July real income and spending rebound, keeping the expansionary trends intact

 

 - by New Deal democrat

Last month I concluded: “This was a very weak report, although not negative. If there were not distortions from the front-running of tariffs earlier this year, it would come very close to meriting a ‘recession watch.’  But by next month, the payback from this previous front-running should have largely abated. If there is a rebound, needless to say that will be good. But if the weakness persists, we may cross the threshold. As it is, because of this uncertainty a ‘yellow flag’ caution, ie., pay extra close attention, is merited.”

This month we indeed got the answer. The “payback” for front-running did abate, and there was a rebound across all the data, meaning the yellow flag can be removed for now.

Nominally income rose 0.4% and spending 0.5%. Since the PCE inflation gauge only rose 0.2%, real income increased 0.2% and real spending rose 0.3%. Real personal income is below only April’s outlier, while real spending set a new record.:

[Note: with the exception of the personal saving rate, and one YoY graph, all of the data in the below graphs is normed to 100 as of just before the pandemic.]

Since real spending on services (blue, right scale) rarely turns down, even in recessions, the focus is on goods (red, left scale). In July they rose a strong 0.9%, also to a new record high:

Additionally, there is authority for the fact that spending on durable goods usually cools risk before spending on non-durable goods. In July, this rose 2.0%, but the absolute level remained below that of March and April, as well as last December:

While this is positive, the YoY trend shows considerable deceleration, as at 3.2%, it is no better than the level last summer:

One way to differentiate between noise and trend is to look at the YoY comparisons.In general while real income and spending do not turn negative YoY until after a recession has started, usually they have declined 50% or more from their YoY peaks within the previous 12 months (e.g. a decline from up 4.0% YoY to being up 2.0% YoY). After the improvements this month, total real spending as well as real spending on goods are nowhere near this threshold, but as shown above real spending on durable goods is very close, as measured on a three month average basis,  it is down -49% from +6.2% YoY to 3.2% YoY.

Next, here is the personal savings rate. I follow this because just before and going into recessions it tends to turn up as consumers get more cautious. In July it remained steady at 4.4%, in line with its typical reading this year:

Finally, let’s take a look at two coincident indicators from this report which the NBER pays close attention to in dating recessions. First, here is real income less government transfers:

This rose 0.3%, reversing the declines in the two previous months. It is now higher 1.6% YoY, and less than -0.1% below its April all-time high. Last month it was on the cusp of being recessionary. With this month’s improvement, the trend this year has stabilized, although over a longer time period, the deceleration that started last year has persisted.

Second, here is real manufacturing and trade industries sales, which is delayed one month and so if for June:

This rose 0.3% for the month, and the upward trend since 2022 is intact, although this remains below April and May levels. 

On an YoY basis this is also up, by 3.1%, in line with the trend in 2024 before tariff front-running came into play:

While both coincident indicators are below their spring peaks, I am not concerned that they signal a present recession, since the leading indicators would turn down first, and they have not done so.

In conclusion, as I said above, last month’s yellow caution flag can be removed, as all components rebounded, some to new all-time highs. There was no consumer retrenchment in savings, and real goods sales improved as well. The only ambiguous measure was real spending on durable goods, which typically is the first area where a downturn would be seen, and even there the more pessimistic reading is flatness rather than down.

Jobless claims suggest continuing, but resolving seasonality issues

 

 - by New Deal democrat

Initial jobless claims have been relatively subdued for the past month. Some have suggested that this is a byproduct of the large immigrant deportations which have recently taken place. Last week I noted that, “comparing the SA and NSA readings in the past several months suggests that [unresolved seasonality] has affected initial claims as well, with markedly fewer claims at the early July peak. But whereas NSA claims continued to decline through August last year, for the past three weeks this year they have held steady.”

This week’s report did not resolve the issue, but tilted it slightly back towards the unresolved seasonality argument.

Initial claims declined -5,000 to 229,000, while the four week average increased 2,500 to 228,500. With the typical one week delay, continuing claims declined -7,000 to 1.954 million:

It is interesting that during the recent weeks in which initial claims have declined, continuing claims have if anything drifted higher - again, suggesting that it is not a lack of immigrants making (initial, and then continuing) claims that is the main driver.

