Ricochet is the best place on the internet to discuss the issues of the day, either through commenting on posts or writing your own for our active and dynamic community in a fully moderated environment. In addition, the Ricochet Audio Network offers over 50 original podcasts with new episodes released every day.
Algorithm Finds No “Trump Effect” on the US Economy, at Least Not Yet
So imagine an alternate reality where Hillary Clinton campaigned a bit more vigorously in supposedly safe Big Ten blue states and eked out a narrow Electoral College victory. So, no President Donald Trump. And also no tax cuts, no deregulation, no trade war. How would the US economy — which Trump describes as “stronger than ever before!” — be performing right now? Or to put it another way: Is there an economic Trump Effect?
Well, not so much, according to “Stable genius: Estimating the ‘Trump effect’ on the US economy” by researchers Benjamin Born, Gernot Müller, Moritz Schularick, and Petr Sedlácek. Their conclusion: “The impact of President Trump on the macroeconomic performance of the US economy has been negligible so far. We measure neither an acceleration of growth nor increased job creation in the US economy relative to an appropriate benchmark.”
With no ready access to the multiverse of the Quantum Realm, BMSS tried a rather clever approach to this economic “what if” question. First they let an algorithm construct a “doppleganger” of the US economy by selecting a weighted combination of other advanced economies (with higher weights attached to Canada, the UK, Denmark, and Norway). This twin rhymed the US economy over the past twenty years, and served as a control for the experiment. They then compared the path of the US economy since the election of Trump to its doppelganger that did not get the “treatment” of electing Trump. And this was the result:
From the BMSS analysis:
The main result can be easily spotted visually – there is no acceleration of the US economy relative to the doppelganger. If anything, the doppelganger outperforms initially and by mid-2018 there is no discernible difference. It is important to stress that the doppelganger has been constructed exclusively on the basis of observations prior to the vote. And yet, the doppelganger tracks the behaviour of the US economy very closely after the vote as well. This means that the US economy behaves just as it did before the election. Trump, in other words, was immaterial for US growth. The doppelganger grew just as fast. This result stands in contrast to the substantial effects that the synthetic control method helps uncover in other situations. For instance, in our analysis of the Brexit vote we find a large and significant effect on UK GDP. In the seven quarters after the referendum, UK GDP declined relative to its doppelganger by close to 2%.
Now to be fair, what this experiment is trying to test is the impact of Trump on the economy vs. a sort of status quo — not the possible economic policies of Clinton. But that said, the result really isn’t that surprising. The tax cuts passed only late last year, deregulation takes a while to move from press release to implementation, and the US vs. the World trade war is only now really gearing up. It’s notable that in testimony yesterday Federal Reserve Chairman Jay Powell seemed reluctant to give fiscal policy much credit for what’s been happening in the economy. (And let’s not forget this is perhaps more accurately the Bernanke-Yellen-Powell economy.) Overall, this has stayed a Two Percent Economy under Trump, and good luck finding a big difference in job growth pre- and post-Trump’s election:
Now all that may change as the stimulus from the tax cuts and federal spending increases causes a growth surge, as both Washington and Wall Street are predicting. But even then, we won’t really know if Trumponomics is working. The goal isn’t supposed to be a sugary surge but rather a long-term, structural supply-side change in the economy’s growth potential. The tax cuts, for instance, aren’t meant to merely give consumers more spending money today but accelerate growth and raise wages by boosting investment and productivity. That will take time, though maybe we are starting to see signs of that investment upturn. For now, though, both proponents and skeptics should probably hold their fire and avoid making any big, sweeping conclusions.
Published in Economics
Pass. Ad hominem arguments degrade the site and I won’t stop pointing them out.
Jamie,
Apparently, any argument that isn’t based on linear analysis is ad hominem to you. You insulted a member and that was truly ad hominem. You insulted the entire site. That is simply rude.
Regards,
Jim
Incorrect. The comment in question made the people who conducted the study, not the study itself the issue: that is by definition as hominem. Mine on the other hand took issue with the content of the comment and not the person who posted it. I didn’t take issue with her as a person but rather the argument presented.
