Will 2023 be “beautiful” (as Ken Fisher asserts) or ugly (as Harvard professor and former IMF chief economist Kenneth Rogoff fears)?
“After 2022’s woes,” says Ken Fisher (“Big Year for Stocks, Bonds,” The Australian, 16 January), this year “markets are primed for beautiful gains. Last year’s fear fest and global bear market has many (people) dreading more (losses this year). But this mentality primes the ASX and global stocks and bonds to shock almost everyone (into) a bull market bringing 15, 20 (or even) 25% gains. Maybe more!”
We can say very few things with certainty about financial markets. One is that they always fluctuate; another is that they fluctuate in erratic cycles; thirdly, they usually fluctuate unexpectedly. For these reasons, neither I nor anybody else can state categorically that Fisher is wrong. Quite the contrary: the major contentions that produce his ebullient prediction are substantively significant and essentially correct. Yet overconfidence leads him to overstate it (see Why you’re probably overconfident – and what you can do about it, 14 February 2022).
Successful investment presupposes a thorough and dispassionate consideration of evidence. Alas, Fisher’s is superficial and biased. He draws his exuberant conclusions by cherry-picking specific points that support his case – and discounting or ignoring general ones that qualify or weaken it. His dismissal of the risks and consequences of recession is particularly cavalier.
In this article, I subject his claims to methodical and even-handed analysis. I demonstrate that
- It’s possible – and, if there’s no recession in the U.S., even likely – that American and particularly Australian stocks will rise modestly this year;
- Whether or not a recession occurs, it’s also possible that the All Ordinaries and S&P 500 indexes will generate material losses.
- If there’s a recession, appreciable losses are probable but not inevitable.
- Fisher’s expectations of strong gains and the commencement of a bull market in 2023 are doubtful (see also How low could stocks go in 2023? 14 November 2022).
The Political Business Cycle
What explains Fisher’s bullishness? Geopolitical developments play a role. “Consider,” he says, “China’s and Hong Kong’s reopening. They will boost Australian economic activity” – and, he implies but doesn’t explicitly say, its stock market. And inbound tourism, particularly its resurgence from China, “will contribute mightily.” But are these factors strong enough to cause a new bull market? Never mind – they’re not the major force that’ll propel stocks. Fisher doesn’t use the phrase, but the “political business cycle” (PBC) clearly underlies his exuberance.
According to Britannica.com, this phrase “is used to describe the stimulation of the economy just prior to an election in order to improve prospects of the incumbent government getting re-elected. Despite numerous attempts to establish their existence, empirical evidence of political business cycles remains rather equivocal” (see in particular Alberto Alesina, et al., Political Cycles and the Macroeconomy, MIT Press, 1997).
The PBC comprises four key contentions. First, “expansionary monetary and fiscal policies have politically popular consequences in the short run, such as falling unemployment, economic growth and benefits from government spending on public services.” Secondly, “however, the same policies, especially if pursued to excess, are found to have unpleasant consequences in the long term, such as accelerating inflation and damaging the foreign trade balance.” These consequences eventually require counter-interventions, such as monetary policies that lift rate of interest.
But that’s getting ahead of ourselves. Thirdly, politicians are oriented almost exclusively towards the short term: they concentrate upon but discount beyond the next election. Hence they “will pursue popular expansionary monetary and fiscal policies immediately before an election. However, being aware of adverse effects of expansionary policies, they will not intend to keep those measures after they get elected.”
Accordingly, and fourthly, after the election politicians will often change course, e.g., slow the growth of (spending and) money supply, and allow interest rates to rise. As a result, the (fixed timing of presidential elections that occurs in the U.S.) will produce cyclical fluctuation of economic activity – the PBC – “because of recurring patterns of government stimulus and restraint in order to induce a boom (at) election time.”
How to apply these general contentions to specific settings? American presidents serve a fixed four-year term. They’re elected every fourth November and sworn into office the following January. The first year of Joe Biden’s administration was 2021, last year was his second year, this year is his third and 2024 will be his fourth. It’s an oversimplification, and as Britannica notes the evidence supporting the PBC is equivocal, but its crux is clear: in order to win re-election, in the second and third year of his term a president will attempt to engineer an economic upswing.
Fiscal and monetary stimuli take at least a year to enact and take effect; accordingly, if all goes well they arrive during the president’s third and final (re-election) year. These stimuli, in turn, blow wind into the stock market’s sails – which, as part of a general “feel good” factor, enhance the president’s prospects of re-election.
