all jobs in the U.S. were concentrated within manufacturing, with employment levels reaching their all-time peak of 19.4 million in 1979. By 1990, the nation’s manufacturing employment share had fallen to 16.2%, which was still measurably higher than the Colorado share of 11.2%. Flash forward to 2026 and both U.S. and Colorado manufacturing employment shares have continued declining at steady rates over the past 35+ years. Through the first half of this year, only 1 in 20 Colorado jobs are in manufacturing, compared to a proportion of 1 in 12 jobs at a national level.

Share

Relative to other states, Colorado’s manufacturing employment share is essentially unchanged between the early 1990s and last year, ranking 40^th in 1990 and 41^st in 2025. Over that period, all but two states (Nevada and North Dakota) experienced decreases in their manufacturing employment shares, aligning with the national trend. In 1990, North Carolina had the highest manufacturing employment share, at 26.4%, while Indiana led all states with a proportion of 15.8% in 2025. The table below provides a snapshot for those two years and which states had the highest and lowest manufacturing employment shares.

There are several factors that can determine how much of overall employment is concentrated within a particular industry, so it is also helpful to look at other Colorado sectoral trends over time in relation to manufacturing. The next chart shows the number of jobs within six select Colorado industries between 1990 and 2025: manufacturing; state and local government educational services (e.g. K-12 school districts and public universities); retail trade; accommodation and food services; professional and technical services; and healthcare and social assistance. Remarkably, these six industries combined to account for exactly 53.3% of Colorado’s total nonfarm employment in both 1990 and 2025.

In 1990, manufacturing (dotted blue line) was essentially tied with retail trade for the industry with the most jobs in Colorado – around 170,000. Accommodation and food services had close to 145,000 jobs. The healthcare and social assistance and state and local government educational services industries each had employment counts exceeding 110,000 in 1990, while the number of jobs in professional and technical services fell short of 90,000.

By 2025, manufacturing was the only Colorado industry within this group with fewer jobs compared to 1990, as the employment count had declined to 146,000 (or a 14% decrease). Subsequently, the healthcare and social assistance industry now has the highest employment total in Colorado, at nearly 350,000 jobs (or a 202% increase since 1990). Professional and technical services has experienced the largest percentage jump in jobs over the past 35 years, growing by 237%. Employment in accommodation and food services and state and local government educational services has essentially doubled since 1990. Even retail trade added jobs between 1990 and 2025; however, that rate of growth has slowed substantially over the past quarter century (under 10% since 2001).

From the prior graphic, Colorado’s manufacturing industry reached an employment peak of 190,000 in 1998, meaning total job loss has exceeded 40,000 in those intervening years. To help clearly illustrate this employment shift in Colorado’s manufacturing industry, the chart below presents the year-over-year changes in job totals between 2001 and 2026. The left side of the graph is marked by substantial job loss in Colorado’s manufacturing industry, spanning two national recessions and ten consecutive years. Excluding the 2020 pandemic, the following years featured a Colorado manufacturing resurgence, as nearly 29,000 jobs were added between 2010 and 2022 (breweries, aerospace, food, and chemical manufacturing were the biggest contributors to those gains). However, Colorado’s manufacturing industry has shed jobs over the past four years. If the trend through the first half of 2026 holds (-2,800 jobs compared to the same period in 2025), the state may lose around 10,000 manufacturing jobs since 2022. That four-year employment decline would likely trail only the information and admin, waste, and temp services industry sectors for the largest decrease over that period.

In addition to evaluating high-level industry trends, it’s also useful to analyze occupational characteristics within Colorado’s manufacturing sector. One way to do this is through the utilization of data from the Bureau of Labor Statistics’ (BLS) Occupational Employment and Wage Statistics (OEWS) program. Specifically, OEWS publishes state-level occupational data by various levels of industry, with the most recent research estimates available for 2025. So using this dataset enables the opportunity to assess staffing pattern information for Colorado’s manufacturing industry and how that compares to other states.

The first table in the following analysis looks at how Colorado’s manufacturing industry employment was distributed in 2025 by major occupational group. Additionally, the table also provides Colorado’s national rank for each major occupational group’s respective employment share. Unsurprisingly, production occupations, which are most commonly concentrated within manufacturing, represented the highest occupational share of Colorado’s manufacturing employment in 2025, at 41.0%. However, that was one of the lowest shares in the nation, ranking 44^th (for reference, Alabama’s share of 58.7% ranked 1^st). On the other hand, there were six occupational groups (highlighted in green) with manufacturing employment shares that ranked Colorado in the top 10 compared to other states: architecture and engineering (9.2%); business and financial operations (7.4%); computer and mathematical (5.0%); sales and related (4.1%); food preparation and serving related (2.8%); and arts, design, entertainment, sports, and media (1.0%).

Combined, the six occupational groups highlighted in green accounted for 29.4% of Colorado’s manufacturing employment in 2025. This actually ranked as the highest such share in the nation, outpacing Massachusetts (28.9%), Maryland (28.8%), Arizona (28.7%), Washington (28.0%), Hawaii (27.9%), California (27.1%), and Connecticut (27.1%). This helps frame Colorado as a unique state when occupational distribution within its manufacturing industry is considered.

Broadly, there are 10 detailed occupations that represented at least 2% of Colorado’s total manufacturing employment in 2025. These occupations, displayed in the following table, include a range of jobs, including: various production workers, manufacturing supervisors, machinery mechanics, software developers, industrial engineers, and sales reps. Notably, all of the production workers listed had 2025 annual median wages that were less than the statewide annual median wage of $59,800 for all occupations.

Finally, shifting from detailed occupations to detailed industries, data from the BLS’ Quarterly Census of Employment and Wages (QCEW) program can be used to highlight Colorado areas of specialization in manufacturing. The last table of this article shows 13 detailed manufacturing industries (6-digit NAICS) in Colorado that met two criteria in 2025: 1) at least 1,000 employees and 2) a location quotient of at least 1.3. A location quotient, which can be utilized as a proxy for industry specialization or expertise, is a measure of employment concentration in an area (state, metro area, or county) compared to the U.S. (a value above 1 indicates a higher concentration relative to the nation).

Share

Geographic Mobility: People Are Moving Less Frequently

The U.S. Census Bureau’s American Community Survey (ACS) contains a treasure trove of incredibly useful and diverse economic data. One dataset measures geographic mobility in the U.S, which estimates how many people change their place of residence compared to the previous year. The following chart shows the estimated share of the U.S. population that remained in the same house as the prior year, by age group, between 2013 and 2024. Older populations, 45 to 64 (green line) and 65+ (purple line), tend to age in place, which is reflected by the high rates of maintaining their same residence (90%+ between 2013 and 2024). Relative to older populations, younger individuals, 25 to 44 (red line) and 18 to 24 (blue line), are more mobile; however, those age groups were also less likely to move in 2024 compared to 2013. For instance, the share of the U.S. 25 to 44 population remaining in the same house increased from 80% in 2013 to nearly 84% in 2024.

In addition to geographic mobility for the U.S. population, these figures are also generated at the state level. As shown below, the share of Colorado’s 25 to 44 population remaining in the same house rose from about 74% in 2013 to 80% in 2024 (another way to think about that: if the 2013 rate was applied to Colorado’s 25 to 44 population total in 2024, there would have been 108,500 more individuals that moved in 2024). Notably, the bulk of that increase has taken place since 2019, when the rate was approximately 76%. While still lower than the 2024 U.S. share of 83.7% for those 25 to 44 that remained in the same house as the prior year, the gap for Colorado has narrowed compared to the nation.

Colorado Housing Supply Trends

The first chart in this section shows the average number of building permits for 1-unit/single-family homes in Colorado over a six-month moving period since 1990. While a permit doesn’t guarantee a structure will actually be built, the data can serve as a valuable proxy for building activity. There were notable surges in permits from the late 1990s to 2005 (see: pre-Great Recession) and briefly following the pandemic. Current levels over the past few years are the lowest since 2015-16, but still much higher than the 2009-11 nadir, when average building permits for single-family homes in Colorado were under 1,000 per month.

On the other hand, the average number of permits for multi-unit home structures in Colorado has less of a clear long-term trend compared to single-family homes. However, the past 10 years reflect elevated levels of multi-unit building permits authorized in Colorado, marked by a peak exceeding 2,500 in late 2021. Similar to single-family homes, building permits for multi-unit housing structures in Colorado fell drastically following the pandemic, but have rebounded throughout 2025 and the first half of 2026.

The information above can be combined into a single chart, to show the relationship between single-family and multi-unit building permits in Colorado. 2001 and late 2021 are the only periods where the six-month moving average exceeded 5,000 permits. Multi-units have made up an average of 45% of total authorized permits in Colorado between 2021 and 2026, while that average was close to 20% in the 20 years prior.

The last chart in this section displays the average count of new monthly listings added in Colorado over a moving 12-month period since mid-2017. Here, new listing levels have been consistently below 10,000 in the state since late 2022, bottoming out around 7,700 in 2023. While recent levels have rebounded compared to 2023, they have remained flat and stuck under the 9,500 mark for an extended period.

Beyond these snapshots of Colorado home building activity and availability over time, I recommended checking out the Colorado State Demography Office’s Housing Shortfall study from last fall. Their four key takeaways were:

• The housing shortfall for 2023 [was] estimated to be 106,000 units.

• The shortage peaked in 2019 at 140,000 units.

• Based on the latest population projections [at time of the study], 34,100 units would need to be added each year to maintain the existing shortage over the next ten years.

• Despite recent decreases in the housing shortfall, housing affordability reached a record low in 2023.

Subscribe now

Affordability

Discussing the topic of housing affordability can cover a wide range of approaches, so I settled on a fairly narrow scope given the time allotted for the presentation. First, I focused on data on the cost to construct a single-family home in the U.S. over time. Since 1998, the National Association of Home Builders has conducted an intermittent Cost of Construction survey. The most recent survey from 2024, containing a representative sample of 4,000 home builders, showed that the average total sales price of a single-family home in the U.S. was $665,000, with construction costs making up 64.4% of that amount (a series high, as shown in the chart below). A rise in material and labor costs since the pandemic were identified as drivers in that uptick in total construction cost as a share of total sales price.

Regarding construction labor cost, one way to evaluate that change over time is to look at the average hourly earnings for the industry. The following figure takes data from the Bureau of Labor Statistics and displays average hourly wages for U.S. construction workers for the month of June from 2012 and 2026. Between 2012 and 2019, average hourly earnings grew by 20% (or 2.8% annually), while that change increased to 35% (or 5.0% annually) from 2019 and 2026. However, this growth also mirrors the change in average hourly wages for the overall private sector during these periods, so the construction industry was not an outlier.