Gentlemen, if one of you is finding a comment embarrassingly rude, it is better to flag that comment and give the moderators a reason than to derail the thread with an argument about it.
Jamie,
She quoted facts about the general political tenor of those who had presented the study. Although, that isn’t damning evidence in itself it is of interest. We don’t have an infinite amount of time to donate to everyone who claims expertise. We make judgments based on many things and I suspect her information was rather relevant. Finishing your comment with the non-sequiturs “Boring. Repetitive. Tense.” is simply insulting and the fact that you argue this is even more insulting.
Regards,
Jim
Moderator Note:
Everyone here loves a good pun, but please confine yourself to puns that don't accuse fellow members of trolling. There are multitudes of puns to choose from that don't.I said, “
Troll on Big River, Troll on.” I thought you might get a chuckle out of that. Unfortunately, a sense of humor is not a necessary requirement for a member but ought to be.Regards,
Jim
I share @jamielockett‘s discomfort with arguments that cast aspersions on an argument based on the purported motives of the person making it. Arguments should be addressed based on their merits.
I am also generally skeptical of attempts to model complex systems such as national economies, and I expressed my skepticism. In this particular instance, the modeling isn’t of a conventional computational sort such as used in, say, climate prediction. Rather, the critical detail is the selection of a control group of countries that, the author’s believe, resemble the United States economically and yet are insulated from the influence of President Trump’s economic policies.
Their conclusion:
I don’t find that particularly compelling, for the following reasons.
First, economies are complex, and I’m skeptical that sufficient important variables are being mirrored in the control group. There are more than GDP and employment to consider, and some other factors take considerably longer to become visible. How has the rate of capital formation and investment changed over the past few years? What is the rate of business startups?
For example, here is a graph of the rate of business startups by year. It ends in 2014; what has happened since then? (I don’t know. But I’d want to know before I took the results of the study cited in the original post seriously.)
Secondly, while I give the authors credit for bringing up one obvious potential objection — that the control countries themselves might have been influenced by the U.S. economy, and so not actually be independent controls — I thought that their response to that potential criticism illustrated a more serious problem with their study. They argued that it was unlikely that the the U.S. influenced the control economies
Fair enough, but is it reasonable to focus solely on policies as input? What if there is a substantial intangible component to President Trump’s election, one that would itself influence business planning, entrepreneurial investment, and business and consumer confidence? Might that have a general influence on economies closely tied to our own?
Thirdly, let’s think about the implications of economies that closely track our own. The researchers appear to have used a simple numerical technique for identifying control countries. The result of their number crunching was a small list of countries with economic profiles that appear to track the U.S. quite closely — at least in GDP and employment, the only two categories identified in the study — over the 25 or so years considered in the study.
Why do those nations track us so closely? The study is mute on this, seeming to imply that, because the selection algorithm is numeric and repeatable, it ought to be adequate. But what if the countries selected track the U.S. precisely because there are undisclosed common factors — things other than President Trump’s recent policies — that have historically linked our economies? Wouldn’t that call the validity of the entire study into question?
Finally, the authors do note that there was a divergence between the control group and the United States, when they write (and the graph shows):
We are now in “mid-2018,” so another way of stating the author’s observations is that, since an initial dip in U.S. performance, the U.S. economy has been performing better than those of the control countries. Otherwise we would not have regained parity from an initial deficit.
The data appears to end just as we re-establish parity. What if we continue to outperform the control group for the next two years, as we have for the past two years?
For these reasons, I’m skeptical. I’m not saying that the findings are wrong, merely that I would put little stock in them.
Henry,
Discomfort is one thing but ignoring facts that clearly show a repeated long-standing bias is not wise. Take Peter Strzok for example. If anyone can sincerely believe he wasn’t severely biased going into the investigation I’d be amazed. Of course, I’ve often been amazed.