Re-Analysing Fisher’s Five Contentions
#1: The Third Year of a President’s Term
A president’s ability to open the fiscal sluices depends upon his capacity to persuade (or coerce) Congress – which might or might not be in a mood to do his bidding (for an analysis of the impact on financial markets of the partisan stripes of presidents and Congresses, see Will Joe Biden be good for investors? Why I disagree with Geoff Wilson, 6 December 2020). Further, the president’s ability to push the monetary pedal to the floor depends upon his ability to cajole the Federal Reserve. These key uncertainties – and plenty of others – greatly muddy the PBC’s waters. Nonetheless, according to Fisher, “the S&P 500 (Index) hasn’t had a negative third year (during) any president’s term since 1939, averaging 18%-plus returns since good data started in 1925.”
Several specifics of this assertion are plainly false. The S&P 500’s return was negative in several other third years of a presidential term (i.e., -1.2% in 1947, -8.9% in 1987 and -3.1 in 2011). Moreover, thanks to Robert Shiller, “good data” have existed since 1871. During the 19th century, presidents lacked the means to influence the business cycle; for that reason, I’ll replicate Fisher’s starting year. Since January 1925, during all rolling 12-month periods the Index’s return has averaged 7.4%; during all calendar years, its return has averaged 6.4%. (These results, and the ones below, exclude dividends.)
Although Fisher’s first claim is clearly incorrect with respect to several particulars, it’s nonetheless true in a general sense: since 1925 the Index’s average return during the second year of a president’s term has been lower (the average of rolling 12-month periods is 2.7%) than the overall 12-month return; and its average return in his third year (12.2%) has been higher than the overall average; accordingly, the average in the president’s third year is much greater than in his second (Table 1a). Notice, however, that my estimate (12.2%) is more modest than Fisher’s (“18%-plus”).
Investors need no reminder that in 2022 the S&P 500 slumped 19.4%. Last year was the second of President Biden’s term, and this year is its third; given these facts, plus the tendency for the Index’s return to improve during a president’s third year, Fisher expects that stocks will rise strongly this year.
Table 1a: Annualised Return, S&P 500 Index, by Year of Presidential Administration, 1925-2022
Fisher makes a significant point, but Table 1a also shows that he greatly exaggerates it. It reveals evidence of a PBC, but it’s weak. The S&P 500’s return is higher in Year 3 than in Year 2, but it’s much lower in Year 4; moreover, the 95% confidence intervals (CIs) around each year’s means include strongly negative returns.
Table 1a shows the standard deviation (a statistical measure of a set of observation’s variability) of each presidential year’s average annualised return. Given that 2023 is Year 3 of the president’s term, what can we expect? Given the S&P 500’s returns’ since 1925 during the third years of presidents’ terms, as well as standard statistical assumptions which these data roughly meet, the probability is 95% that the Index’s return in 2023 will be ± 2 standard deviations from its mean, i.e., it will range from 12.2% – (2 × 17.5%) = -22.8% and 12.2% + (2 × 17.5%) = 47.2%.
Fisher could be right – it’s possible that this year the S&P 500 might lift 15-25%. Indeed, based upon the variability of its returns over the past century, it could soar more than 40%. But if past is prologue, that’s very unlikely because it’s pushing the CI’s upper bound. Equally, however, if past variability persists then he could be wrong by a country mile: the Index might fall markedly for a second year in succession.
It’s reasonable to infer from Table 1a that this year’s return will be better than last year’s: my estimate is 12.2%. That’s just one-half of Fisher’s ebullient contention. Moreover, the great variability of past results necessitates caution. We can’t reasonably expect – as Fisher overconfidently proclaims – that this year’s return will be strongly positive. We can’t even say with great confidence that it will exceed 0%. Markets will fluctuate, and they will oscillate unexpectedly – and the S&P’s past variability clearly makes it imprudent to advance bold claims such as Fisher’s.
#2: America’s PBC Extends to Australia
According to Fisher, his first contention “spreads globally. Australian stocks also fared best in U.S. presidents’ third years, with very similar returns.” Table 2a shows that he’s substantially correct – but the same crucial caveat that I’ve already mentioned continues to apply (for full details of the Australian data that underlie Table 1b, see Recessions usually crush shares – but investors can always reduce their ravages, 31 October 2022).