My second focus for discussing housing affordability revolved around highlighting a tool from the Federal Reserve Bank of Atlanta, called the Home Ownership Affordability Monitor (HOAM). Per the Atlanta Fed, “this tool tracks the relative changes in home ownership affordability at a higher frequency and more granular level of geography than other current housing tools. It measures the ability of a median-income household to absorb the estimated annual costs associated with owning a median-priced home.” While documented within their methodology section, HOAM utilizes a couple ways to track unaffordability: 1) as the annual cost of homeownership exceeding a 30% share of the annual median household income [in an area] and 2) as an index value, where “a value below 100 indicates a median income family would not be able to afford a median-priced home given the current interest rate”.

Going back to 2005, HOAM provides these affordability measures for over 700 core-based statistical areas (CBSA). Since May 2022, over 50% of these areas across the U.S. have been determined as unaffordable, with the series reaching a peak rate of 79.5% in July last year. This is in contrast to the ten years between 2012 and 2021, when the average share of unaffordable CBSAs was a little over 20%.

This same approach can be applied to the 17 CBSAs in Colorado, which includes metro areas as large as Denver and Colorado Springs, to smaller areas like Alamosa, Rifle and Sterling. With the exception of a handful of months, all 17 areas in Colorado have had an index below 100 since May 2022.

Other Considerations/Looking Forward

Obviously, a crucial component to building and creating housing is having access to construction workers. The next chart looks at the percent change in U.S. construction employment over the past four years, including the primary components that make up the construction industry (residential building construction and specialty trade contractors; nonresidential building construction and specialty trade contractors; and heavy and civil engineering construction). There was a clear slowdown in construction job growth in 2025 and 2026 (through June) compared to prior years; although this mirrors trends observed in other industries. However, residential building construction and specialty trade contractors employment contracted in 2025, with accelerated losses experienced through the first half of 2026. The other two construction sub-sectors have stable and positive job growth this year.

Looking at U.S. residential construction industry employment over time reveals that the only years where job growth had previously declined was from 2007 to 2011, during the Great Recession, and in 2020, during the pandemic. A restrictive U.S. immigration policy is likely a contributor to employment decreases in 2025 and 2026, but it is not possible to comb out those effects with this dataset.

And while there were a few more things I covered in the presentation, please allow me to take a Princess Bride approach, and sum up:

• Actions from the Federal Reserve, in response to rising inflation and a quirky labor market, will be key due to the relationship between the Federal Funds Effective Rate and the 30-Year fixed rate mortgage average in the U.S. However, the odds of a rate increase are significantly higher than a rate cut in the coming months.

In Summary

• Geographic mobility is slowing in the U.S. and Colorado, particularly for those aged 25 to 44, who have historically been in that prime, home-buying demographic.

• Colorado has a housing supply issue.

• Housing costs national and for Colorado have surged, putting a strain on affordability.

• U.S. residential construction has contracted in 2025 and 2026, potentially impacting the ability to build/relieve the supply issue.

• Current macroeconomic conditions and uncertainly are likely headwinds to any meaningful reduction in mortgage rates in the second half of 2026, adding to the affordability issue.

Share

Between December 2025 and May 2026, Colorado’s labor force total contracted by nearly 48,000 (from 3.254 million to 3.206 million persons), which translates to a growth rate of -1.5% over a five-month period. As shown in the following table, the recent percent decline in labor force is the largest in Colorado’s history (going back to 1976) outside of the 2020 pandemic. Other sizeable drops (of at least -0.7%) over a short period of time have occurred during national (2009) or state (1987) recessions.

Compared to other states, Colorado’s recent labor force decline is 5^th largest, trailing only Connecticut (-1.7%), Rhode Island (-1.8%), Arizona (-1.8%), and Vermont (-2.0%). However, it’s worth noting that a total of 36 states experienced a reduction in labor force between December and May, while the U.S. percent change over that period equals -0.8%. So it’s not like Colorado is necessarily a national outlier with this recent trend, but the magnitude of this downward shift in labor force is unprecedented for the state.

The chart below shows the annual percent change in Colorado’s labor force between 1977 and 2026. Through the first five months of this year, Colorado’s labor force has declined by 2.1% compared to the same months in 2025. That decrease is three times greater than during 2020. Additionally, there have been only four other times when Colorado’s labor force dropped year over year: 1986, 1987, 2010, and 2025.

As to the cause(s) for this rapid decline in Colorado’s labor force (starting in 2025, but accelerating in 2026), that remains murky. Slowing population growth, driven by atypically weak domestic net migration, is a possible reason, but that data is lagged by a year and therefore difficult to line up with the most recent trends in labor force. Additional population impacts due to restrictive national policies on international migration could also be a factor – although based on data from the 2024 American Community Survey, a little over 10% of Colorado’s population was foreign-born, below the national average of around 15%. Another potential cause could stem from an aging population that is exiting the labor force – after all, members of the Baby Boomer Generation (born between 1946 and 1964) are between the ages of 62 and 80 this year. However, when looking at state-level shares of private sector employment over the age of 64, Colorado was a middle-of-the-road state in 2024, as 7.5% of workers met that age threshold (Utah had the lowest share of older private sector workers at 6.5%, while Hawaii had the highest share at 9.7%). Finally, weak economic conditions can be added as a potential culprit for Colorado’s shrinking labor force, but that explanation is muddied by the fact that the state’s unemployment rate has been essentially unchanged over the past year, ranging between a narrow band of 3.8% and 4.0%. This trend runs counter to prior instances with large decreases in Colorado’s labor force, which coincided with subsequently steep rises in unemployment.

What the heck is noise infusion and why do we care? Essentially, noise infusion is a method utilized to protect respondent confidentiality (either self-reported through a survey or provided via administrative data) within various statistical products.

Share

According to a recent LinkedIn posting from John Abowd, former Chief Scientist and Associate Director for Research and Methodology at the Census Bureau, several “data products [will be] affected by the recent Department of Commerce order banning noise infusion”, including:

• All tables and Public Use Microdata Samples (PUMS) from the American Community Survey

• All tables and microdata from the 2020 Census of Population and Housing

• Every key product from the Longitudinal Employer-Household Dynamics (LEHD) program:

• Quarterly Workforce Indicators (I used this dataset within a previous Substack article to highlight age characteristics for Colorado industries; additionally, check out this 2014 paper from Mr. Abowd about the application of noise infusion within QWI, if interested)

• OnTheMap (vital source of information to measure commuting/work-home patterns at various geographic levels)

• Postsecondary Employment Outcomes (fantastic resource for looking at postsecondary earnings outcomes and industry flows 1, 5, and 10 years after graduation)

• Job-to-Job Flows (I had an entire Substack article dedicated to this concept and for discussing Colorado trends over time)

• Veterans Employment Outcomes (unique dataset that measures employment outcomes for the veteran population)

• Popular statistical products used to measure and evaluate business activity:

• Business Dynamics Statistics

• Business Formation Statistics

• All tables from County Business Patterns and Zipcode Business Patterns

• And other products:

• Opportunity Atlas

• Geospatial Environmental data

• Many releases from the Federal Statistical Research Data Centers

• All experimental data products under current development

As you can see, the impact of this relatively obscure-sounding order could be extensive and significantly hamper our ability to measure the local and national economy, population, and labor force.

For a layperson perspective, check out this NPR article from June 12. For a deeper dive, I recommend reading this recent Substack from Nathan Goldschlag, Director of Research at the Economic Innovation Group and former economist at the Census Bureau.

While healthcare and social assistance is still dominating employment growth in the U.S. (nearly 200,000 jobs added since February), increases over the past three months have also been buoyed by other industries. Accommodation and food services payrolls have grown by approximately 130,000 between March and May, compared to around 40,000 from January 2025 and February 2026.Trade, transportation, and utilities; government; and administrative, waste, and temporary services are three industries that have combined to add 212,900 jobs over the past three months, an impressive feat considering those same three industries shed an aggregated number of over a half million jobs spanning the 14 months prior. Finally, two goods-producing industries, construction and manufacturing, have both gained at least 20,000 jobs since February. Of note, manufacturing, which lost 110,000 jobs between January 2025 and February 2026, has experienced employment growth of 22,000 jobs over the past three months.

Teleworking

Through HTOPS, the Census Bureau estimated that approximately 34% of U.S. workers teleworked at least 1 day in June 2025, while 66% did not telework or work from home. The most frequent occurrence of teleworking was 5 of more days, indicating full-time remote work – approximately 15% of all workers fell into this category. 12% of workers experienced a schedule that relied on 1-2 of teleworking, with the other workdays in-office.

The Census Bureau also collects and publishes a ton of demographic/characteristic data for these HTOPS topics, including age, sex, sexual orientation, race and ethnicity, education, marital status, family/household income, and others. Notably, there was little difference in teleworking between female and male workers, with rates of 33% and 35%, respectively. However, there were significant differences in telework rates when comparing by education and family income levels.

For those workers with less than a bachelor’s degree, only 20% teleworked at least 1 day in June 2025, compared to a telework rate of 55% of those with a bachelor’s degree or higher. The gap in the telework rate is likely driven by differences in concentrations of industry and occupational employment between these two groups and the likelihood for remote work.

For similar reasons, those workers with higher family/household incomes ($100,000 and over), were about two times likely to telework at least 1 day in June 2025 compared to individuals with lower family/household incomes (less than $100,000).

Admittedly, the Bureau of Labor Statistics also produces data on U.S. teleworking trends (most recent month available, over time, and by occupation/industry), so I’m not sure if HTOPS telework dataset is really breaking new ground here.

Share

AI Usage and Opinions

The HTOPS for June 2025 asked respondents a series of questions on use usage and opinions of AI. Like with the telework data above, these data distributed by different characteristic/demographic groups. Because I thought it would be interesting to look at generational divides with an emerging technology, the following AI usage and opinion statistics highlight trends for three broad age groups: 18 to 39, 40 to 54, and 55 and older. As shown in the chart below, more than half of households had used an AI tool within the three months prior to being surveyed in June 2025. Usage rates above 60% were recorded for both the 18 to 39 and 40 to 54 age groups, while 45% of those 55 and older had recently used an AI tool.

The survey provided a range of options the respondent could select how they utilized AI tool, including: finding factual information; assisting with schoolwork; assisting with projects; performing a task instead of hiring someone; assisting with creative tasks; and brainstorming or idea generation. A respondent could select multiple categories. The next chart shows the frequency of category chosen by age group as it relates to using AI for assistance. Unsurprisingly, the 55 and older cohort didn’t use AI that much for schoolwork compared to the younger individuals, but the frequency of this category for those under the age of 55 was still under 30%. Relatively higher rates were observed for all age groups when using AI to assist for either work projects or creative tasks.

