The "real" Fed funds rate and recessions

 - by New Deal democrat

I've seen several posts recently that might be summarized as "don't worry, be happy" because, despite a series of Fed funds rate hikes, the "real" inflation-adjusted Fed funds rate is still negative, on the order of -1%.  I beg to differ.

Let me cut to the chase: below are two graphs covering the lat 60 years (before about 1954, there really wasn't any official "Fed funds" rate). The "real" Fed funds rate is shown in red. The YoY change in the Fed funds rate is shown in blue:

Typically, the Fed funds rate was higher than the inflation rate before recessions. BUT, notice that there were three lengthy "false positives," i.e., the "real" Fed funds rate was also higher than the inflation rate during most of the 1960s, 1980s, and 1990s expansions, two of which were among the best economies the US has ever had.  Further, there was one "false negative," as the Fed funds rate never exceeded the inflation rate before the 1980 recession.

In other words, using the "real" Fed funds rate as your metric gives you correct calls for only 5 of the last 9 recessions, although to be fair, if you simply imply that a "loose" Fed funds rate is inconsistent with an oncoming recession, that improves your score to 8 out of 9.

But the simple fact is, using the *velocity* of Fed funds hikes is the most reliable guide over this time period.  When the Fed funds rate has been raised by 1.75% YoY or more during this period, at very least a sharp slowdown has occurred (1984 and 1994), and on all other occasions a recession followed.  Without such an increase, no recessions occurred.

Presently the Fed funds rate has been rising at 0.75% YoY. This is similar to the first half of the 1960s expansion.

So, is my conclusion "don't worry, be happy"?  No. The issue is found at the far left of the graphs, in the mid-1950s. That was the tail end of the low-interest-rate, low-inflation period similar to today.  Because we don't have data for that entire expansion, we really can't make a conclusion. But (not shown) the Fed funds rate did increase, somewhat slowly, by a total of 3%.

And if there isn't a Fed funds rate before 1954, there was a Fed discount rate, going all the way back to 1914 (NY Fed rates shown), and uniform for the country since 1935. Here's what that looks like compared with long term government bonds:

No yield curve inversions from 1930 to the mid-1950s. During lengthy periods of time, no changes at all. And yet four recessions happened -- one of which, in 1949, followed a two year period of up to 20% inflation(!) from 1946 to 1948, during which time the Fed pretty much sat on its hands. 

In other words, I strongly suspect that relying upon the "real" interest rate to forecast recessions only works on those occasions where the Fed raises rates in order to lower a rate of inflation that it considers too high. 

I'll take a more detailed look at what the Discount Rate might teach us in another post.  

A tantalizing question about existing home sales: have they peaked?

 - by New Deal democrat

I normally don't pay that much attention to existing home sales, even though they are about 90% of the overall market.  That's because their economic impact is not nearly so important as new home sales, because of all of the construction costs and purchases made by the new homeowners over the next year or two.

But let's take a look, at least partly because existing home sales *may* -- and it's a very big "may" -- be in the midst of a turning point.

The overall picture in the report is same old, same old: sales flat, prices up over 5% YoY, inventory still depressed. I do not envy at all buyers in this market, where the average house sells in a mere 27 days!

But let me return again to my basic housing mantra:

first, sales turn

then prices turn

finally, inventory turns

I discount the "months' inventory" metric, because historically it has turned because sales turn, and not because of a meaningful change in inventory.

So -- first, sales. Here's the long term view through last year:

Note that sales turned flat at roughly 5.50 units in summer 2015, and have hovered near that mark +/- 5% since.

Here's the last 12 months through May. June came it at 5.38 units annualized:

See that peak in November? We'll come back to that.

Second, prices. Here's an important graph of median new vs. existing home prices through last year:

The median new house used to sell at a 10% premium. Since the great Recession, that has blown out to 30%. The price of the median existing home has continued to rise strongly, up to $276,900 in June.

Finally, inventory. Again, here's the long term graph:

June's inventory, reported this morning, was 1.95 million units. That is the first YoY increase since 2014.

_________

Now let's get to the tantalizing question: did existing home sales peak last November,  and did inventory bottom last month? Let's follow the mantra:

Sales bottomed in 2010

Prices bottomed in 2012

Inventory made a temporary bottom in 2012, but then has continued to fall -- until now?

In the meantime, let's hypothesize that maybe sales peaked in November 2016.

Prices are continuing to rise.

We would expect to see that inventory will continue to rise.