Regards,
Jim
Jim,
I’m loath to get involved in heated exchanges, and I won’t comment on this one. But I’ve also thought about the difference be a true ad hominem attack and something that looks quite similar, which is an attack on the trustworthiness of a character rather than the argument he is presenting.
In the case of Strzok, I would conclude that he was dishonest and untrustworthy. Based on that, I’d be disinclined to believe anything he said. However, if he presented an actual argument about something, rather than simply claiming to state facts the veracity of which only he could know, I’d be happy to engage the argument.
I wouldn’t consider the statement “the guy has lied to me so often that I don’t believe anything he says” is an ad hominem attack. “His argument is wrong because he’s always biased” would be.
H.
Mods, Jamie, and all,
Now that this redaction has been worked to death I’m logging off and going to sleep. Night all.
Regards,
Jim
But you already did get involved, and you did comment.
Moderator:
I for one find comments like that pun-itive.
I tend to agree with this analysis now that I’ve read some of the details of the study, although I would point out that the complexity of the economy lends one to the conclusion that any President would have a minor effect at the margins compared to the rest of the factors in the economy.
That said I find this studies conclusion plausible since the effects of Trumps major economic policies (corporate tax cuts and regulatory reform) bear fruit on a longer time frame than we have yet seen. This is a good thing! I means that the economy might be even further unleashed and just in time for 2020.
Well, if anyone could actually create an economic model that would accurately predict the future performance of the economy, wouldn’t that person be off making billions of dollars in the stock market instead of doing “studies” on political issues?
Not one model in the history of econometrics has ever actually predicted anything with greater accuracy than throwing darts at a board at random. In fact, I wouldn’t limit the GIGO observations to models. I have one simple rule – all macro-economics is bunk. And that is not my reaction to this study or the Trump economy. I became firmly convinced of that rule back when I was earning my economics degree at UCLA in the 70’s. Nothing that has happened since has changed my opinion one whit.
Well, yeah, but there is one effect that can and does happen immediately, and that is confidence. Just getting rid of Obama and not replacing him with an Obama clone had an immediate and very positive effect on the economy. Consumer confidence went up. Business confidence skyrocketed. Investment increased. Hiring increased. All of this happened before any tax cuts or deregulation. In fact, it started before Trump even took the oath of office. We could have replaced Obama with a monkey (some say we did), and it still would have been great for the economy.
Because picking individual stocks is not the same as forecasting broad economic trends. Most economists I know place their own money in Index Mutual Funds with a minor position in Tbills as a hedge against inflation. This would fit with both the history and most economic models which show that the stock market is on a steady inflation adjusted growth trajectory.
Further, the largest growth field in investing over the last two decades has been in quants and mathematicians building software models for investment firms. Those positions are extremely welL compensated. So there are those that go out and make millions and even billions based on their economic forecasting.
Except that hasn’t been measured. The real economic growth rate is basically the same as it was under Obama – 2ish percent. So is the employment growth rate.
This might be confirmation bias on behalf of conservatives who suddenly feel more confident since they’re back in power.
It would be interesting to see if the method in question always finds no presidential effect. Of course it probably can be used to test past presidents because their economic performance is integral to making the algorithm in the first place.
I think the graph of employment rate is very convincing that at least in that variable there is no difference between Trump and before Trump. Arguing that his election and policies have not had a discernable effect on employment rate. So showing a lack of change in trend in other economic variables is probably the best way to determine if a particular policy has had an effect.
Oh Snapp!!
If I’m not mistaken, I think this post has garnered James the most comments ever . . .
We have to be careful when looking at graphs because formatting matters. The labor force participation graph I shared earlier (repeated immediately below) seems to suggest that Obama arrested a decline and now we’ve leveled off. I shared that view because it seemed to most directly address Misthiocracy’s question. If I were trying to influence a conclusion I might format the data differently.