The highest frequency across age groups from these usage measures came from using AI to find factual information. All three cohorts had a frequency of at least 55% with this category, led by the 55 and older age group (66%). However, there was a sizeable gap in frequency between the age groups for using AI as a brainstorming or idea generation tool (18 to 39 = 59%; 40 to 54 = 37%; 55 and older = 21%). Finally, a low share of respondents across each age group stated that they used AI to perform a task instead of hiring someone.

Like with AI utilization, the survey also provided a range of options the respondent could select regarding their opinion of AI, including: AI tools make me more productive; I control how my data is used by AI; AI has changed my field of work; I trust information from AI systems; I feel prepared to use AI at work; I feel prepared to use AI in daily life; and I am worried about AI’s impact on my career. The following graphic shows the frequency of category chosen by age group as it relates to opinions on AI and work. Less than 20% of respondents across all age groups thought that AI had changed their field of work. However, there was a relatively elevated share of younger respondents that were worried about AI’s impact on their career (34% for those between 18 and 39 and 25% for individuals aged 40 to 54). As of June 2025, under one-third of respondents across all age groups felt prepared to use AI at work.

Between the age groups, there appeared to be a shared skepticism in how they could control how their data is used by AI and that they could trust information from AI systems (for these two categories, no frequency exceeded 20% for any age group). However, there were relatively positive opinions on the ability of AI tools to make respondents more productive, which ranged from 32% for those 55 and older to 49% for individuals between 18 and 39 years old.

As these responses reflect AI use and opinions from almost a year ago, it will be interesting to see if/how these trends shift over time when more recent data is made available.

Colorado labor force continues decline in early 2026

On Wednesday the Bureau of Labor Statistics released new labor market data for states through February. I’ll keep this fairly short since I’ve already extensively covered Colorado’s labor market over the past month, but early 2026 so far looks like a continuation of 2025 for the Centennial State.

• Loss of 7,200 total nonfarm jobs in February, followed by a small, revised gain of only 1,300 jobs in January (seasonally adjusted).

• Sizeable declines in total employment and the labor force through the first two months of the year (shown in the chart below).

• Colorado’s seasonally adjusted employment-population ratio and labor force participation rates in February (64.0% and 66.6%, respectively) are at their lowest levels since the 2020-21 pandemic.

Subscribe now

Total nonfarm employment

When looking at the total nonfarm jobs estimate for Colorado, which includes both private sector and government employment, the state lost 7,700 jobs in 2025, translating to a growth rate of -0.3%. Going all the way back to 1961, last year marks only the 7^th time that Colorado has experienced an annual decline in total nonfarm jobs, as shown in the table below. In fact, there are only five other states that have had fewer instances of total nonfarm decline over the past 65 years (Arizona, Florida, Nevada, New Mexico, and Utah).

Additionally, Colorado’s job growth rate of -0.3% last year ranked 44th nationally. There have been only three other years in which Colorado has had a worse ranking in job growth compared to other states: 2003 (46th), 1988 (48th), and 2002 (49th). The following table displays how Colorado’s national ranking has been distributed since 1961. Nearly 70% of years have featured a Top 20 growth rate ranking for the state, while a ranking of 31^st or worse has only occurred about 18% of the time.

Which industries drove job loss for Colorado last year?

At a high level, Colorado’s private sector shed 10,100 jobs last year (growth rate of -0.4%), while government added 2,400 jobs (growth rate of +0.5%). Out of 20 industry sectors (17 private, along with federal, state, and local government), there were a total of 13 that lost jobs in 2025. That was led by the hodgepodge industry of administrative, waste, and temporary services, at -6,700 jobs, and retail, with a decline of 5,500 jobs. A handful of industries had payroll decreases of at least 2,000 last year, including: information (-3,300), accommodation and food services (-3,100), manufacturing (-2,900), federal government (-2,100), and wholesale trade (-2,100). The chart below presents this list of 13 Colorado industries that lost jobs in 2025. Going back to 1991 there has been only two other years with more industries experiencing year-over-over job declines: 2009 during the Great Recession (15 industries) and 2020 during the height of the pandemic (16 industries)

Which Colorado industries had job gains last year?

Conversely, only 7 Colorado industries added jobs in 2025. As shown in the following graphic, healthcare and social assistance was the big winner in job growth last year (both for the state and the nation), as the employment count grew by 15,000. Local government was the other industry with sizeable gains in employment, with an increase of 4,000 jobs in 2025. The remaining 5 industries added between 200 (finance and insurance) and 1,200 (transportation, warehousing, and utilities) jobs.

Share

What the heck happened in June and July!?

As I was reviewing the revised jobs data I noticed bizarre shifts in the figures during the summer. Based on the seasonally adjusted total nonfarm estimates, Colorado lost 24,200 jobs between May and June, and then regained 19,400 jobs between June and July (shown below).

At first glance it looked like an estimation error – after all, that decline in June would be the largest one-month drop in state’s history excluding April 2020. However, following further analysis, nearly every state experienced job loss in June, so it appears to be a potential impact at a national level. That said, half of states also had job declines in July, which runs counter to the Colorado trend for that month discussed above. It’s possible this pattern could be related to tariff disruptions introduced in early April last year (i.e. “Liberation Day”); but that connection is uncertain given how convoluted the timeline is to follow.

How did Colorado’s CES pre-benchmark estimates perform last year?

I’ve mentioned this in previous articles, but the data the CES program produces throughout the year are based on a survey of businesses. These estimates are then revised annually; utilizing adjustments to things like seasonal factors, company births and deaths, and benchmarks to administrative data from unemployment insurance tax reports (this link goes into more detailed information for those who want something to read before they go to sleep). Each year, the Bureau of Labor Statistics produces a report that provides an analysis of the size of each state’s revisions in total nonfarm payrolls for the month of September. For the third straight year, Colorado had relatively large downward revisions in September nonfarm employment, measuring over a -1% change compared to initial estimates. Disregarding sign (+ or -), Colorado had the 13^th largest revision to September nonfarm employment in 2025.

Colorado’s labor force growth was historically weak in 2025

In the aforementioned article from a couple weeks ago, I posited that we might see an increase in Colorado’s unemployment rate towards the end of the year since it seemed unlikely that the state’s labor force growth in the second half of 2025 would remain negative after annual revisions. However, revisions show that Colorado’s seasonally adjusted labor force still declined between June and December, although by a lower amount than previously estimated (change of -4,100 compared to -23,600). On an annualized basis, Colorado’s labor force (which includes both employed and unemployed persons) decreased by 6,500 last year, equivalent to a growth rate of -0.2%. Going back to 1977 when comparable data was first available, Colorado’s labor force decline in 2025 of -0.2% is the second largest in the state’s history, trailing the -0.7% decrease in 2020. Additionally, it’s rare for Colorado to experience a drop in labor force, as that has only happened four prior times (1986, 1987, 2010, and 2020).

Compared to other states, Colorado’s labor force growth rate ranked 41^st last year. That marks the ninth time in 49 years that Colorado has fallen outside of the top 25 nationally with that metric.

To conclude, while Colorado’s seasonally adjusted unemployment rate is currently below 4% (3.9% as of January 2026), that figure is masking atypical weakness in the state’s labor market, particularly for job and labor force growth identified above. So monitoring these trends in 2026 may require more nuance and scrutiny than in past years.

Jobs report for March

The topline numbers published by the BLS for March looked surprisingly good at first glance. Total nonfarm employment increased by 178,000 jobs compared to February, the largest one-month gain for the U.S. since December 2024. Healthcare and social assistance (+89,900), construction (+26,000), transportation and warehousing (+21,000), and manufacturing (+15,000) led all industries with employment growth last month. However, February’s jobs estimates were revised down by 41,000, so the change compared to January is now -133,000 rather than the initially reported decline of -92,000. Excluding last October, when the government shutdown drove a nonfarm payroll decrease of 140,000, December 2020 was the last time with U.S. over-the-month loss of at least 130,000 jobs. While these figures are seasonally adjusted, it is possible that weather and a California nursing strike were factors for the relatively weak and strong employment changes over the past two months.

Additionally, the nation’s unemployment rate dropped to 4.3% in March compared to 4.4% in February. This “improvement”, however, was determined by a rather large change in the labor force – declining by almost 400,000 – coupled with a relatively smaller decrease of 64,000 in total employment.

Related to jobs and labor force figures, the Federal Reserve recently released a short and interesting analysis about how weakening/near-zero U.S. labor force growth can impact breakeven employment and potential GDP growth. The report highlights two impacts that are being driven by growth in the nation’s labor force that has substantially slowed due to immigration policy and an aging population:

1. “Near-zero labor force growth implies that breakeven employment growth (i.e. the pace needed to maintain a steady unemployment rate) would also be near-zero—making negative job growth almost as likely as positive job growth in a given month.”

2. “[Near-zero labor force growth] implies that any growth in potential GDP will need to come entirely from productivity growth. These features would represent a significant shift in U.S. labor market dynamics and the composition of economic growth.”

Subscribe now

All-time low response rate for the household survey

The aforementioned unemployment and labor force data comes from the Current Population Survey (CPS), also known as the household survey (this not only informs national labor market trends, but also every region and state in the U.S.). The CPS, a joint effort conducted by the BLS and the Census Bureau, attempts to collect these vital labor market statistics from 60,000 U.S. households each month. Since this is a survey, high response rates are crucial in order to produce quality and accurate data. However, as I covered in an article in August, significantly reduced funding for statistical agencies and a growing reluctance to participate in surveys over the past 10-15 years have led to rapidly deteriorating response rates for the CPS and other key surveys. Last month, the CPS hit an all-time low response rate of only 63.9%. The chart below shows the CPS response rate from March 1994 to March 2026. Between 1994 and 2012, the average monthly response was 92%. CPS response rates started to routinely fall below 90% starting in mid-2013 and by 2019 the rate had declined to 83%. Over the past seven years, the decline in the response rate has accelerated, spurred on by the pandemic and the other factors I noted above.

Share

U.S. hires rate continue to look historically weak for several industries

Finally, the BLS also released February statistics from the Job Openings and Labor Turnover Survey (JOLTS) last week. While the monthly data can be volatile (due to an abysmally low response rate of less than 40% since 2022), I thought it might be worthwhile to look at the hires rates for some select U.S. industries. JOLTS, which attempts to collects data from about 21,000 businesses each month, defines hires as “all additions to the payroll during the entire reference month”, minus transfers/promotions, those returning from strike, and temporary or contract employees supplied to a company. The hires rate is determined by dividing the number of hires by total payroll count in a given month. For instance, JOLTS estimated there were about 4.85 million seasonally-adjusted total nonfarm U.S. hires in February, compared a total nonfarm employment total of around 158.5 million. That proportion results in a seasonally-adjusted total nonfarm hires rate of 3.1%.