In other words, it's at least possible that we have reached the point, with existing homes, where increased interest rates and price increases that exceed income growth, have caused the cycle to turn.

The much more economically important new home sales will be reported in two days. This metric, along with purchase mortgage applications, has been the most positive of all housing measures for the past few years. The less volatile single family permits have not made a new peak in four months. Will new home sales continue to defy gravity, or will they too show signs of stress? We'll see.

The President of the United States is a Russian asset

 - by New Deal democrat

That the President of the United States is a Russian asset needs to be openly acknowledged. He may be a naive, negligent or unwitting asset, a coerced asset, or a willing and enthusiastic asset, or some combination thereof, but at this point there is no getting around that he is a Russian asset.

My readers who have followed me from progressive blogs presumably have no trouble accepting this.  But I know that I also have many readers from investment or economic sources, many of whom are probably Republicans. To them I ask two simple questions: (1) in what way has he acted in any way inconsistent with being a Russian asset? and (2) if you evaluated him the same way you evaluated SEC and other filings in order to determine whether or not to purchase a stock, to what conclusion would you come?

What possible reason could there be for a President of the United States to insist on meeting the Russian President both without any witnesses in the room, and also no means to verify what was discussed? Why would a President who is known for bombastically unloading on just about everybody else on the planet, refuse to utter, over a period lasting years, a single negative word about one singular matter: the conduct of the Russian state?

In the past week I have only heard three potential arguments against the fact posited by the title of this post.

The first comes from a comment here, which in summary says:

[H]e thinks he's doing great work; he thinks Putin's terrific; he thinks this will all be justified as a brilliant move once the nation and the world catches up with his brilliance...and, especially, he thinks he's enacting his supporters' wishes, sticking it to the uppity European deadbeats and mending fences with the "real" leaders.....
The fact that he concealed it all from his staff (nobody else in the meeting, etc.) [could just] mean ... [that] it was Putin's idea: let's be alone "so we can really talk" (with some nods to the dangers of "Fake News").

But this just means that the President is a naive or negligent asset, in so deep over his head that he does not know he is being played.

The second, from Al Jazeera, says that "Trump is not Putin's Puppet," postulating that he has parted company with Russian policy when there is profit to be made. But that implies that he has been both willingly and knowingly abetting Russian policy when there *is* profit to be made. And that, by further implication, suggests that he has already done so in his pre-Presidential career, which of course supports the theory that he engaged in money laundering for the Russian mafia, which is coercive both as a subject of blackmail and also as a threat of violence against his family members.

The third is that, in terms reminiscent of Hitchhiker's Guide to the Galaxy, that no asset would be that obvious. Well, a useful idiot certainly would be.

In short, I do not see how any person, discerningly sifting the evidence, could not come to the conclusion that the President of the United States is a Russian asset. We need to face that fact openly and publicly.

Weekly Indicators For July 16 - 20 at Seeking Alpha

 - by New Deal democrat

UPDATE: the article has now been published, and you can find it Here.

I get paid per view, so, you know, go for it!

__________________

All posts at SA have to be approved by their editors. Two weeks ago I was given instructions for how I could "fast-track" publication, if I submitted Friday evening.

Worked like a charm last week.

This week's edition has been sitting in their "pending" bin since late Friday. I've heard crickets.

If it isn't published there by tomorrow morning, I will publish it here so that it doesn't go stale.

The US is not in an economic Boom: midyear udpate

 - by New Deal democrat

At the beginning of this year I asked: Is the US economy going to enter a Boom in 2018?

To recap, there is no standard definition of a Boom. But in my lifetime there have been two occasions when the "good times" feeling was palpable, and the economy was working extremely well on a very broad basis: the 1960s and the late 1990s tech era. During both times,  employment was rampant and average people felt that their situations were going well.

Back in January I identified five markers that, taken together, marked off the two eras as unique: the low unemployment rate, the duration of a very good rate of growth of industrial production, strong growth in real average and real aggregate hourly wages, and increasing inflation.

Let's update all of these through midyear.

First, in both the 1960s and late 1990s, the unemployment rate (note that the U6 underemployment rate wasn't reported in its current configuration until 1994, and so is not helpful), hit 4.5% or below for extended periods of time:

While these weren't the only two periods of low unemployment, they are among those that stand out.

Needless to say, we've hit that marker.