The long-term graph below makes a slightly different impression. If we attempt to visualize a trend line through this data, we might conclude that LFPR continues to decline through most of 2015, reaches an inflection point in late 2016, and arguably increases in 2017. If we accept that influencing trends in the US economy is generally slow business, we must conclude that the beginning of an upward trend would be all the Trump administration could possibly affect in their time in office and it would look like what we see in this data. I agree that LFPR data doesn’t prove a positive Trump effect, but it’s certainly consistent with one.
Of course, the same thing can be said for tariff policy.
It’s hypothetically possible that positive effects of tax and regulatory reforms will be counter-balanced by increased tariffs and subsidies, and at the end of the day the macroeconomic trendline will remain about the same as at the end of the Obama era. If that trendline remains upwards, then great.
(Unless overall federal spending decreases, I’m holding to the hypothesis that President Trump is still a Technocratic Democrat deep-down, and that the only real difference between Trump and the Democratic Party establishment is where federal attention is focused.)
Well… from what I know, markets are approximately weakly efficient. Full efficiency of markets would mean no asymmetric information and would result in markets being fully stochastic. But approximate weak efficiency means it is possible but very difficult to exploit asymmetric information and predict something better than chance. Arbitrageurs make a profit off exploiting asymmetric information, a profit which is then priced into the market for everyone else to “eat up the arbitrage” so that everyone else’s stochastic models can work well enough.
If arbitrageurs can profit off figuring out something that others haven’t, it shouldn’t be impossible for sufficiently diligent researchers to do the same and make a model predicting somewhat better than chance. But there’s little reason to hope it will be easy, or that the models will do much better than chance.
So… how much better than chance do we need to tell us anything of significance? Would a little better than chance be meaningful in the right context?
So this is what you say you see in the top graph that zooms in on the LFPR from 2013 to the present? Because I think you are miss reading it if that is what you are seeing. Trump assumes the presidency in January 2017 which on the X axis would be to the right of 2016 half way between 2016 and 2018 marks. So the inflection point you mention (the jump from 62.3 to 63 I assume) has to start in 2015 and crosses the new year 2016 and ends before 2017 which is still in Obama’s second term. Following that the number oscillate between 63% and 62.6% through today thus encompassing all of Trumps presidency so far as well as his election in November of 2016.
Larry,
Yes, replacing a President who supported a high tax, hyper-regulatory, anti-business agenda with one that appears to be following the opposite course was quickly embraced by an economy that had been frozen in place. The political jury is still out but the view from the average voter appears that they think more highly of the ‘strange orange monkey’ than the punditry does. That will be empirically confirmed in November 2018.
First, one can twist and turn the model until the numbers give you a chance not to yield any credit to the ‘strange orange monkey’. If you are caught manipulating the model then you can revert to a nihilistic point of view that makes the case for just closing down all economics departments as they are all useless. To suggest that both the discipline of economics and the ‘strange orange monkey’ deserve a little credit is obviously conservative confirmation bias or so we are told.
Regards,
Jim
True. Although I think the few areas were a President can have a measurable effect is trade and tax policy. I’m hopeful that Trump’s corporate tax cut will filter down over the next 12 – 18 months and we will see some real productivity gains resulting in real wage growth.
There is no actual measurable effect indicating this. Yet.
I remain hopeful that the corporate tax cut will have a measurable effect in time for 2020.
Fitting trendlines has as much potential to be misleading as formatting the chart, but I whipped up one for the data from 2000 through 2018. Make of it what you will.
Yeah, I know some of those forecasters. And with all of their models, and all the money they charge for them, fund managers still can’t beat the S&P 500. They are like those guys on late night TV who promise that their seminar will teach you how to make millions in real estate with no investment of your own money and no credit. If they could really do it, they wouldn’t be selling the knowledge in infomercials.
On your other point, I never said anything about picking individual stocks. If you could predict the trend of the economy, you wouldn’t need to bother with individual stocks. Just predicting the dot.com boom and bust of the late 90’s and the real estate boom and bust of the early 2000’s would have made you a fortune, if you could time it accurately. Just those two major events, and nothing more. But you can’t predict it. No one can.