In order to account for some of the monthly volatility previously mentioned, the following chart shows the U.S. nonfarm hires rate from 2002 to 2026 based on a six-month moving average. The current hires rate of around 3.25% is the lowest it has been since 2013, and only slightly above levels during the Great Recession (2009-10). Peaks rates (above 4%) occurred after the height of the pandemic when the national and global economy began re-opening.

Hires rates are also available at the industry level, which can help explain what could be driving the subdued rate for total nonfarm. There are three U.S. industries that have hires rates that are currently at or below historic lows: construction; retail; and real estate, rental and leasing.

Additionally, there are two other industries – manufacturing and professional and business services – that presently have historically weak hires rates, but that are still above all-time lows.

Subscribe now

Colorado counties with the largest populations (as of July 1, 2025)

This new set of data reflects population levels as of July 1, 2025. Based on these estimates, there are 11 Colorado counties (out of 64 total) that have populations of at least 100,000. El Paso County has the largest population at around 757,000, just edging out Denver County with 740,000 people. Arapahoe, Jefferson, Adams, and Douglas counties (all part of the Denver MSA, more on that later) have the next largest populations, ranging between 399,000 and 674,000. The other Colorado counties with the populations above 100,000 are sprinkled throughout the state and include Weld and Larimer (378k each), Boulder (328k), Pueblo (169k), and Mesa (163k). Mesa, home to Grand Junction, is the only county in this list not located within a stone’s throw of I-25. These 11 counties had a combined population of 5.12 million, representing 85.2% of Colorado’s population total of 6.01 million last July.

The number of Colorado counties with population declines spiked last year

In addition to new estimates for July 1, 2025, the Census Bureau also released revisions to population estimates for July 1, 2020 to July 1, 2024. Using these July 1 estimates over the past five years, we can calculate how many Colorado counties experienced population declines from one year to the next. As shown in the chart below, from 2021 and 2024, there were between 19 and 23 Colorado counties that had year-over-year decreases in population. However, that figure spiked to 33 counties when measuring county-level population changes between July 1, 2024 and July 1, 2025. Notably, there were 10 counties with population gains every year between 2021 and 2024, but saw that trend reversed in 2025: Arapahoe, Delta, Garfield, Montezuma, Montrose, Park, Pueblo, Routt, San Juan, and Summit.

Share

Colorado counties with the largest positive and negative changes in net domestic migration last year

There are two primary factors, or components, that can determine how an area’s population changes over time. The first component is vital events, which is just a fancy way to say a population can change due to births and deaths. The second component is net migration, which measures net movements of the international and domestic/U.S. born population from one area to another (i.e. the number of people who move into an area minus the number of people who move out of that area). As discussed in a prior article, Colorado had a historically low change in total net migration between July 1, 2024 and July 1, 2025, with domestic net migration actually decreasing by 12,100 over that period. With this new set of population estimates, we can now see which Colorado counties had the largest increases and declines in domestic net migration last year (note: some of these county-level changes can be driven by people moving from one Colorado county to another, rather than out of state). Weld, Douglas, and Larimer counties all had positive changes in domestic net migration of at least 2,000 people, while Boulder (-2,248), Denver (-8,023), and Arapahoe (-9,859) had the largest drops in domestic net migration figures last year.

Colorado counties with the largest positive and negative changes in population from 2020 to 2025

The final county-level chart looks at those Colorado counties with the biggest changes in population between July 1, 2020 and July 1, 2025. Over this five-year period, Colorado’s total population expanded by about a quarter-million people, or a growth rate of 3.9%. Weld County added the most to this growth, with a population increase of nearly 47,000 over the past five years. Unsurprisingly, other large counties were also significant contributors to the state’s gain in population since 2020, including: Douglas (+39k), Adams (+34k), El Paso and Denver (+23k each), Arapahoe (+18k), and Larimer (+17k). Twenty-three Colorado counties had declines in population between July 1, 2020 and July 1, 2025, led by Jefferson and Boulder counties, both with losses exceeding 2,000 people. A geographically diverse group of counties round out the seven largest population decreases, including: Eagle (-1,366), Otero (-976), Pitkin (-738), Logan (-567), and Rio Grande (-451). While not displayed in the following graphic, Custer (18.3%), Elbert (16.5%), San Juan (14.5%), and Weld (14.2%) led all Colorado counties with the fastest population growth rates, far outpacing the state. However, of those four counties, only Weld has a population of more than 31,000 people.

Population trends and data for Colorado metro areas

As previously mentioned, along with county-level data, the Census Bureau also released population estimates for the 393 metropolitan statistical areas (MSA) within the U.S. and Puerto Rico. Colorado has seven MSAs, five of which are concentrated in a single county and two that span multiple counties. In total, 17 Colorado counties are within an MSA, while 47 counties fall outside of that designation. The table below shows some interesting population statistics for these areas – with population total data covering the beige cells and population change data within the purple cells.

The MSAs combined to represent 88% of Colorado’s population in 2025, with Denver-Aurora-Centennial by far the largest area with over 3 million residents (ranking 19^th nationally). Denver-Aurora-Centennial also had the largest change in population between July 1, 2024 and July 1, 2025, adding over 10,900 people. Greeley (+7,100), Colorado Springs (+4,700), Fort Collins-Loveland (+3,100), and Grand Junction (+1,400) were the other MSAs with gains in population last year, while Pueblo and Boulder each saw declines of around 900. Combined, population in Colorado’s seven MSAs grew by about 25,500 in 2025, compared to a loss of 1,450 in the 47 counties that make up the non-metro total. Finally, expressed as a growth rate, Greeley (1.9%), Grand Junction (0.9%), and Fort Collins-Loveland (0.8%) had the fastest changes in population last year, ranking 20^th, 93^rd, and 99^th, respectively compared to all other MSAs.

Subscribe now

To start, we’ll look at estimates for Colorado jobs in 2025, specifically figures produced by the Current Employment Statistics (CES) program. While I won’t bore you with how the sausage is made, data from the Quarterly Census of Employment and Wages (QCEW) plays a critical role in determining how CES estimates are revised each year and were therefore used for this predictive exercise (this article from 2024 goes into detail about the importance of QCEW if interested).

When looking at the total nonfarm jobs estimate for Colorado, which includes both private sector and government employment, the state likely lost around 16,900 jobs in 2025. The majority of that loss should be concentrated within Colorado’s private sector, which may show a decline of 14,900 jobs last year, with government making up the remaining decrease. Going back to the early 1990s, 2020 is the only previous year in which Colorado experienced job loss for total nonfarm, and the private sector, and government. The following two charts show those expected job losses for Colorado in 2025, expressed as absolute values and as year-over-year percent changes.

Next we’ll review the expected changes for Colorado’s ten private industry supersectors. Top takeaway here is that nine of the state’s ten private industry supersectors likely shed jobs in 2025, while only education and health services experienced gains last year. This would follow a similar trend for the U.S. that I discussed last month. As shown in the next chart, private education and health services (green bar) increases in 2025 are expected to be around +15,600, while the other Colorado supersectors (red bars) should combine for a total loss exceeding -30,000 jobs. Professional and business services may lead those declines, dropping approximately 9,000 jobs last year. Information, manufacturing, and trade, transportation and utilities could all experiences decreases of at least 3,500 jobs in 2025.

Similar to a previous graphic, the next chart displays the expected year-over-year percent changes for Colorado’s private industry supersectors. The growth rate for private education and health services growth rate likely exceeded 4% in 2025, while the percent changes for the other nine supersectors could range between -0.6% (leisure and hospitality) and -4.9% (information).

Share

Finally, let’s discuss some potential changes to Colorado’s unemployment rate after benchmark revisions. As I mentioned in my first article back from Substack hiatus, “between June and December, Colorado’s labor force total dropped by nearly 23,600, or a growth rate of -0.72%.” I noted that a drop like that over a six-month span was extremely rare (only 12 prior instances since 1976) and was a key driver in Colorado’s seasonally adjusted unemployment rate significantly decreasing from 4.7% to 3.8% over that period.

The U.S., which has already undergone benchmark revisions, actually saw its seasonally adjusted labor force increase by +0.65% between June and December 2025. That is meaningful since Colorado historically outpaces the nation in labor force growth. For instance, the number of people in the U.S. labor force grew by 8.5% between June 2015 and June 2025, while the Colorado rate was nearly twice as fast at 16.6% (note: to be counted in the labor force, an individual either has to be employed or unemployed and active, available, and able to seek work). Given that historical trend, while not impossible, it is unlikely that Colorado would suddenly experience a reversal in labor force growth, particularly compared to the nation. Therefore, I’ll walk through a brief counterfactual scenario of what would happen to Colorado’s unemployment rate if the state’s change in labor force mirrored the U.S. between June and December of last year.

• Based on current, pre-benchmark estimates, June’s seasonally adjusted labor force total for Colorado is 3,283,000, while December’s total is 3,259,400.

• Increasing the June estimate by +0.65% would result in Colorado’s December labor force total being 3,304,300, or a gain of 44,900 above the current, pre-benchmark level.

• In this scenario we’ll assume that the additional number of 44,900 individuals would shift from not in the labor to unemployed status (rather than employed).

• That would result in Colorado’s number of unemployed persons (on a seasonally adjusted basis) in December increasing from the current, pre-benchmark level of 123,800 to 168,700.

• Finally, dividing 168,700 unemployed individuals by our new labor force total of 3,304,300 would result in a revised Colorado December seasonally adjusted unemployment rate of 5.1%, compared to the current, pre-benchmark rate of 3.8%.

I honestly have no idea how Colorado’s labor force and unemployment estimates will be revised in a couple weeks, as there are many factors that can influence these labor market components and the process is more difficult to predict than for CES. However, an unemployment rate around 5% near the end of 2025 would make more sense, especially given the expected decline in jobs highlighted above. Excluding the 2020 pandemic/2021 rebound, Colorado last had an unemployment rate of at least 5% in June 2014, following the slow and long recovery from the Great Recession.

As expected, the rise in oil price due to the U.S.-Iran War led to a sizable increase in the average midgrade retail gasoline price in Colorado, which surged to $3.82/gallon for the week ending March 9 compared to $3.11/gallon the week prior. That 22.7% gain in Colorado’s average midgrade retail gasoline price is the largest weekly change for the state since comparable data began in 2000. Below are other instances with weekly increases of at least 10%.

· March 9, 2026: +22.7%

· September 5, 2005: +15.9%

· March 7, 2022: +12.4%

· May 25, 2020: +10.9%

While the West Texas Intermediate (WTI) daily closing price ranged between $80 and $100 per barrel last week, thus exhibiting some relative restraint in growth, there’s plenty of evidence that points to elevated levels going forward. I recommend listening to this Odd Lots podcast for a great deep dive into the problem at hand and different potential pathways for global oil prices.