On a YoY basis, which is more important for forecasting, both the one week and four week average of initial claims were down -1.3%, while continuing claims were higher by 4.5%:

Note that in both 2023 and 2024, initial claims were declining all during August. This year on a SA basis they declined in late July and have instead trended higher during August. My suspicion is that within the next two weeks both the one week and four week average for initial claims will revert to being higher YoY once again. We’ll see.

Next week we will get the jobs report for August. Since jobless claims generally lead the unemployment rate, here’s the latest on what that comparison looks like, as YoY% changes:

One year ago the unemployment rate was 4.2% in August, declining to 4.1% in September. Last month it was also 4.2%. Initial claims suggest that it will remain 4.2% or perhaps decline to 4.1% in the next month or two, while the less leading but more comprehensive total of initial+continuing claims suggest it may rise to 4.3%.

Manufacturers’ new durable goods orders is a bright spot

 

 - by New Deal democrat

No new significant economic data today, but yesterday we did get a look at manufacturers’ orders for durable goods.

This is one of those noisy indicators that sometimes is leading (2000), sometimes is not (2007), and sometimes gives false positives (2016 and 2019), but has historically been considered a leading indicator for the economy:

Yesterday’s report for July showed a -2.8% monthly decline, which still kept the value higher than at any point before May. Since much of the volatility has to do with airplane orders (Boeing), the core capital goods measure, which excludes aircraft and defense, is typically more important. This rose 1.1% monthly, to the highest level since late 2022:

The general trend this year remains positive, which is a good - if surprising - thing; particularly as spending on goods, especially durable and consumer goods, is an important short leading indicator I am watching more closely since I have gone on “Recession Watch.”

Because so many durable goods are imported, real consumer spending on such goods has risen much faster than domestic production of such goods, so the best way to compare the two is YoY, which is shown in the graph below:

There has been some marked deceleration in real consumer spending on durable goods in the past few months, after the tariff front-running of earlier this year.

Real consumer spending will be updated on Friday. Will it resume its prior uptrend, or continue below the levels of earlier this year?

Repeat home sales indexes provide final confirmation that the housing market is in deflation

 

 - by New Deal democrat

Last week YoY existing home prices increased only 0.2%. Since those were not seasonally adjusted, it was apparent that the actual peak in such prices was about early this spring. Yesterday we saw that new home prices continued to deflate. This morning’s repeat home sales reports from the FHFA and S&P Case Shiller are the final confirmation that the housing market is in deflation.

On a seasonally adjusted basis, in the three month average through June, the Case-Shiller national index (light blue in the graphs below) declined -0.3%, while the FHFA purchase index declined -0.2%, matching their declines from the previous report. The FHFA index (blue in the graphs below) has now been declining for three straight months, and the Case-Shiller Index (gray) for four. the third. (note: as per usual, FRED hasn’t updated the FHFA information yet):

In other words, there has been actual *de*flation in the house price indexes since February, by -0.7% in the FHFA Index and -1.2% in the Case Shiller Index:

On a YoY basis, price gains in both indexes not only continued to decelerate, at 1.9% for the Case Shiller index, and 2.7% for the FHFA index; but these were the lowest YoY% increases since 2012 for both indexes excluding 5 months in 2023 for the Case Shiller index:

Because house prices lead the shelter component of the CPI by 12 - 18 months, this strongly indicates that they will continue to decelerate over that period. Here is the same graph as above (/2.5 for scale) plus Owners’ Equivalent Rent from the CPI YoY (red):

The last time the Case-Shiller and FHFA Indexes were in this range, excluding the Great Recession, was in the 1990s, during which time Owners Equivalent rent was in the 2.5%-3.5% range (vs. 4.1% as of the most recent CPI report).

Similarly, although I won’t bother with the graph, the latest “National Rent Report” from Apartment List from the end of July showed precisely zero YoY rent increases.

As of now, all phases of the housing market are either at or near their low points (sales, permits, starts), or declining (prices, construction, employment, and new spec units for sale). After a brief spurt earlier this year, for the past few months industrial production has also been flat. Before 2000, when both goods producing sectors of the US economy were flat to negative, the economy contracted promptly. On Friday, we will find out if consumer purchases of goods are flashing a warning (or not) as well.

New home sales: the final shoe in the housing sector has dropped

 

 - by New Deal democrat

In this month’s report on new home sales for July, the most important news was at the tail end, which I’ll get to last.