Given that this year is the third year of the U.S. presidential term, what result can we expect the All Ordinaries Index to generate? According to Table 1b, we can expect a gain of 15.1%.
That’s a good result by historical standards, and a very good one compared to last year, but it’s at the lower end of Fisher’s expectation. Moreover, and as the CIs indicate it’s entirely possible that this year’s result, like last year’s, will be negative. Given the variability of past results, it’s also possible that this year’s loss will exceed last year’s.
Table 2a:Annualised Return, All Ordinaries Index, by Year of U.S. Presidential Administration, 1937-2022
#3: Negative Second Years Beget Strong Third Years
“Even better” than his first conjecture, says Fisher, “the nine negative second years of presidents’ terms (like 2022) turbocharged third years – delivering 28.7% median returns.” Precisely because last year was sub-par, he reckons this year will be above-par.
Again, Fisher makes a significant point – and greatly overstates it: I counted 12 negative second years and calculated a mean of 12.9% (Table 3a). It’s true that negative second years of a president’s term tend to beget above-average third years. But my estimate is less than half of Fisher’s. Moreover, the considerable variability of past results again includes the possibility – even the reasonable likelihood – of another negative result in 2023.
Table 3a: Annualised Return, S&P 500 Index, Rolling 12-month Periods in Years When Administration’s Second Year is Negative, 1925-2022
Fisher doesn’t mention it, but for the sake of thoroughness of analysis and in fairness to him, it’s important to note that this result also applies to Australia (Table 4a). Given the S&P 500’s negative result in 2022, we can expect that the All Ordinaries index will lift 11.4% this year. But we shouldn’t be surprised if it falls considerably. As before, so it remains: the past variability of results produces a very wide confidence interval.
Table 4a: Annualised Return, All Ordinaries Index, Rolling 12-month Periods in Years When U.S. Administration’s Second Year is Negative, 1937-2022
#4: The Mid-term Upswing
Fisher advances a fourth assertion: “the nine months beginning with October of a midterm election year (that is, the half-way point of a president’s term) are the most positive streak in U.S. stockmarket history with 20% returns and positive (returns occurring 92% of the time).” To make sense of it, we need very briefly to review the timing of congressional and presidential elections. Elections of the House of Representatives occur every two years; accordingly, every second election (such as the one in 2020) will occur at the same time as a presidential election. The most recent House election, in 2022, was a “mid-term” in the sense that it occurred at the half-way point of the president’s term of office. Because no federal election occurred in 2021, it was a non-election year.
I’ve labelled each month of each year since January 1925 as either “Presidential,” “MidTerm” (MT) or “No Election” (NE). Fisher’s third assertion concerns the S&P 500’s returns from the October of a midterm election year (which I’ve labelled MT10) and during the following eight months – which I’ve labelled MT11 and MT12, and NE-1-6. I’ve extended the timeframe to include all twelve months of the mid-term year (MT1-12) and the following year (NE1-12). Figure 1a plots the S&P’s average return and CI in each of these 24 types of month.
Figure 1a: Annualised Return, S&P 500, by Month during Mid-Term and Non-Election Years, 1925-2022
Figure 1a uncovers clear but weak evidence of the Congressional variant of the PBC.
In MT1-MT3, the S&P’s average 12-month rolling returns closely approximate the average return (7.4% per year) in all rolling 12-month periods. In MT4-MT10, average annualised returns fall gradually below the overall average, slumping to -1.1% in MT7. During the next 12 months, average annualised returns steadily rise to and above the overall average, reaching 13.1% per year in NE7; thereafter they steadily fall, etc.
Fisher correctly makes a significant point: the S&P’s returns rise from the October of a midterm year to the middle of the next year. 2022 was a midterm year; on that basis, he expects the S&P’s returns to rise steadily from October 2022 to mid-2023. And as of mid-January, he’s been correct.
Once again, however, I add the vital caveat that this relationship is weak. Specifically, the CI’s average lower bound is -30% and its average upper bound is 43%. By this criterion, Fisher’s expectation that annualised results will climb up to 25% on an annualised basis by the middle of the year could be correct. Equally, it could be grossly inaccurate. The past variability of results has been so great that a wide CI surrounds our estimate.
#5: What If There’s a Recession?
Many people, ranging from the heads of global organisations to the owners of small businesses, expect a recession in 2023. I’ve not just written in detail on this topic (see, for example, How low could stocks go in 2023?): I’ve specified how Leithner & Company has prepared for the possibility of a downturn (see in particular How we’ve prepared for the next bust, 28 November).