Share

Inflation measures and the price of goods not called oil or gasoline

Data for the two primary inflation measures in the U.S. were published last week. The Bureau of Labor Statistics (BLS) released the Consumer Price Index (CPI) for February, while the Bureau of Economic Analysis released the Personal Consumption Expenditures Price Index for January. Both headline statistics still remain well above the Federal Reserve’s preferred rate of 2% as part of their “dual mandate”. Along with various U.S. and sub-national indices, the CPI program also publishes “average prices for select utility, automotive fuel, and food items”. I thought it would be interesting to look at some of these prices over time, so this section highlights trends for the following goods: meat derived from cows; coffee; chicken breasts vs. eggs and oranges vs. frozen orange juice; and electricity. The following prices reflect only U.S., city average, data, as more granular levels of information are not available. These prices are also not adjusted for inflation.

Prices for ground chuck, round roast, and steak sirloin

This first chart shows the per pound price of ground chuck, round roast, and steak sirloin going back to 2006 (diagram of a cow and available cuts of meat available here). Steak sirloin, a desirable cut of meat, generally has a price 1.5-3.0x higher than ground chuck and round roast. In the six years prior to the pandemic, the price of steak sirloin was around $8-$9/lb., while in early 2026 the price has exceeded $14/lb. All three cuts of meat have experienced price increases of at least 16% over the past year.

Coffee prices

The next chart displays the U.S. city average per pound price for ground roast coffee since 2010 (prices were not reported for most of 2018 and 2019). In December 2019, just before the start of the pandemic, average prices were around $4/lb. Three years later the price had increased to nearly $6.50/lb., marking the first time surpassing $6/lb. since 2012. While coffee price stabilized between late 2022 and mid-2024, it has risen precipitously over the past 21 months. As of February 2026, the per pound coffee price approached $9.50, a 30% gain compared to the same month a year prior. Most of that recent rise in price can be attributed to tariffs in 2025, as coffee is grown within very limited portions of the U.S.

Disconnect in price changes between related food products

The following collection of charts highlight an interesting disconnect in price changes between food products that should feasibly be correlated, but are not. The first graphic shows the per pound price for a boneless chicken breast (green line) and the price for a dozen, large, grade A eggs (black line). As detailed within the chart, which spans between 2014 and 2026, there have been three separate instances of bird flu resulting in spikes in egg price (and subsequent declines). In 2015, the price of a dozen eggs jumped from $2 to $3 in just three months. The increase was significantly more dramatic in 2022, rising from $2 over $4.50 in ten months. However, that was quickly dwarfed by the change in price between June 2024 and March 2025, when a dozen eggs jumped from costing an average of $2.70 to $6.20. Since eggs come from chickens, I assumed that the price of chicken breasts over time might reflect those aforementioned spikes, but that’s not the case. An uptick in chicken breast prices did occur in 2022, but the peak occurred months before the peak in the cost of eggs. That leaves general post-pandemic price increases as the likely culprit, especially since there were no meaningful changes in chicken breast prices in 2015 or 2024.

The other comparison is between the price of frozen concentrate orange juice and oranges. Since 2018, the per pound price of navel oranges, when available through the BLS, has ranged between $1.20 and $1.80. In fact, the price has only increased by 12.6% between February 2018 and February 2026. However, the price of frozen concentrate orange juice, which is presumably comprised of oranges given its name, has risen by 91.6%(!) over that same eight year period.

Electricity prices

The final chart looks at the U.S. city average electricity price per kilowatt-hour (KWH) since 1978. That price did not exceed 10 cents until late 2005. Moving forward, the current price is approaching 20 cents, having grown rapidly between 2021 and 2025. For example, between September 2020 and September 2025, the electricity price per KWH grew from $0.137 to $0.188, an increase of 37%. While there have been other steep changes in electricity prices over a five-year period (for instance, a gain of 40% was experienced between September 2003 to September 2008), the recent activity has attracted a lot of attention. Causes for the increase in electricity price are diverse and nuanced; however, it’s difficult to not look past the parallel boom in data centers as the drive for AI-production intensifies. Per data from the BLS’ Quarterly Census of Wages and Employment program, the number of U.S. businesses in the data processing, hosting, and related services industry (NAICS 518120) have expanded from 30,620 in the third quarter of 2020 to 56,224 in the third quarter of 2025 – a gain of 84%. The majority of that growth in businesses is likely driven by data centers, which require and consume massive amounts of energy.

Oil prices

As has been widely reported, the United States has been engaged in a war with Iran since February 28. Given the region’s importance for the production and transportation of oil, there has been a lot of attention paid to the price of oil over the past week. While there are a few measures of oil prices, the following chart shows the West Texas Intermediate (WTI) daily closing price since February 20 (as dollars per barrel). From February 20 to February 26 the price ranged between $65.20 and $66.50, following a trend of relatively low and stable prices over the prior eight months. However, as rumors of a potential U.S.-Iran conflict accelerated, the price ticked up to around $67.00 to close out Friday, February 27. Between March 2 and March 6, with the war fully underway and duration unknown, the WTI price rose from just over $71 to nearly $91 to conclude the work week. The WTI futures price as of 10pm on Sunday, March 8 has exceeded $115/barrel.

To assess how this rapid rise in oil price (i.e. shock) compares historically, I looked at periods when the WTI Friday closing price had increased by at least 40% compared to the average Friday closing price of the prior 26-week period. The available data only goes back to 1986, but there have been five instances that meet this price increase threshold: the start of the Gulf War in August 1990; late May 2009, following the Great Recession; March 2021, following the pandemic and re-opening of global economies; March 2022, coinciding with Russia’s invasion of Ukraine; and last week.

Of these prior examples, the 1990 Gulf War and the start of the Russo-Ukrainian War in 2022 are most similar to the current disruption of oil prices due to geopolitical causes. However, it’s likely that the oil price shock due to the U.S.-Iran War could soon dwarf the increases experienced during the 1990 and 2022 conflicts. For instance, between February and August 1990, the average WTI price was around $19/barrel, compared to a peak of $40/barrel in September, nearly two months after the start of the Gulf War – a gain of 107%. Based on the 26-week average of around $62/barrel prior to the beginning of the U.S.-Iran War, an increase of 107% would equal a WTI price of $129/barrel – an amount that could be feasibly achieved within a couple days based on the accelerating pace of growth and no attempts to cap price increases.

Historically, oil prices rarely exceed $100/barrel, having first occurred in February 2008 and in only 111 of the past 941 weeks (rate of 11.8% during that period). Interestingly, trying to tie barrel prices to local retail/gas pump prices is not straightforward, as the two don’t always move in sync. That said, as displayed in the next table, Coloradans (and elsewhere, of course) will experience a steep increase in the amount they will be paying at the pump to fill up their vehicle. Based on data from U.S. Energy Information Administration (EIA), the weekly average midgrade (89 octane) retail gasoline price in Colorado was $3.11/gallon for the week ending March 2 (most recent currently available). Colorado midgrade gas prices have fluctuated between $2.50 and $3.50 per gallon over the past year, which would translate to around $30-$42 to fill up a completely empty 12-gallon tank. Many Denver gas stations are already showing prices that exceed $3.50/gallon and a $4/gallon cost is likely to happen soon. The last time the Colorado midgrade average exceeded $4/gallon was in October 2023, following an 18-month period of elevated gas prices (peak was $5.22/gallon in June 2022, three months after the start of the Russo-Ukrainian War). Other related indicators to watch in the coming weeks are diesel fuel costs and airline prices.

Luckily, the financial markets are taking this all in stride. Just kidding…the S&P 500 and Dow Jones Industrial Average declined by 1.2% and 2.6% last week, respectively, and may both open down around 2% Monday morning (based on Sunday night futures). Japan’s Nikkei has dropped by over 6% through the four hours of Monday trading on the other side of the world. Given the resilience of the markets over the past year, I can’t say I’m confident in predicting how they’ll react in the coming weeks, but it is certainly going to be more difficult for the U.S. to backtrack from a Middle East geopolitical conflict (see: 2001-2021) than tariff rates.

Prediction markets

Predictions markets like Kalshi and Polymarket have gained popularity and exposure over the past couple years and they appear to sit on that hazy line of “how is this legal?/is this even being regulated?/insider trading is totally a thing, right?” (I recommend this Ringer article from January for a deep dive). It appears that line was definitely crossed during the onset of the U.S.-Iran War on February 28. According to multiple news sources (links to: NPR and NY Times), there were a number of well-placed (re: suspicious) bets in Polymarket on war predictions that netted individuals notable earnings (this also happened in January). Given the recent inundation of advertisements and commercials for these “services” and the amount of money circulated (over $12 billion traded between the two platforms in December based on this NY Times article), perhaps additional (any?) regulation and guardrails could be implemented? One could hope…or is that just a sucker’s bet?

Concerning jobs report

The Bureau of Labor Statistics (BLS) released labor market data for February on Friday. The big news was an announced decline of 92,000 payroll jobs in February compared to January. While some of that decrease was driven by a healthcare strike in California, it represents a sizeable loss in jobs on top of a fairly weak 2025. Over the past 10 months, the total number of U.S. nonfarm jobs has dropped by 32,000. Over the month declines occurred in five of those 10 months, including December 2025, which was revised down from a change of +48,000 jobs in the January BLS employment situation release to a change of -17,000 jobs in this recent report.

Additionally, the nation’s seasonally adjusted unemployment rate rose one-tenth of a percentage point in February to 4.4%. Notably, while last month’s unemployment rate is identical to the rate in December, February’s labor force participation rate (the LFPR is the share of 16+ population employed or unemployed and able, available, and actively seeking work) is only 62.0% compared to 62.4% just two months prior. While that difference may appear small, consider the following. If February’s LFPR was 62.4% (meaning a shift of individuals from not in the labor force to unemployed status) that would increase the number of unemployed people last month to nearly 1 million. That change would bump up last month’s U.S. jobless rate to around 5.0%, a level last achieved in mid-2021.

Related to unemployment rates are statistics on duration of unemployment. The BLS splits duration into a number of broad ranges: less than 5 weeks; 5 to 14 weeks; 15 to 26 weeks; and 27 weeks and over. The 27 weeks and over category is also referred to as long-term unemployed (note: this unemployment data is derived from the Current Population Survey and not tied to those receiving unemployment insurance benefits). As presented in the chart below, the U.S. long-term unemployed as a share of total unemployed has been climbing over the past couple years and is now above a proportion of 25%. This indicator is important to track as it can reflect the difficulty individuals can have in regaining employment over a long period, particularly as job opportunities dry up in a weakening labor market.