Second, during both the 1960s and 1990s, production grew at or over 4% a year for extended periods of time, not just right after the end of a recession  

  

While the YoY% growth of industrial production has been accelerating this year (up to +3.8% in June), it has still not hit 4%.

The rate of growth of real average earnings for non-managerial employees, both individually and in the aggregate are the third and fourth markers of the two Booms.  In contrast to other expansions, real average hourly earnings also grew at roughly 1% YoY or better:

Meanwhile, real aggregate earnings grew at a rate of 4% YoY or better:

Real average hourly earnings have not grown at all in the past year. Real aggregate earnings are growing at the tepid rate of 2.5%.

The fifth and final  marker of a Boom -- probably as the byproduct of the first four -- is an increase in the YoY rate of inflation:

This has been occurring, although it is probably due more to the price of gas than to any wage pressures.

So far this year, only the first and last markers are present: low unemployment and an increasing YoY inflation rate. But industrial production is not growing as fast as during either of the two Booms, and real wage growth has continued to be lackluster to say the least.

In short, while the production side of US economy is doing pretty well, the consumer side of the economy remains tepid, and in particular wage growth is non-existent. As of midyear 2018, the US economy is not Booming.

US layoff rate at all time lows

 - by New Deal democrat

I don't write about new jobless claims much anymore, mainly because it has been boringly good for a few years (outside of hurricane disruptions!).  But there are times like this week when I am particularly thankful that I am concentrating on the one thing - the economy - that is doing unequivocally well.

Back in 2009 and 2010 when I was arguing with the Doomer dead-end recession double-dippers, I used to hear from people who were still in danger of, or worse, had been laid off. I sympathized, and just noted that, however bad things seemed, they had been even worse in 2008 and early 2009, and conditions were still going to improve.

I haven't heard stories like that in a few years. One of my milestones was when Atrios over at the Eschaton blog stopped reporting on the weekly jobless claims as "xxx,000 new lucky duckies."

Well, this morning, we had yet another new 48 year low in initial jobless claims, at 207,000. We had lower numbers, in the high 100,000's, in the 1960s, but that was against a US population that was only half of what it is now.

In other words, we have never had a lower *rate* of layoffs for as long as records have been collected:

As of this morning, roughly only 1 in 800 workers is filing a new jobless claim each week.

Occasionally  I still see it noted that the percent of jobs which qualify for filing for unemployment insurance is lower than it used to be. That is true, but the DoL keeps track of that number as well, and we can calculate what the number of initial claims would be normed for  this:

If the same percent of jobs now qualified for unemployment insurance as at any point in the last 50+ years, it would translate into about 270,000 new claims per week -- still a very good number.

So today I will celebrate one unabashedly good thing about the US economy: it is almost unheard of now to get laid off.

Disappointing housing permits and starts point to housing, GDP slowdown

 -  by New Deal democrat

For the second month in a row, the preponderance of evidence from housing permits and starts is that increased interest rates and continuing increased prices are beginning to take a bite out of the market. 

FRED doesn't have the graphic updates yet, so let's look at the charts provided by the Census Bureau.

First, here are housing permits:

Total permits declined to a 9 month low, and are actually down YoY. The less volatile single family permits rose, but are just above its 9 month low from one month ago. 

This remains the first time that single family permits have declined about 4% since 2010. But because the peaks for single family homes were only in February, and overall in March, not enough time has passed to be confident that this downturn is truly significant. Declines of 4% or more took place several times in the 1990s and 2000s without signaling the top of the market, as in 1994-95, 1996, and 2004:

It is noteworthy that in 2017, there was a similar pattern of new highs in permits during the winter months, and a decline in the spring and summer. So there may also be some residual seasonality that has not been accounted for (data shown through May):

The more volatile total and single family housing starts also declined to 9 month lows:

There were also significant downward revisions to the last several months. This had a significant effect on the three month rolling average, which cuts down on volatility. Initially, last month's three month average was a new expansion high. With the revisions, the high is now back in March.

So there is not enough evidence of a significant downturn to set off any recession alarms. It would take a decline of at least 5% from peak in single family permits for me to change my evaluation of the housing market. At the same time, as I noted last month, even though the economy didn't roll over, in two of the last 3 times -- 1994 and 2004 -- where there was a similar decline in permits outside of recessions, real GDP did slow down:

Further, note that we are getting the same message from the flat but not inverted yield curve, which over the last few decades has signaled a substantial slowdown in GDP growth several quarters later.

Thus,while the evidence is hardly conclusive, there is an increased preponderance of evidence  that housing is slowing down, and that this slowdown will show up in GDP within the next year.