As per my usual intro, while new home sales are the most leading measure of the housing market, they are very noisy and heavily revised - which turned out to be important this month - and which is why I generally pay more attention to single family permits. Still, if averaged over three or more months they are valuable indicators of the underlying upward or downward pressure on the economy going forward one year or more. Further, as per usual, sales turn first, followed by prices and inventory, which is typically the last shoe to drop.

So let’s turn to each metric in order.

With mortgage rates remaining in the 6%-7% range, sales of both new and existing homes have also been rangebound for over two years. In July that continued to be the case, as new home sales declined -4,000 (from a June level upwardly revised by 29,000!) to 652,000, near the bottom of that range: 

After sales peaked, prices also stalled, and then began a very slow deflation on the order of -1% -5% YoY. That trend not only continued but amplified in July, as on a non-seasonally adjusted basis prices declined -$3,400 to $403,800 (gold, right scale in the graph below). More importantly, on a YoY basis prices declined -5.9%, the steepest such decline since last November (blue, left scale):

Recall that last week the median price for existing home sales was only up 0.2% YoY. Tomorrow we will get both the FHFA and Case Shiller repeat sales indexes, which also have shown recent, seasonally adjusted, declines. If that continues, it will be the steepest such retreat since the Great Recession’s housing bust.

But the most important news was at the tail end. Last month’s initial report indicated that the inventory of homes for sale had risen to 511,000, a new post-pandemic high.

Revisions made a major difference this month. The data now indicates that the inventory of new homes for sale in fact peaked in March at 504,000, with both May and June being revised down, the latter by -9.000 to 502,000. And this month inventory for sale was reported declining another -3,000 to 499,000:

Why is this so important? Because in the past recessions have happened after not just sales decline, but the inventory of new homes for sale (red, right scale) - which also consistently lag - also decline, as builders pull back:

To reiterate: in the typical housing cycle mortgage rates lead sales, which slightly lead permits and starts, which in turn lead prices and housing units under construction (gold in the graph below), which finally lead employment in residential construction (blue) and housing for sale (red):

For the record, I pay very little attention to months’ supply, becuase it depends on both units sold and for sale. There does not appear to be any “magic level” that serves as a turning point, and indeed, it frequently only turns down after a recession has begun, as homes for sale decline even more than homes sold:

So, the important conclusion is: all of the important sequence of metrics in the housing sector have now turned down. The final shoe has dropped. Now we pay more attention to short leading indicators like major durable goods purchases and initial jobless claims, for signs that the turn in the bigger economy is near.

Weekly Indicators for August 18 - 22 at Seeking Alpha

 

 - by New Deal democrat

My “Weekly Indicators” post is up at Seeking Alpha.

There were no big changes this week, but what continues to stick out in the data as far as I am concerned is just how strong consumer spending continues to be: weekly retail spending was up nearly 6% YoY, and restaurant reservations - a very easy thing to cut back if consumers feel pinched - are up 10%. That simply is not compatible with a big slowdown.

As usual, clicking over and reading will bring you up to the virtual moment as to the state of the economy, and bring me a penny or two in my pocket for organizing the data for you.

The rebalancing of the housing market continues, as existing home price increases have halted, and inventory finally exceeds 2020 levels

 

  - by New Deal democrat

The housing market, for all of its economic importance, tends to move as slow as molasses. This is especially true as to prices, where sellers are loathe to realize a loss, even when compared to hypothetical gains they could have had by selling earlier.

The pandemic, of course, through the entire market out of whack, since there was a period of time that it was for all intents and purposes shut down. It is only now rebalancing.

After the Fed began hiking rates in 2022, mortgage rates also rapidly rose from 3% to the 6%-7% range, where they have remained ever since. Since sales follow mortgage interest rates, existing home sales rapidly declined to 4.0 million annualized, and have remained in that range, generally +/-0.20 million for the past 3.5+ years:

In July, the rangebound behavior continued, with sales of 4.01 Million annualized (blue, right scale). Note that new home sales (gray, left scale) similarly declined and have similarly stabilized in the 625,000-725,000 annualized range.

The trend I have been looking for in the past several years is the rebalancing of the new and existing homes markets. Existing home inventory has been removed from the market for over 10 years (likely due in part to absentee rental owners buying increasing chunks of inventory), and really accelerated during the pandemic. This caused an acute shortage of houses for sale, which in turn led to bidding wars among buyers and a spike in prices.