Another unemployed group to watch during a subdued labor market are those who are entering the labor market for the first time, known as new entrants. These aspiring workers, with no prior experience, are going to be young and may also be engaged in educational pursuits like high school or college. Similar to duration of unemployment, the BLS has different broad categories for reasons for unemployment: job losers, job leavers, reentrants to the labor force, and new entrants to the labor force. The final chart shows that U.S. new entrants have recently exceeded a share of 10% of total unemployed, close to a peak rate over the past 30+ years. While some of this relatively high proportion may be driven by low rates of layoffs and fires (which would be reflected in the job losers category), it’s a concerning statistic for new jobseekers and a trend supported through other types of analyses.

Daylight savings

Let’s end on a slightly lighter note – the annual ritual of “springing forward” so that our clocks adjust an hour to take advantage of more daylight (except if you’re Hawaii, parts of Arizona, and most of the world). This exercise presumably affects short-term productivity, health, and sleep cycles, but on the plus side sunsets are now after 7pm. Anyways, for those interested, here’s an entertaining and educational six-minute YouTube video from 14 years ago aptly titled, Daylight Savings Time Needs to End.

The 30-year fixed rate mortgage average fell below 6%

Many news outlets reported that the U.S. mortgage rate dropped under 6% last week. While 6% isn’t a magical number, it does represent the first time in three and a half years that the rate has fallen below that threshold. As shown in the chart below, the 30-year mortgage rate in the U.S. has gone on a wild ride over the past 5+ years, moving from below 3% in late 2020, to nearly 8% by the fall of 2023, and steadily downward since that point. However, the rate is still elevated going back to 2010, so it’s not likely buyers will come rushing back anytime soon.

Mortgage rates can be influenced by the federal funds rate, but they don’t necessarily always move lockstep with each other. That said, the federal funds rate reached a recent peak of 5.33% in August 2023, with the 30-year mortgage rate hitting a high of 7.79% just two months later. Since then, the federal funds rate has declined to 3.64% as of January 2026, as the Federal Open Market Committee (FOMC) announced a series of rate cuts over the past two years. The number and magnitude of rate cuts in 2026 are uncertain due to several factors, and any action this month is unlikely.

Share

Additional statistics on Colorado’s private sector job loss since September 2023

In last week’s article, I mentioned that “[Colorado] has shed nearly 2,000 private sector jobs between September 2023 and September 2025, based on these QCEW employment estimates.” Since employment data is one of my areas of expertise and interest, I wanted to provide some additional statistics to highlight the fact the state has lost private sector jobs over the past two years.

First, that occurrence is pretty rare. Based on data going back to the early 2000s (and presented in the following table), Colorado private sector employment has decreased over a two-year span a total of 79 months. 77 of those 79 months coincide with national recessions, while the other two (June and September of 2025) do not take place during recessions.

Next, similar to U.S. data we recently reviewed, the healthcare and social assistance industry has masked broader job losses in Colorado’s private sector over the past two years. The chart below shows that between September 2023 and September 2025, Colorado’s private sector declined by 1,945 jobs. However, over that time, healthcare and social assistance gained 31,342 private sector jobs, while all other industries experienced a combined loss of 33,287 private sector jobs.

Last, the following tables show the Colorado industry subsectors with the 10 largest gains and declines in private sector jobs between September 2023 and September 2025. Four of the top five subsectors with increases over that two-year period are concentrated within healthcare and social assistance. The list of bottom 10 subsectors (largest losses) is diverse. Out of the 83 Colorado industry subsectors with private sector employment, 19 added at least 500 jobs over the past two years, while 25 shed at least 500 jobs. The Bureau of Labor Statistics will publicly release Q3 2025 QCEW data for all states on March 10, so it will be interesting to see how Colorado compares.

Government shutdown in October likely impacted GDP growth for the fourth quarter

On Friday, February 20, the Bureau of Economic Analysis (BEA) released GDP (gross domestic product) data for the fourth quarter of last year as well as the annual estimate for 2025. Top line, Q4 real GDP grew at an annualized rate of 1.4% (with the government component representing the largest drag as shown in the chart below), while 2025 increased at a rate of 2.2% compared to 2024. Per a technical note within the release:

“The full effects of the partial federal government shutdown on the fourth-quarter estimates cannot be quantified because they are embedded in the regular source data that underlie the estimates and cannot be separately identified. However, BEA did estimate the effects of a reduction in the labor services supplied by federal employees. BEA estimates that this reduction in services provided by the federal government subtracted about 1.0 percentage point [emphasis added] from real GDP growth in the fourth quarter.”

Trusting government statistics

Here, I just wanted to point to a good article from Claudia Sahm’s Substack on “Trust in Numbers”, which encourages investment into our statistical agencies, by protecting the integrity of the data and the institutions that produce it. Recommend checking it out!

Share

More tariff news

In last week’s article I mentioned that the Supreme Court may rule further on tariffs in the coming weeks and that came to fruition on Friday. By a 6-3 margin, the Supreme Court ruled that the President exceeded his authority in invoking the International Emergency Economic Powers Act (IEEPA) of 1977 to set tariffs on imported goods from more than 100 countries. Following the Supreme Court’s ruling, global tariffs of 10%, then changed to 15%, were announced. Interestingly, one push for a shift in U.S. tariff policy early last year was to help close the nation’s trade deficit. Recent data from the U.S. Census Bureau shows that the trade deficit was little changed in 2025 compared to prior years.

If you’re like me and having difficulty keeping up with tariff whiplash, here are some resources, analysis, and commentary I discovered over the past few days:

· Fantastic analysis from Joseph Politano’s Substack on the Supreme Court ruling and what comes next

· A day-of reaction from Paul Krugman in his Substack, as well as Paul Krugman in video-form on YouTube (5 minutes long)

· Podcast with the Ringer’s Derek Thompson and economist Jason Furman on “what happens now?” (39 minutes long)

· And while this isn’t tied directly to recent news, two researchers from Northwestern University just released their working paper on “The Macroeconomic Effects of Tariffs: Evidence From U.S. Historical Data”.

Colorado QCEW data for Q3 2025 shows continued job loss

Colorado’s Labor Market Information office released third quarter 2025 data from their Quarterly Census of Employment and Wages (QCEW) program on Wednesday. If you’re not familiar with QCEW, I’ve covered that dataset in prior articles (here and here), so I encourage you to look into those for some background information.

The newest QCEW data shows continued job loss for Colorado in the third quarter of last year, following a pattern of subdued growth since the beginning of 2024. The following chart displays the over the year change in the number of jobs in Colorado for all industries (private and public sector), from Q1 2023 to Q3 2025. Note that the state has shed jobs in three of the past four quarters, with those losses ranging between -9,800 (Q4 2024) to -13,000 (Q3 2025). This represents a drastic shift from 2023, when Colorado was adding an average of nearly 69,000 jobs compared to the year prior, and even to the first three quarters of 2024, when year-over-year levels slowed to around a 21,000 gain on average.

The data in the next chart focuses on Colorado’s private sector and measures the rate of change in jobs by month. Here we can see that the average growth rate for Colorado’s private sector over the past 21 months (January 2024 to September 2025) is 0%, with a range spanning from +0.8% (January 2024) to -0.9% (November 2024). While not shown in the chart below, the state has shed nearly 2,000 private sector jobs between September 2023 and September 2025, based on these QCEW employment estimates.

Finally, the last chart looks at information for Colorado at the industry subsector (3-digit NAICS) level. This graphic, which includes data from the third quarter of 2001 to the third quarter of 2025, presents the share of Colorado subsectors with over the year job loss for a given quarter (for reference, there are typically around 90 industry subsectors with published employment data). During the past four quarters (Q4 2024 to Q3 2025) more than half of Colorado’s industry subsectors have experienced a decline in employment compared to a year prior. That frequency of over 50% (shown in red) is notable as that has only previously happened in Colorado during recessionary periods. Additionally, the past four-quarter period is important as that data will be utilized to update Colorado’s 2024 and 2025 estimates during the annual benchmark revisions that will be publicly released in a month or so.

Consumer Price Index data for January released

The BLS reported that the consumer price index (CPI) increased by 2.4% in January compared to the 12 months prior, on a not seasonally adjusted basis. That rate of 2.4% represents a slight slowdown compared to prior months. However, the government shutdown in October had an immediate and short-term impact on CPI estimates, so it might be best to treat these figures with caution for a little while.

Tariff news

On February 11, the U.S. House voted to rescind tariffs imposed on Canada last year. The resolution, which passed by a 219-to-211 margin, may end up just being a symbolic measure, but does represent a notable congressional shift compared to last year. The Supreme Court may rule further on tariffs in the coming weeks, but the outcome is far from known. In related news, the Federal Reserve Bank of New York published an analysis on “who is paying for the 2025 U.S. tariffs?” They found that “nearly 90 percent of the tariffs’ economic burden fell on U.S. firms and consumers.”

Share

Selected highlights from the annual benchmark revisions of U.S. jobs data

I’ve selected four charts to capture the copious amount of new and revised data released by the BLS on February 11.

Every year, coinciding with the January data release, the BLS also publishes the changes to prior year’s employment and jobs estimates (check out the “overview” section in the provided link for technical, background information). While jobs data going back to 2021 were revised, the chart below shows the level change in U.S. total nonfarm employment for each month in 2024 and 2025 after the annual benchmark revision. For instance, 2025 total nonfarm employment estimates for every month were revised down by between 785,000 jobs (January) and 1,045,000 jobs (September). While historically a large revision, these figures are in-line with preliminary estimates the BLS published late last year.

There were four industry supersectors that had December 2025 downward benchmark revisions of at least 100,000, displayed in the next chart. The total number of U.S. jobs within trade, transportation, and utilities in December were revised down by 357,000, far outpacing all other industries. Leisure and hospitality had the second largest downward revision for December at 186,000 jobs. Professional and business services and manufacturing each had December downward benchmark revisions of just a little more than 100,000 jobs. Combined these four industry supersectors accounted for 74% of the U.S. December downward revision of 1,029,000 for all industries.

As previously mentioned, the benchmarked data is released along with January estimates. With the new figures, the U.S. added 359,000 jobs between January 2025 and January 2026, which translates to a growth rate of just 0.2%. However, there is a large difference when looking at the change in jobs between the private and public/government sector. Over the past year, the U.S. private sector has grown by 615,000 jobs (growth rate of 0.5%), while government employment has shrunk by 256,000 jobs (a decline of -1.1%).