Real retail sales and industrial production update

 - by New Deal democrat

I'm having a hardware issue, so this will be a quick hit. Hopefully sometime tomorrow everything will be back to normal.

Yesterday we had another excellent retail sales number, up +0.4. YoY real retail sales are up +3.7%. Since these sales are a good short leading indicator for employment, that suggests the good monthly jobs readings are likely to continue, at least for a few more months:

This morning industrial production was positive, but with a significant caveat. Production increased +0.6% in June, extending the recent good trend (total is black, manufacturing is red in the graph below):

The caveat is that May was revised down by -0.4%, so on net June was up +0.2% over the level as of last month's initial report.

Bottom line: the run of good, positive news continues.

Weekly Indicators for July 9 - 13 at Seeking Alpha

 - by New Deal democrat

My Weekly Indicators post is up at Seeking Alpha.

The biggest story continues to be what is happening with the long leading indicators.  This week several moved in opposing directions.

Since my compensation is tied to the number of readers, by all means indulge your curiosity! You'll have to register, but registration is free.

Real average and aggregate wages: July 2018 update

 - by New Deal democrat

As we close out this week devoted to jobs and wages, with the consumer price index having been reported yesterday, let's take a look at real wages.

By now you've probably read elsewhere that YoY wages for average workers actually declined slightly (-0.1%):

But the flatness goes back further: all the way to February 2016:

Real wages have only grown 0.4% in the last 2 years and 4 months.

So why is the consumer economy still growing?  In part because the savings rate has declined:

Another reason for the growth, at least this year, which I'm unable to graph, is that on net there was about $4 Billion per month cut in withheld taxes, much and perhaps most of which is being spent.

Finally, let me update the graph on real aggregate wage growth, which shows that since their bottom way back in October 2009, they have risen by about 26%:

Basically, the improvement in the employment rate, plus the growth in average hours worked, is making up for the stagnation in real hourly wages.

A (mainly) business cycle explanation for this year's better jobs growth

 - by New Deal democrat

The pace of job creation declined from averaging over 200,000 a month between Q2 2014 through Q2 2015 to 180,000 or less during most of 2017. This year, it has picked up noticeably to over 200,000 per month again:

Why?  A basic analysis of the business cycle supplies an answer. To quickly refresh, long leading indicators tell us about the direction of the economy more than 1 year out; short leading indicators by less than 1 year. Payrolls, along with industrial production, are the premiere coincident indicators.

So let's see where we have been over the last few years.

Professor Geoffrey Moore identified four long leading indicators in his 1993 book on forecasting: corporate bond yields, housing permits, real M2, and corporate profits deflated by unit labor costs. These four historically were bundled into the proprietary "US Long Leading Index" by the Economic Cycle Research Institute, or ECRI.  After disappearing from view since 2010, late last year an updated graph was published. Here it is:

Note that growth in the index nearly stalled in early 2016, before rising considerably later that year, then declining in early 2017.

In other words, the index forecast a stalling economy into early 2017, with growth picking up later in 2017 into early 2018.

But we can graph the four long leading indicators ourselves. Here is their YoY% change over the last five years:

We can see their steep slowdown beginning in 2014 and continuing into early 2016, and an upturn in late 2016 through early 2017 before sharply decelerating again.

And here's what happened with real GDP:

Note that YoY GDP growth started to decline one year after the LLI turned down, bottom six months after the trough in the LLI, and accelerated through 2017.

Thus, for example, in my forecast for 2017, which I published in January of the year, I called for the economy to be quite strong in the first half, and continued positivity in the second half.  Following a normal pattern, strength in the long leading indicators should transmit into strength in short leading indicators as we get towards one year later.

So did we see an upturn in short leading indicators in 2017? We sure did.

Here, for example, are three of them: stock prices, real retail sales, and initial jobless claims (inverted), all normed to 100 in January 2016:

After largely stalling in 2016, they turned up somewhat in early 2017 before really taking off later in 2017 into early this year.

Here's another short leading indicator: the ISM manufacturing new orders index:

Note how it turned up sharply going into 2017 and remained that way throughout the year.