A rebalancing of the market more than anything would require an increase in inventory at least to pre-COVID levels, and a deceleration of price increases, or even outright decreases. Which means that the level of sales themselves was far less important than what the median price for an existing home and inventory are telling us about the ongoing rebalancing of the housing market.

The secular decline in inventory reached a nadir in 2022. This series is not seasonally adjusted, so it must be looked at YoY. In July inventory increased by only 1,000, but more importantly, inventory finally exceeded its 2020 level for the same month, up 5,000:

Still, pre-2020, inventory was typically in the 1.7 million to 1.9 million range, which means that although it is lessening the chronic shortage still exists.

But even more important is what happened, and has continued to happen, with prices. As shown in the below graph, the average price of a new home (gray, left scale, not seasonally adjusted) rose almost 40% between June 2019 and June 2022 before slowly declining about -7% through June 2025. Meanwhile, the average price of an existing home (blue, right scale, not seasonally adjusted) rose about 45% between July 2019 and July 2022 and another 5% through July of this year, as was reported yesterday:

With seasonal adjustments are not made, my rule of thumb is that a peak (or trough) occurs when the YoY% change is less than half of its maximum change in the past 12 months. Here are the comparisons in the past 12 months:

July 4.2%

August 3.1%

September 2.9%

October 4.0%

November 4.7%

December 6.0%

January 4.8%

February 3.6%

March 2.7%

April 1.8%

May 1.3%

June 2.0%

As of yesterday’s report for July, the YoY% change in average prices was only 0.2% higher than one year ago.

Last month I concluded with “I still expect moderation in price increases and more importantly, for inventories finally to exceed their 2020 levels.”

This month, both happened. Inventory finally exceeded 2020 levels, and further, it is safe to say that if we had seasonally adjusted measurements, we could conclude that prices for existing homes peaked sometime this spring, and have started to decline.

Even so, prices of existing homes are still up about 50% from 2019 levels, vs. new single family homes, which are up less than 30%. Which means that while the July existing home sales report confirmed the ongoing rebalancing of the market, there is still some distance to go.

Jobless claims suggest our recent good news has been more unresolved seasonal quirks

 

 - by New Deal democrat

Initial jobless claims have been plagued by apparent unresolved seasonality in the past several years, post-pandemic. I suspect that is still playing out, as evidenced by the past few weeks of claims.

Initial claims rose 11,000 to 235,000 last week, a seven week high. The four week moving average rose 4,500 to 226,250, the highest in four week. Meanwhile, with the usual one week delay, continuing claims rose 30,000 to 1.972 million, the highest since early November of 2021:

Last week I speculated that the steep decline in initial claims in July might have been due to seasonality issues around layoffs in the education sector, or might be a side effect of large scale deportation raids in some sectors. This week’s report makes me think it is more likely the former than the latter.

To show you why, here is the four week seasonally adjusted average since spring 2022 (blue), together with non-seasonally adjusted initial claims, averaged biweekly (red):

On a non-seasonally adjusted basis, initial claims always peak in January after the Holiday season, with a secondary peak when the school year ends in June. They make their lows around Labor Day, as the new school year begins. In the last several years, the seasonal adjustment has given us low readings in January (i.e., fewer layoffs than was usually the case pre-pandemic), but elevated readings in June and early July at the end of the school year, gradually declining through autumn.

This year the June employment report strongly suggested that end of school year layoffs were slightly askew compared with the last several years. Comparing the SA and NSA readings in the past several months suggests that has affected initial claims as well, with markedly fewer claims at the early July peak. But whereas NSA claims continued to decline through August last year, for the past three weeks this year they have held steady.

We’ll see if that continues to be the case in the next few weeks.

Returning to our regularly scheduled analysis, here are the YoY% changes that are more important for forecasting purposes. Initial claims were 1.3% higher than one year ago, while the four week average was down -3.9%. Continuing claims were higher by 6.1%:

This suggests that layoffs remain subdued, while hiring has seriously slowed down, but not enough to suggest that a recession is close at hand.

Finally, we’re far enough along in the month to take a look at the implications for the unemployment rate. The below graph looks at all three metrics by YoY% change:

This suggests that the the unemployment rate will remain very close to its 4.1%-4.2% of one year ago.