The next chart highlights the over-the-year change in U.S. jobs by private sector industry group between January 2025 and January 2026. Healthcare and social assistance (+757,700) accounted for more two times (!!) of the 359,000 total private sector jobs added over the past year. Accommodation and food services (hotels, restaurants, and bars) has been the other significant contributor to U.S. private sector employment since last January, adding 140,200 jobs. And that’s the extent of the good news. The goods-producing industries (mining/logging, construction, and manufacturing), have combined to shed 58,000 jobs over the past 12 months. Wholesale/retail trade has experienced a total loss of 87,000 jobs, while transportation, warehousing, and utilities employment levels have declined by 107,700 jobs. The remaining private sector industries not listed (including: information; financial activities; business and professional services; private educational services; arts, entertainment, and recreation; and other services) have combined to lose around 30,000 jobs since January 2025.

Finally, the last chart looks at the over-the-year change in U.S. jobs by public/government sector industry groups. Job gains have been driven by local government, with an increase of 89,700 jobs in local government, excluding education, and 50,000 jobs added in local government, education (which is primarily concentrated within K-12 school districts). Slight job losses (-9,100) have been measured in state government, excluding education, with larger declines (-62,500) experienced in state government, education (this can include postsecondary universities and colleges). And the largest driver in the decrease of U.S. government jobs over the past year is within federal government. U.S. Federal government employment has dropped by 324,000 between January 2025 and January 2026. That is the 6^th largest 12-month decline in U.S. federal government employment going back to 1950 (larger decreases: June 2011 [-333k]; May 1991 [339k]; April 1981 [-343k]; May 2001 [-486k]; May 2011 [-545k]).

State-level benchmark revisions won’t come up until late March/early April, but watch this space for some of those takeaways.

Due to the short, partial government shutdown, the Bureau of Labor Statistics (BLS) had to delay their January employment situation release from Friday, February 6 to Wednesday, February 11. While a look at last month’s data is important, I’m probably more interested in how 2024 and 2025 figures will be revised, as that coincides every year with the January release. In September, BLS published estimates that showed employment was likely going to receive a sizeable downward revision – we’ll find out Wednesday what that officially looks like.

U.S. job openings fall in December

While also delayed a couple days, the BLS published December estimates from their Job Openings and Labor Turnover Survey (JOLTS). Notably, national job openings fell in December to around 6.5 million from 6.9 million in November (on a seasonally adjusted basis). That marks the lowest number of U.S. job openings since September 2020 and a 13% decline compared to a year prior. I’ll caution not to make too much of a one-month change, but the trend has certainly been moving in the wrong direction.

Layoffs picking up?

The global outplacement and executive coaching firm Challenger, Grey, & Christmas has been producing layoff reports for decades. On February 5 they reported that “U.S.-based employers announced 108,435 job cuts in January…the highest [total] for the month since 2009, when 241,749 job cuts were announced.” While 108k is a far cry from 242k, you typically want to avoid any comparisons to 2009 and the Great Recession when assessing layoff activity. Interestingly, the report noted that AI was cited for 7% of January job cuts, but that it’s “difficult to say how big an impact AI is having on layoffs specifically.”

Share

Bitcoin roller coaster

Bitcoin, a cryptocurrency, experienced a significant decline in value late last week, dropping nearly 20% between Wednesday and Friday morning. Since then, Bitcoin has recovered over 70% of those losses in price. However, when compared to a year ago, the value of Bitcoin is down 26% compared to a year prior. Over that same period, the S&P 500 is up +14%, NASDAQ +17%, and gold a staggering +70%.

The Super Bowl

Finally, the Super Bowl will be played this weekend in a matchup of the Seattle Seahawks and the New England Patriots. If you turn on the TV or any other entertainment option with ads, you’re likely to see about 30 commercials featuring a variety of online betting options. Per data from the Department of Revenue, the following amounts were spent on wagers in Colorado over the past three February’s on American football: 2023 ($27.6 million), 2024 ($29.4 million), and 2025 ($26.3 million). I’m guessing wagers this year will be in a similar range, but would have definitely surpassed $30 million if our beloved Broncos had made it to the Super Bowl. For a deep dive on sports wager data in Colorado, check out my article from last year. And with that – go Seahawks!

Share

New population/net migration data

On January 27, the U.S. Census Bureau released updated national and state population figures for the period of July 1, 2024 to July 1, 2025. The main takeaway is that the U.S. experienced a significant slowdown in population growth (+1.8 million, or +0.5%) during the past year, with the primary driver being the “historic decline in net international migration”.

From a Colorado perspective, I thought the State Demography Office (SDO) put together a nice release highlighting state-level trends. First, Colorado’s population is now over 6 million (the 5 million population threshold was surpassed in 2010). Second, Colorado’s population growth was slower than the U.S., coming in at a rate of only 0.4%. Third, population change is determined by three components: births, deaths, and net migration (which is the number of people moving into an area, minus the number of people moving out of that area). Colorado has historically been a state with relatively strong net migration. However, meaningful slowing in net migration has occurred in Colorado over the past year, as total net migration was just over +3,000. The chart below is from the SDO release and shows Colorado net migration totals for the domestic and international populations, which have both experienced significant changes over the past 5-10 years. The shift of domestic net migration over the past year to negative territory is certainly alarming; however, the SDO does caution that some of that decline could possibly be driven by international migrants.

Another U.S. Census Bureau source for evaluating state net migration trends comes from their state-to-state migration flows dataset, the most recent data being from 2024 and based on information from the American Community Survey. The SDO has also converted this data into a dashboard – certainly worth checking out. Job-to-job flows data from the U.S. Census Bureau also provide value in this space and can give insight on these trends before the official population data is released. I put together a lengthy analysis back in September 2024 utilizing the job flows data for those interested.

Labor market statistics for December

Also on January 27, the Bureau of Labor Statistics (BLS) released December state-level labor force and jobs data. The late 2025 government shutdown wreaked havoc for all series releases (as well as the underlying data), so it’s pretty incredible that the state December statistics were only delayed by a few days. One takeaway for Colorado is the rapid decline in the unemployment rate since summer, moving from 4.7% in June to 3.8% in December. In fact, Colorado is now tied for the 21^st lowest unemployment rate in the nation, after spending most of 2024 and 2025 in the bottom third. However, I would caution about celebrating that as completely positive news, as the decline in Colorado’s unemployment rate has been driven by a subsequent decrease in the state’s labor force totals (note: to be counted in the labor force, an individual either has to be employed or unemployed and active, available, and able to seek work). Between June and December, Colorado’s labor force total dropped by nearly 23,600, or a growth rate of -0.72%. Going back to 1976, there have only been 12 other instances when Colorado experienced a decrease in the labor force of at least -0.72% over a six-month period (detailed below). That said, Colorado’s labor market statistics will undergo routine, annual revisions over the next two months, so it will be interesting to see how much the current data and trends change. I anticipate that the revised data will be released sometime between mid-to-late March.

· 1987 (2x), during a regional recession

· 1989 (1x)

· 2009 (4x), during the Great Recession

· 2020 (5x), during the COVID Pandemic

Nominations for crucial economic positions

Finally, two nominations for economically-important roles were announced late last week: Brett Matsumoto for BLS Commissioner and Kevin Warsh for Federal Reserve chair.

I wrote at length about the firing of former BLS commissioner Erika McEntarfer in August, and in the interim a highly unqualified individual was nominated, and then withdrawn from consideration. The new nominee, Brett Matsumoto, has worked in the BLS for a decade and seems to be a good fit for the role (anytime someone is described as a “data nerd”, I’m on board). He still has to go through a confirmation process and there’s likely concern about further politicization of BLS statistics and procedure, but it is at least a step in the right direction.

Regarding the nomination of Kevin Warsh for Federal Reserve chair, I’ll point to a Substack article Claudia Sahm recently released for some useful insight. This position also has to go through a confirmation process.

While many individuals and entities have expressed their opinions on this matter over the past two days (for example: Paul Krugman, former BLS Commissioner William Beach, the National Association for Business Economics, Economic Policy Institute, and the American Economic Association), I thought it might be useful to add some additional context based on my expertise as a labor economist and knowledge of the BLS and their numerous datasets.

Share

First, it should be noted that the BLS Commissioner is a politically-appointed position, and can removed from the position at the President’s discretion. Commissioner McEntarfer was nominated by President Biden in mid-2023 and confirmed by the U.S. Senate in early 2024 by a vote of 86-8. I met and conversed with Dr. McEntarfer many times when she worked as head of research for the Longitudinal Employer-Household Dynamics program at the U.S. Census Bureau, and I can say that she is incredibly smart, very nerdy, and extremely passionate and knowledgeable about data – basically all of the main qualities you’d want from someone leading the BLS. However, while the position is politically-appointed, the role and the BLS operate in a strictly apolitical fashion, with the primary mission of the organization centered on “measur[ing] labor market activity, working conditions, price changes, and productivity in the U.S. economy to support public and private decision making.” Put another way, if you were to ask the BLS if they see a glass as half full or half empty, they would reply with, “we see an 8-ounce glass containing four ounces.”

So with that preamble out of the way, I’d like to spend the rest of this article briefly describing some factors that impact BLS estimates, as well as explanations for why the revisions for May and June payrolls aren’t that surprising and certainly don’t reflect any efforts to manipulate or distort.

Budget

Considering all of the information it collects, calculates, reports, and disseminates, the BLS operates on an insanely small budget. After adjusting for inflation (2025 dollars), the BLS was appropriated $768 million during the 1999-00 federal fiscal year (FFY). That amount peaked at just over $900 million in the 2010-11 FFY, and is now estimated to have hovered around $700 million over the past couple fiscal years. Based on the last FFY with actual appropriation data (2023-24), the BLS experienced a real decline of approximately 14% in their budget compared to 20 years prior. Not only does the BLS budget go towards paying their national and regional staff and supporting the various statistical programs, but it’s also used to allocate dollars to Labor Market Information (LMI) offices within all 50 states and U.S. territories, that act as BLS “agents” as part of a federal-state partnership. Essentially, a large proportion of LMI budgets are federally-funded by the BLS, specifically as it relates to the operation of programs like Occupational Employment and Wage Statistics (OEWS), the Quarterly Census of Employment and Wages (QCEW), Current Employment Statistics (CES), and Local Area Unemployment Statistics (LAUS).

In addition to budget, the BLS has experienced a decline in their full-time employee count over the past decade, which has probably accelerated in 2025 due to efforts to drastically reduce the federal workforce. There are likely less than 2,000 full-time employees working for the BLS presently (national and regional staff, does not include federally-funded LMI employees). To put that number into context, there are several Colorado state government departments with more employees than the BLS.

On top of reduced budget and staff, the volume of information and data the BLS has to maintain, analyze, and evaluate becomes larger each year. For example, the number of U.S. business establishments in 2004 was 8.4 million, compared to 12.1 million in 2024 (an increase of 44%). U.S. employment has risen by 26% over that same period (from 129.3 million to 154.9 million), while the labor force has increased by over 20 million. Consider too that the BLS produces labor market data on nearly 400 Metropolitan Statistical Areas and over 3,000 counties, among other geographic distinctions. Much of this data is being produced on a monthly basis.