So, to recapitulate, we had an upturn in the long leading indicators beginning in Q2  2016, which translated into an upturn in short leading indicators especially later in 2017, and now we see the strength spilling into the coincident indicator of job creation in the first part of 2018, as well as the other premier coincident indicator, industrial production:

I don't mean to suggest that the reason for strong job growth this year is mono-causal. I suspect that hurricane and California fire repairs played an important role in new production and spending during the last few months of 2017. The turnaround in energy production has also  helped. And the tax cut, while very inefficient -- because it is so lopsided in favor of the wealthy -- has shown up in decreased withholding tax receipts this year, which presumably means that amount of money is available to consumers to spend.

So, what's ahead? Well, if you go back and look, the long leading indicators almost all hit zero at the end of last year, before a weak rebound in the first quarter of this year. The stock market has gone basically sideways for six months, initial claims for the last few, and manufacturing new orders, while positive, are not so strong as before.

Thus, back in January my forecast was for weaker, but positive, growth in the second half of this year compared with the first half.

For what it's worth, consistent with my view, at the time ECRI's LLI was publicly published late last year, ECRI's Chairman, Lakshman Achuthan, indicated that they portended a global slowdown.

And that means I expect job growth to slow back down later this year.

May JOLTS report: as good as could be asked for

 - by New Deal democrat

The simple take on yesterday's JOLTS report for May is that it was excellent:

• Hires made a new high
• Quits made a new high
• Total separations made a new high
• Openings were just shy of their expansion high
• Discharges were just shy of their expansion low

In short, pretty much everything good that could happen during a positive labor market, happened.

So let's update where the report might tell us we are in the cycle.

Let's start with the simple metric of "hiring leads firing." Here's the long term relationship since 2000, quarterly through the end of March:

Here is the monthly update for the past several years through May:

In the 2000s business cycle, both turned down well in advance of the recession. That isn't the case now.

Hires had been within a 2% range since last May, and had not made a new high since last October, making YoY comparisons more challenging as of this month, but both hires and separations jumped considerably (interestingly, just like May last year, which makes me wonder if there is a little unresolved seasonality).

Further, in the previous cycle, after hires stagnated, shortly thereafter involuntary separations began to rise, even as quits continued to rise for a short period of time as well:

[Note: above graph show quarterly data, ending Q1, to smooth out noise]

Here is a close-up of the same data on a monthly basis for the last 2 years through May:

This in no way resembles a late cycle report from the 2000s cycle. Nevertheless, the large majority of overall economic data suggests we are late in the cycle. Thus, my anticipation remains that we will see weaker, and even negative YoY readings on hiring in particular at some point in the second half of this year. We'll see.

A business cycle theory of labor force participation and wage growth

 - by New Deal democrat

I've devoted a lot of time and thought, and typed a lot of pixels of commentary, about wage growth in the last few years. Some of it has panned out: based on past expansions,I expected YoY wage growth to bottom consistent with an unemployment rate of about 6%. A little later I refined that to an underemployment rate of 9%. In retrospect that is indeed about when wage growth bottomed out in this expansion.

But even three years ago, I expected wage growth to rather quickly reach 3% YoY once that level of underemployment was breached to the downside.  Obviously, that didn't happen.

Rather than ignore the call that didn't pan out, I have tried to understand why.

One important part is the changed behavioral set-points of both employers and employees. Ever since the 1980s, when unions were effectively broken, employers have found that they can get away with paying less and less to maintain employees. Those employees learned to expect less and less in the way of raises from employers. During each successive expansion, employers have tightened the screws more and more, until by now giving raises has become a taboo, where employers would rather sacrifice at the least short term profits from more production than give in to the necessity of raising pay.

But if the taboo against raising wages is a secularly increasing phenomenon, on another level I think we can still tease out a lot of information in terms of the order and direction of employment and wage trends.

To that end, I want to propose a general theory of labor force participation and wages within business cycles. To wit, at least in the modern era since 1982, the pattern has been:

1. the unemployment rate peaks, and begins to decline.

2. prime age labor force participation bottoms, and begins to rise
3. nominal wage growth bottoms, and begins to rise.

BUT...

4. If labor force participation grows too quickly, wage growth languishes.

In other words, as a recession deepens, not only are more people laid off, but an increasing number of people leave the labor force entirely. At some point, as more of the people remaining in the labor force get work, people who have been on the sidelines are drawn back into the labor force and look for work. As even a bigger share of this expanded pool of potential workers find employment, the wages needed to entice them to a job begin to rise. But sometimes the increased number of those entering the workforce causes the pool of available workers for hire to become temporarily saturated, and in that case the pressure for increased wage growth ebbs.

Now let me lay out the supporting graphs. All of these date from 1982 to the present.