Real nonsupervisory payrolls and income in danger of tariff-driven stagnation

 

 - by New Deal democrat

Let’s take a look at the “real” purchasing power of average working and middle class Americans.

The July jobs report showed that average hourly earnings for nonsupervisory workers rose a little under 0.3% (blue in the graph below). Consumer inflation (red) rose 0.2%, so “real” average hourly earnings rose 0.1%. The below graph is the monthly changes in each over the last 24 months, showing that nominal monthly wage gains have slowly decelerated from over 0.3% to about 0.25%, while inflation, with the exception of a few months, was somnolent during 2024 and the first few months of 2025:

The net result is that real average hourly wages for nonsupervisory employees have risen on trend through last month:

Here is the nominal YoY% change in each, showing the slow deceleration of nominal wage gains, along with - until recently - the similar slow deceleration in consumer inflation, driven mainly by slowing real and fictitious rent appreciation:

The danger going forward, obviously, is if tariff-driven inflation picks up, while wage gains continue to decelerate.

For the economy as a whole, the more important metric is real aggregate nonsupervisory payrolls. Last month these jumped by 0.6% nominally, translating into a 0.3% growth in real terms. Thus in absolute terms (blue, right scale) real aggregate nonsupervisory payrolls set a new record, although the pace of improvement has slowed to only a 0.3% gain in the past four months. On a YoY% basis (red, left scale) they are up 2.2%:

To repeat, with almost 100% reliability, a peak in real aggregate nonsupervisory payrolls has preceded recessions in the past 50+ years by a few months. This suggests that no recession is likely in the next few months, although there is the same danger of slowing aggregate payroll growth and accelerating tariff-driven inflation.

Finally, let me update a metric I haven’t noted in awhile, but which showed up as the source for the Oval Office press conference last week in which T—-p touted his “real” economy: namely, the monthly real median household income compilation by Motio Research.

 

This group picked up the torch after Sentier Research discontinued the series. In the graph below, I show the data from 2017 to the present:

T—-p used the series to show how real median household income had increased strongly during his first term (true, until Covid) and stagnated during Biden’s term (true for the first two years, but it rose 2% during the last two years). 

He also showed a graph beginning in January or February of this year, also showing a big increase. This was very misleading. Through May, real median household income hadn’t grown at all this year, and was actually *down* -0.1% since last September. The entirety of the increase came in June, when real median income increased 0.3% in one month (Motio hasn’t updated July yet).

Since the June increase could be one noisy month, the overall trend for the past 9 months has been stagnation in real median household income. Yet another reason to be very concerned if tariffs hit consumer finances harder.

Housing remains recessionary. Why hasn’t one happened yet?

 

 - by New Deal democrat

The post pandemic period has been an exception to many past relationships. This morning’s data on housing construction raises the issue as to whether housing is going to be included in those exceptions as well. That’s because the data has been classically recessionary for a number of months, and yet the economy has not rolled over. 

In general, this morning’s report on housing permits, starts, and construction continues the trend of posting low multi-year numbers, but the downward trend may be abating.

Permits (gold in the graph below) declined -39,000 to 1.354 annualized, while the more noisy starts (blue) increased -70,000 to 1.428 annualized. The former in particular remains very close to its post-pandemic lows. The metric that is the least noisy of all and conveys the most signal, single family starts (red), rose 1,000 to 870,000 annualized:

In the above graph, I normalized permits and single family permits to 100 as of their post-pandemic peaks. I did the same for starts, but used their peak three month average. Starts are down 20.0% from their peak, permits 29.5%, and single family permits 30.0%. 

As I showed you last month, the historical pre-pandemic absolute levels of all three of the above metrics indicates that the current levels of decline from peak were typical of those in place at the beginning of most of the recessions of the last 50+ years, although in two cases - 1991 and the Great Recession - they were down 50% or more.

On a YoY% basis, steep declines in permits and starts, generally more than 10% YoY, have persisted right up into recessions:

By contrast, at present, permits are down -5.7% YoY, single family permits -7.9%, and the noisy starts actually higher (vs. a very low July 2024 comp) by 12.9%.