It’s also not cheap to employ individuals who specialize in highly technical data analysis and maintenance of that information. Using the BLS’ own data from the OEWS program, here are the 2024 U.S. median wages of some occupations you’d likely see working within the Bureau: statisticians ($103,000); mathematicians ($121,680); data scientists ($112,590); economists ($115,440); database and network administrators and architects ($107,480); and computer and information research scientists ($140,910).

Subscribe now

Response Rates and Other Measurement Issues

The BLS relies heavily on surveys in order to estimate labor market activity, pricing data, and other metrics. For instance, the unemployment rate is derived from a survey of approximately 60,000 households each month, called the Current Population Survey (CPS). The monthly payroll estimates from the Current Employment Statistics (CES) program primarily come from an establishment survey of over 600,000 worksites across the nation. Large surveys are not inexpensive to effectively manage and follow-up requires effort and organization, which can be significantly hampered when operational budgets remain flat or decline over time.

The figure below shows the response rates for the CPS and establishment survey from April 2015 to April 2025 (data available via the BLS if interested). Notably, the response rates for both surveys have dropped substantially since the height of the pandemic in 2020. From 2015 to 2019, the CPS response rate ranged from about 88% to 82%, while the establishment survey response rate averaged around 60%. Flash forward to 2025 and the response rates for both surveys are visibly lower – a little under 70% for the CPS and less than 45% required for payroll estimates. Basically, it has become incredibly difficult to get responses to surveys when compared to pre-pandemic, which adds an additional layer of complexity and volatility to these monthly estimates.

Other data series from the BLS have also recently been impacted by budgetary constraints. Within the past couple months, the BLS has announced reductions to the scope of the Producer Price Index and Consumer Price Index, both which are valuable in assessing how costs have changed over time. Additionally, in 2024 there were serious discussions about reducing the sample size of the CPS to 55,000 households, but luckily that was tabled (although the likelihood of that conversation being revisited is probably high). While some of these impacts and decisions are unavoidable for the BLS, they can have a profound influence on the estimates.

BLS Transparency

The BLS is, and always has been, highly transparent in their methodologies and estimations process. While it may not be fully accessible (i.e. easily understandable by a layperson), the information is publicly available for anyone to view and digest. Much of this is provided in the Handbook of Methods, which go into great detail about each BLS program (for example: CES, CPS, LAUS, OEWS, and QCEW). The BLS also produces many articles on a wide range of topics that can dive into research, trends, or methodologies. Basically, anything you’d ever want to know about the BLS, their data, and the processes utilized for generating estimates is in plain sight.

Monthly Revisions

Finally, every BLS employment situation report has an initial payroll estimate for the previous month (in this case, July), along with a second and third estimate (June and May, respectively), otherwise known as revisions. These revisions primarily “result from additional reports received from businesses and government agencies since the last published estimates.” To piggyback off of the transparency section above, the BLS publishes these revised data for nonfarm payroll employment going back to 1979. The following chart shows the number of U.S. payroll jobs added each month so far in 2025. January through April had no changes to employment levels in the report that was released August 1, as those months had already undergone three rounds of payroll estimates. The monthly change for May was revised sharply downward from +144,000 to +19,000, with a similar impact to June (+147,000 to +14,000). The July estimate was +73,000 jobs added compared to June. While not shown in the chart, it’s worth mentioning the job changes for March received substantial downward revisions between the first and third estimates (1^st, published in the April report: +282,000; 2^nd, published in the May report: +185,000; 3^rd, published in the June report: +120,000, or a decrease of 57% vs. the 1^st estimate).

Relative to earlier in the year (average of +122,750 per month), 2024 (average of +167,700 per month), and historically, these new and revised estimates may seem like a dramatic and unanticipated change, but for anyone actually paying attention this was fairly predictable.

• As covered previously, budget is a longstanding problem for the BLS. The issue has become exacerbated as response rates have significantly declined post-pandemic, which adds volatility to the monthly estimates. Large revisions (i.e. 2^nd and 3^rd estimates) have become more commonplace as these series struggle to fully capture labor market activity on the first pass.

• Trade policy has been upended by the U.S., which has now switched to a tariff-heavy model. While this seemingly fluctuates on a daily basis, the effective tariff rate was around 18% on August 1, compared to under 3% earlier in the year and over the prior two decades. In fact, at present the effective tariff rate is the highest since 1934. These rates really started to accelerate in early April, and have more or less stayed above 10% since then. Businesses historically dislike change and uncertainty, and this was reflected in employer confidence surveys at the time. It’s therefore not surprising that hiring and job growth would slow down significantly immediately following this period.

• Alongside trade, U.S. policy on immigration has significantly shifted. If that longstanding supply of labor is drastically reduced, then that will also have a meaningful impact on job creation and the ability for the nation to experience employment gains.

• For a number of different reasons the nation’s labor market was able to rapidly recover from the effects of the pandemic. Remarkably, in 2024 80.7% of U.S. individuals aged 25 to 54 (known as prime working age) were employed. That marked only the 9^th time in history when that ratio exceeded 80% for the U.S. (1996-2001, 2019, and 2023). Given that the peak rate is 81.5% (2000), it’s difficult to see how it could have increased much further in 2025 and future years. Essentially, full employment has been achieved by the prime working age population, which places an additional cap on potential job growth.

The U.S. Federal Statistical System is the envy of the world and is emulated by numerous countries. It’s crucial that agencies that oversee statistics and the creation and distribution of data remain independent, unbiased, and can operate free from the political apparatus. The BLS is the gold standard for producing quality labor market data and has always functioned within a framework of transparency, accuracy, and integrity.

Share

As shown in the first chart, Colorado added only 14,700 jobs last year, a steep drop-off compared to an increase of nearly 68,000 in 2023. Over the last 20 years that is the lowest gain in the number of jobs for the state in a non-recessionary year (bars shown in green).

The change in jobs on a year-over-year basis can be converted to growth rates, which are displayed in the following chart. The solid, dark blue bars represent years in which Colorado’s QCEW job growth was at least 1.5%, and the pixelated bars are years that fell below that mark (2008-2010, 2020, and 2024). While 2008-11 (Great Recession) and 2020 (Pandemic Recession) overlap with known U.S. recessions, Colorado’s relatively weak growth rate of 0.5% in 2024 occurred during an expansionary period for the nation. In 2023 the state’s growth rate was 2.4%, very similar to the years leading up to the pandemic.

One way to isolate drivers of Colorado’s weak job growth in 2024 is to look at trends by ownership type. Simply, ownership type can be separated four ways: federal government, state government, local government, and the private sector. The next chart shows the change in Colorado jobs for 2023 and 2024 by ownership type (federal government = blue bar; state government = red bar; local government = green bar; and private sector = purple bar). Federal government jobs in Colorado grew by 1,900 in both 2023 and 2024, so that ownership type had no effect on the state’s relatively weak job growth last year. State government employment gains slowed slightly in 2024 (2,900 vs. 3,800), while the difference in local government was more notable (7,100 vs. 13,500). However, Colorado’s private sector experienced a massive shift in job growth last year, registering only 2,800 jobs added compared to a change of nearly 50,000 in 2023.

Mirroring a prior chart, Colorado’s private sector job growth over the past 20 years is presented below. The state’s private sector growth rate of 0.1% stands out as historically weak, trailing only 2009, 2010, and 2020. Job growth in 2023 was 2%, marking the 15^th time since 2005 Colorado’s private sector reached that threshold.

When breaking out Colorado’s year-over-year private sector job growth over the past eight quarters, two significant shifts become quickly obvious. First, the change in the growth rate between the fourth quarter of 2023 (1.9%) and the first quarter of 2024 (0.4%) represented a rapid slowdown for the state. Second, job growth for the fourth quarter of 2024 turned negative (-0.7%), meaning job loss, rather than relatively slow job gains.

The final chart shows how Colorado’s QCEW private sector job growth ranks compared to all other states over the past 20 years. In 2024 the state’s growth rate of 0.1% ranked 45^th nationally (tied with Washington and Iowa) and only lead Illinois (0.0%), Oregon (-0.1%), and Massachusetts (-0.2%). The rank of 45^th last year matched Colorado’s relatively weak national performance in private sector job growth from 2010; however, the U.S. was deep in a lengthy recession at that point. 2009 and, notably, 2023 were the only other years over the past 20 years in which Colorado’s private sector job growth ranked worse than 20^th.

Subscribe now

To further understand the drivers behind Colorado’s weak private sector job growth in 2024, it’s important to look at the different industry sectors within the state. The table below displays the nineteen 2-digit NAICS industry sectors in Colorado with private employment. The third column in the table shows the change in the number of jobs for each industry sector in 2024, while the fourth column represents the related growth rate. Private health care and social assistance clearly added the most jobs last year (10,537) and was the only industry with gains exceeding 3,000. Job loss, however, occurred in 11 of the 19 industry sectors, led by administrative/support services (-5,189). Accommodation and food services (-1,061), retail trade (-1,304), manufacturing (-2,399), and information (-3,621) all experienced employment declines of at least 1,000 last year. Professional, scientific, and technical services, long a key driver for Colorado job growth, lost jobs in 2024, marking the first time that has happened since 2010.

Finally, a strength of the QCEW data is the ability to analyze more granular level of industry detail available than other BLS employment series. The next table digs deeper into industries with employment declines last year, specifically looking at the 16 detailed (4-digit NAICS) Colorado industries that shed at least 600 jobs in 2024. The first three columns of the table identify information about how each of the 16 industries is classified, while the fourth column shows how much employment declined for each of those industries in 2024. Overall, job loss between these industries was spread through multiple broad sectors of the state, touching retail trade, manufacturing, tech, finance and insurance, restaurants, and more. Employment services and computer systems designs and related services each lost over 3,000 jobs last year. Combined, these 16 detailed industries shed over 23,000 jobs in 2024.

Note: Late last year the BLS temporarily suspended the publication of Colorado’s QCEW data, but announced the resumption of that data publication in February. While the specific reasons for these series of decisions are unknown, the analysis contained within article is treating all 2024 QCEW data for Colorado as an accurate representation of the state’s labor market and related job growth, since “quality concerns with Colorado data have been sufficiently addressed”, per the BLS. That said, there are atypical shifts in the number of Colorado establishments over the past year. For instance, private sector establishments in Colorado dropped from 256,379 in the fourth quarter of 2023 to 232,818 in the fourth quarter of 2024, a decline of 9.2%. It’s unknown if economic factors are causing that change.