First, the unemployment rate leads prime age labor force participation:

The unemployment rate (blue, inverted in the graph below) also leads wage growth:

Further, the prime age labor force force participation rate leads wage growth:

Thus, we have our progression: decreasing unemployment rate --> rising prime age participation --> increased wage growth. 

But while it seems pretty clear that, over an economic cycle, participation leads wages, within that cycle there may be waxing and waning rates of participation increase. That's what is shown in this next graph, which shows the YoY change in the prime age labor force participation rate:

Note that the last several years have shown the sharpest growth in prime age labor force participation since the 1980s, except for 1996.

Further to the point, in the below graph, prime age labor force participation in inverted, with any YoY rate of increase greater than +0.3% showing as a negative:

We can see that when there is a surge in labor force participation, wage growth tends to wane. In the 1980s, when there was persistent growth in participation YoY over more than 0.3%, wages actually declined.

Let's now zoom in on the last 5 years:

Prime age labor force participation has grown by more than 0.3% YoY in 22 of the last 24 months. While there has been some acceleration in wage growth during the first part of this year, it has still not hit so much as 3%.

To put my whole theory together, I think employers increasingly developing a the behavior of a taboo against raising wages ("wages are also sticky to the upside" in econospeak) is a powerful explanation of the decades'-long secular trend; while the pattern I have set forth in this post is an equally good hypothesis to explain the within-cycle behavior of labor force participation and wages over the last 35 years. 

A teaser about wages and labor force participation

 - by New Deal democrat

I was going to put up a short piece about wages this morning, but it has turned into a longer, more comprehensive piece, so in the meantime, here are some teasers to ponder.

1. There is a direct relationship between the economy generally, and child care costs specifically, and couples' decisions about whether or not to have more children:

The below graph comes from Vox.com. It is the top eight reasons that couples give for not having (more) children:

Note that 6 of the 8 reasons have to do with the economy, and 4 of those specifically have to do with the costs of child care.

2. While correlation is not causation, nonetheless in the modern era, there has been a clear correlation whereby prime age labor force participation leads nominal wage growth:

3. Despite #2 above, in the shorter term there appears to be an inverse correlation between the rate of prime age employment growth and relative wage growth:

I'll flesh this out in the more comprehensive post.

Understanding the simple strategy for the 2018 elections

 - by New Deal democrat

It's Sunday, so I feel free to stray from nerdy economics ....

Pretend for a moment that you are a political strategist. Your party is the party in power. The opposition has been enraged since the moment of your standard-bearer's election. In the special and off-year elections since, they have been showing up in unprecedented droves for offices up and down the ticket, from Governor and US Senator to state representative and local council.

The mid-term election is bearing down, and you know very well that midterm elections are fundamentally referendums on how the party in power is doing. The odds look overwhelming that the opposition turnout tsunami, at least, is likely to continue.

As a strategist, what of the following courses of action do you recommend?

(a) make soothing noises, hoping that the anger of the opposition is mollified enough that your candidates squeak through

(b) engage in searingly divisive behavior that will rile up your base
(c) get rip-roaring drunk and hide under a table curled up in a fetal position until it's over, so that you don't remember anything

Whether or not you choose to do (c), it's pretty obvious that (b) is the correct answer, isn't it?

Opposition supporters already hate your with the heat of 1000 suns. So what if you repeatedly do such cruel and outrageous things that they become incandescent with rage. They're going to turn up anyway, regardless of what you do. 

Since the opposition is going to turn out in droves, your best option is to get your base to turn out in droves as well. You don't do that by disappointing your base, or putting them to sleep (Cf. Obama, 2010). Here's Rasmussen's poll of "strong approval vs. strong disapproval" from that time:

Note how enervated Obama's base was, with only about 27% +/-2% during that time.

No, you do that by getting your base incandescent with rage as well. You want to pick the bloodiest fights you can, so that your base feels they need to show up and support you and your allies, lest their enemies prevail. Here's Rasmussen's equivalent poll for Trump right now:

Note that the level of strong disapproval is about the same as for Obama, but the level of strong approval has been inching up, and is currently at about 33% +/-2%.

Once you understand this simple strategy, all of the horrible things that have been done by Trump and the GOP in the last few months make perfect sense. The more mean, cruel, divisive, bloody fights that are boiling over this November, the more the GOP base will turn up. Since the Democrats' base is going to turn up anyway, this gives you your best chance to preserve your majorities.