As I have written many times in the past several years, the best “real” measure of the economic impact of housing is units under construction (blue in the graph below). This month they rose 1,000 from last month’s four year low to 1.357 annualized. They remain down 21.9% from their peak (graph compares with single family permits, where both are normalized to 100 as of their respective post-pandemic peaks):

Again, as I’ve written in the past two months, more often than not in the past, by the time units under construction had declined by this much, a recession had already begun. The only two exceptions were the late 1980s, where the pre-recession decline was -28.2%, and 2007, where the pre-recession decline was -25.6%.

One reason why the steep decline in housing has not caused a recession yet is that other durable goods purchases, and in particular motor vehicle purchases, have not followed suit. In the below graph I show the historical pre-pandemic YoY% change in housing units under construction (blue) vs. purchases of autos and light trucks (gold, averaged quarterly):

With the brief exception of the 1981 “double dip” recession, motor vehicle purchases were down YoY for several quarters before the recession began, although in the case of the 2008 Great Recession it was only by about 2%. By contrast, so far this year purchases of cars and light trucks is running slightly *higher* YoY:

As I indicated yesterday, the trend in absolute light vehicle sales this year so far is flat. if that hasn’t changed by the 4th Quarter, we might very well have the negative YoY sales signal from motor vehicles that is lacking at present.

Finally, let’s take a look at housing units under construction (red) vs. the final shoes to drop typically before recessions have started, houses for sale (gold) and residential construction employment (blue), in comparison with units under construction. I won’t bother with the historical view this month, but we did get important revisions in the payrolls report earlier this month:

It appears that the proverbial shoe has dropped with regard to residential construction employment, which as revised has declined every month since March, albeit only by a total of -0.3%. On the other hand, the number of new single family houses for sale made another new high last month.

I would expect all three series to be negative YoY by the time a recession begins. That could happen by the end of this year.

The muddied historical picture of PPI vs. CPI

 

 - by New Deal democrat

Forecasting has always been hard, and moreso since the supply chain issues of 2021-22 made reading the interest rate signals from the long leading indicators muddled. But at least the short leading indicators were intact.

But now we have the additional wrench in the works in the form of a mafia-style blowout being the operative behavior from the US Administration. If sowing chaos were a winning economic move, banana republics everywhere would be wealthy. There’s a good reason why they’re not, and that’s because chaos and corruption make it impossible for producers to foresee the results of their economic actions.

I see no reason to disagree with the idea that the net result of T—-p’s chaotic imposition of tariffs, plus the $Trillions in deficits that will be run up in short order by the recent tax and spending bill will ignite stagflation - although we can’t see the finer details yet.

But are there a few shadows on the wall that are becoming evident from last week’s consumer and producer inflation reports?

To start, as I wrote last week, both housing spending and motor vehicle purchases have declined in recent months:

The longer trend in real spending on motor vehicles has been flat since the turn of the year.

A similar dynamic turns up when we compare real spending on durable goods vs. nondurable goods:

The former have trended sideways since last December, while the latter have continued to increase. This is a typical historical progression a year or so before recessions.

Let’s turn to the inflation reports now.

From the end of World War 2 until the turn of the Millennium, a YoY% increase in the producer prices for finished goods higher than consumer price increases was a realizable sign of an oncoming recession, the only significant exception being the deep slowdown of 1966:

This was the era dominated by the goods-producing sector of the economy. If producers could not pass on their increased costs to consumers, they would have to cut production and/or employment, which had the effect of causing a recession.

That hasn’t been the case since the beginning of the “China shock” at the turn of the Millenium:

Producer goods cost increases in excess of consumer price increases have been a nearly constraint feature during expansions for most of the past 25 years, although typically real GDP YoY (black) turned down several quarters after the surge in producer goods prices. Note that at present, even with the poor PPI report for last month, on a YoY basis consumer prices have still increased more.

This is likely partly due to the adoption of “just in time” inventory management, which meant that there were less inventory overhangs that needed to be cleared that in the past, and also partly due to the relative fading of the goods producing sector vs. services in the US economy.

But what of the PPI for services? We only have about 15 years of data, and few trends are evident:

There is some slight evidence that the PPI for services *may* slightly lead the CPI, and also some slight evidence from 2016, 2019, and 2021 that when PPI for services has increased more than CPI, real GDP has slowed down within a quarter or two.

If PPI continues to rise vs. CPI, I would expect to see a further slowdown, possibly showing up in the services sector more than during expansions in the past.