Macro Trends/Forecasts

Jeff Miller

"The Man Who Called Dow 20,000" --CNBC

Causal Modeling for Current Economic Issues

Everyone in the investment world knows that correlation does not equal causation.  That knowledge does not inoculate them against errors!

Understanding your investments begins with some idea about what is happening.  It provides context for events and a basis for evaluating scenarios of risk and reward.

Analyzing causation is frequently a challenge, and economics is an extreme case.  So many variables move in the same direction at about the same time.  Traps abound.  Serious students spend entire semesters on advanced courses in causal modeling.  We cannot take the time or space for that, but we can make helpful progress just by understanding some basics.

Some Illustrations

Here is a simple but powerful example that I often used in this course: the traffic director.  Play along with me for a bit and we’ll circle back to recent market events.

  1. A man dressed in khakis, a baseball cap, and a whistle around his neck stands in the middle of a busy intersection.  He waves his arms, gesturing traffic to make turns, stop, or proceed through the intersection.  Over his head is a traffic signal indicating the same instructions.  The drivers do as directed.
  2. A man dressed the same way as in #1 makes similar instructions at the busy intersection.  In this case, the traffic signal is dark.  The traffic follows his instructions.
  3. A man in a police uniform stands in the middle of a busy intersection.  He makes signals that contradict those of the traffic signal.  It is a regular time for a factory to end work, and there is a flow of many vehicles exiting the parking lot. The traffic follows his instructions.
  4. A man dressed in an orange vest and holding a walkie talkie raises his hand to stop traffic.  Drivers can see some heavy equipment up ahead.  The drivers stop.  After a minute, some cars pass, and the vested man gives the signal to go ahead.
  5. A man dressed in khakis, a baseball cap, and a whistle around his neck stands near a group of young men wearing football pads.  When he points, gestures, or whistles, they all react immediately.

These cases each involve a different causal path, even though the results seem similar.  How can we distinguish causation?

Methods for Detecting Causation

Determining a causal path can be very complicated, but a few techniques take us a long way.

  • Controlling for one possible independent variable. If we are suspicious about whether the man in the intersection is directing the traffic, we can eliminate other possible sources of causation.
    • We could change his attire.
    • We could turn off the traffic signal.

If traffic still obeyed without the traffic signal, that would suggest causation.  If it obeyed regardless of attire, that would strengthen the idea that drivers respected the directions when there was an obvious need for someone to step in.  If traffic ignored the man when the signal was working, causation would be refuted.  If traffic followed the directions of a uniformed person, and contrary to the signal, that would be good evidence of causation.

  • Plausibility and logic.  If there is reason to believe in the authority of the person doing the directing, that adds strength to the inference of causation.  If the circumstances call for someone to take action, that also adds strength.
  • Timing.  The first essential is that the potential causation must occur before the effect.  This may seem obvious, but it is a bit tricky.  Let us suppose that a country is in the process of electing a new leader.  Positive economic and social events occur.  The starting point for measuring these effects is murky.  It could be any of the following:
    • The moment the new leader takes office seems most obvious – but…
    • After the new leader has time to implement some policies (the first 100 days?) since it may take time to unwind mistakes of the past.
    • Even later, since there could be a delayed reaction to the new leader’s wisdom.
    • When the new leader is elected but before he takes office.  After all, people may start anticipating positive changes.
    • When the new leader demonstrates a solid advantage in pre-election polls.

Applying these techniques to the examples

  1. In the first example we simply do not know the answer.  The director and the traffic could both be looking at the signal, which is actually guiding their behavior.  Technically, this is called a spurious relationship (a term frequently mis-used by non-statisticians).
  2. In the second case the argument for causation is strong, since there is no plausible alternative explanation.
  3. Police guidance against the light is strong evidence of causation based upon logic and circumstances.  It is situationally persuasive.
  4. The walkie talkie operator seems to be the cause of traffic movement, but is he?  The person on the other end of the connection is giving directions to him.  He is merely conveying those directions.  Testing this would require a control where the two actors gave different directions.
  5. Situation and attire both demonstrate causation.  The Michigan Wolverines follow Coach Harbaugh (famous for wearing khakis, a cap, and a whistle) and they do it pronto.

Applications to current market questions

There is great temptation to take a position and seize upon events to buttress your case.  Let’s take a closer look.  I am not going to “prove” each point.  These are items for investigation.  I will state my own conclusion in each case.  Disagree if you will, but please provide a causal model that makes sense.

Did the markets lead the Fed’s interest rate cut?

I say “no.”  The Fed and the market were looking at the same traffic signal.  In fact, the Fed continues to see a stronger economy than the market.  Please note the two dissents, and that the market was mostly looking for 50 bps.  This fits the chart I have posted several times showing the preponderance of market errors in “predicting” Fed moves.

Did the Fed action flatten the yield curve?

No.  The trading immediately afterward and the next morning showed a modest steepening in the curve.  The timing is not consistent with this conclusion.  I have heard some convoluted explanations from committed Fed bashers.  You have to believe that those super-smart bond traders were considering the evidence for nearly a day before acting in unison.

What was the cause of the spike in bond prices?

Given the timing of the collapse in rates and stocks and the rebound after the Chinese acted to stabilize currency moves, the answer seems obvious.  As I have been writing for months, we are getting a real-time lesson in economics.  If you want to understand either stocks or bonds, you need to understand the trade dispute.

Does the lower US interest rates signal weakness in the US economy?

Not directly.  European rates have gone negative and there is a relationship with US rates.  Even if the US “wins” its trade war, the impact on Europe has an indirect effect.  Someone wanting to make this argument would need to show independent US effects, just as in the walkie-talkie example.

What to Do?

I don’t know what will happen and neither does anyone else.  The pundits who think they understand Trump’s tactics or can see inside the minds of the Chinese are just trying to get TV time.  It is amazing how many newly minted experts appear whenever there is a fresh problem.

My best guess is for an eventual resolution, probably in multiple stages.  My reason is the long history of conflict and compromise.  When all parties benefit, they find a way. 

 

Jeff Miller provides Economic Analysis as well as Market Forecasts as one of the original contributing analysts at FATRADER. A quantitative modeling expert and former university professor who is the #1 Economics contributor at Seeking Alpha, Jeff is regarded as an expert on economics, market reaction to news events, and computer-based trading.
Sign Up Now - Free 15-Day Trial! Go! Fatty!