Fisher acknowledges the widespread – even ubiquitous – expectation that “a global recession is coming …” He cites KPMG’s survey of 1,325 Australian business leaders, published in October, which found that 86% expect a recession in this country by this summer. He also cited a survey by the Conference Board that found that virtually all (an astounding 98%) multinational CEOs expect a recession in the U.S., and that 99% foresee a contraction in Europe.
Yet he’s unperturbed: “… even if recession hits, mass fretting will minimise (its) impact.” By this he seems to mean that “firms have increasingly prepared for recession since last (northern) spring, (and) anticipation is mitigation.” Fisher also seems to doubt that a recession will occur: “In my 50-plus years as a professional investor, I’ve never seen any recession that was widely anticipated.” He concludes: “a small or no recession would be a big positive.” I agree – and hasten to add what Fisher omits to mention: a recession, and particularly long and deep one – would be a big negative.
It’s telling that Fisher cites no historical data or trends to substantiate his relaxed attitude towards recession in 2023. I suspect that’s because he knows full well that any such citation would greatly weaken his bullish case.
Table 1b: Annualised Return, S&P 500 Index, by Year of Presidential Administration, 1925-2022
Table 1b replicates the analysis in Table 1a but bisects the data into two portions: those months within intervals that the National Bureau of Economic Research identified as recessions; and months outside of recessions (for full details, see Recessions usually crush shares – but investors can always reduce their ravages). It shows that the “presidential year” effect is robust, but that the “recession effect” is stronger (compare the corresponding means in the “recession” and “no recession” columns).
If there’s no recession, and whether we consider rolling 12-month periods of the 12 months to December, our point estimates don’t increase much: it’s reasonable to expect that in 2023 the S&P 500 Index will increase ca. 12.5-15%. Again, however, we must be cautious: large standard deviations imply confidence intervals that include strongly negative returns.
If a recession does occur, estimated returns plummet: they become moderately (average of -7.1% for all 12-month periods) and strongly (average -21.7% for all calendar years) negative.
I’ve criticised Fisher for ignoring the variability of past returns; hence it’s imperative that I consider them. The lower bound of the CI for all rolling 12-month periods is a gut-wrenching, GFC-style loss of -7.1% – (2 × 25.4%) = -57.9%, but the upper bound is 43.7%. In other words, if past variation is prologue, and even if a recession occurs at some point during this year, then the S&P 500’s return could conceivably scale the heights Fisher expects. A similar point applies to the CI for the 12-month periods to December. But the key point remains: the huge variability of past returns under these conditions makes it vital to consider the CI as well as the point estimate.
Table 2b: Annualised Return, All Ordinaries Index, by Year of U.S. Presidential Administration, 1937-2022
Table 2b replicates the analysis of Australian data in Table 2a. On the one hand, Fisher’s key contention remains robust: whether or not there’s a recession in the U.S., the All Ordinaries generates a better return during the third year than in the second year of an American presidential administration. On the other hand, if a recession occurs in the U.S., then during Year 3 the All Ords generates average annualised losses of ca. 16-18%. Thirdly, however, the CIs indicate that past results have varied greatly; accordingly, despite a recession in the U.S., the Ords could lift this year. For all rolling 12-month periods, the confidence interval’s lower bound is -16.3% – (2 × 16.5%) = -49.3%, but its upper bound is -16.3% + (2 × 16.5%) = 16.7%.
Table 3b reproduces the analysis in Table 2a. Given that the S&P 500 fell last year (Year 2 of a presidential term), if a recession occurs this year, we can expect the S&P’s return to decrease slightly (1.9%). This surprisingly mild point estimate, plus the very large standard deviation, results from a small number of particularly variable cases.
Table 3b: Annualised Return, S&P 500 Index, Years When Administration’s Second Year is Negative, 1925-2022
Finally, Figure 1b replicates the analysis in Figure 1a but bisects the data into relevant months that occurred during and inside recessions. To reduce the number of squiggly lines, I’ve omitted the CI’s lower and upper bounds. Crucially, however, they’re very wide: for non-recession months, the CI’s average lower bound is -21% and its average upper bound is 43%; for recessionary months, its average lower bound is -45% and its average upper bound is 25%.
Figure 1b: Annualised Return, S&P 500, by Month during Mid-Term and Non-Election Years, 1925-2022