HomeInformation EfficiencyEquity market timing: the value of consumption data

Equity market timing: the value of consumption data


Jupyter Notebook

The dividend discount model suggests that stock prices are negatively related to expected real interest rates and positively to earnings growth. The economic position of households or consumers influences both. Consumer strength spurs demand and exerts price pressure, thus pushing up real policy rate expectations. Meanwhile, tight labor markets and high wage growth shift national income from capital to labor.
This post calculates a point-in-time score of consumer strength for 16 countries over almost three decades based on excess private consumption growth, import trends, wage growth, unemployment rates, and employment gains. This consumer strength score and most of its constituents displayed highly significant negative predictive power with regard to equity index returns. Value generation in a simple equity timing model has been material, albeit concentrated on business cycles’ early and late stages.

A Jupyter notebook for audit and replication of the research results can be downloaded here. The notebook operation requires access to J.P. Morgan DataQuery to download data from JPMaQS, a premium service of quantamental indicators. J.P. Morgan offers free trials for institutional clients.
Also, there is an academic research support program that sponsors data sets for relevant projects.

The below post is based on proprietary research of Macrosynergy Ltd. It ties in with this site’s summary of systematic value generation based on macro trends.

Why household and consumption trends matter for equity markets

The most common equity valuation approach is the dividend discount model. It estimates the fair value of a stock as the discounted present value of expected future dividend payments. This implies that a stock price should be negatively related to expected real interest rates and positively to earnings growth. Both forces depend on the macroeconomic environment in general and on the quality of economic growth in particular.

In this post, we focus on the conditions for private households and consumption. Strong household spending, labor income, and low unemployment shape the equity-relevant macro environment in two principal ways:

  • Consumer strength bodes for strong aggregate demand and increasing price pressure, lessening the need for monetary policy support and driving real interest rates higher. This is particularly important in the context of business cycles. In mature cycle stages with high operating rates and signs of overheating, high household income and spending growth often lead to monetary tightening.
  • High wage growth and tight labor markets are indicative of a rebalancing of national income from capital to labor. All other things equal this reduces earnings growth. Empirical research shows that long-term trends of equity return over decades have been heavily influenced by allocation between labor and capital income (view paper here).

The proposition is that the quality of growth matters for equity markets. Strong household spending, wage growth, and labor markets mark, all other things equal, unfavorable macro conditions for equity performance. Conversely, strengthening corporate profitability, cheap labor, and monetary policy support would mark favorable conditions.

Market participants do not plausibly exploit household conditions systematically across countries. This merely reflects that investment managers cannot continuously process and act on all information, as explained by the theory of rational inattention (view post here). Following and evaluating macroeconomic data, beyond short-term surprises at release, is unpopular and tedious. This means that a systematic approach that uses meaningful point-in-time information on the state of consumer strength holds promise for predicting equity market performance.

A basic set of quantamental indicators

To assess the value of consumer-related data for equity market timing, it is critical to work with point-in-time data. Some indicators are released with long lags relative to their observed periods, and almost all are revised over time. Hence, the analyses below are based on indicators of the J.P. Morgan Macrosynergy Quantamental System (JPMaQS) for 16 countries and currency areas with liquid equity index futures for the past 30 years. Economic information of JPMaQS refers always to the state of public knowledge of the latest instance of a measure based on a concurrent data vintage at a daily frequency. The countries chosen for the analysis are Australia (AUD), Brazil (BRL), Canada (CAD), Switzerland (CHF), the euro area (EUR), the UK (GBP), Japan (JPY), South Korea (KRW), Mexico (MXN), Malaysia (MYR), Poland (PLN), Sweden (SEK), Singapore (SGD), Thailand (THB), Taiwan (TWD), South Africa (ZAR).

There is no single statistic that represents the strength of consumer spending and income prospects in a timely manner. Hence, we picked five plausible indicators with complementary advantages that jointly should give a good idea of consumer strength:

  • Excess consumption growth: This is based on point-in-time metrics of real private consumption, seasonally adjusted, % over a year ago, using either quarterly data or 3-month moving averages (view documentation here). Apart from the U.S., these data are taken from the national accounts and hence feature longer publication lags. To derive excess consumption growth, we subtract the past five years’ median of real GDP as a benchmark (view documentation here).
  • Excess import growth: This indicator uses merchandise imports in local currency, adjusted for seasonal effects, working days, and holidays, as % of the last six months over the previous six months at an annualized rate (view documentation here). The choice of 6 months reflects these data’s high monthly or even quarter volatility. Import demand is only partly related to households and reflects both prices and volume changes. However, external trade statistics are released very timely and dependent on nominal consumption, particularly in smaller countries. The natural benchmark for excess import growth is the sum of the past five years’ real GDP growth median plus the effective inflation target (view documentation here).
  • Excess wage growth: This concept is predicated on real-time measures of wages or salary growth % over a year ago, quarterly, or as a 3-month moving average (view documentation here). There is no common international standard for higher-frequency wage data, and JPMaQS focuses on series that are mostly watched by the market. The reference for excess wage growth is the sum of the effective inflation target and past productivity growth, whereby productive growth is proxied by the past five years’ real GDP growth median minus the past five years’ median of growth of the workforce (view documentation here).
  • Excess unemployment rate: We use the JPMaQS concept of unemployment gaps, defined as the unemployment rate, seasonally adjusted: 3-month moving average (or quarterly) minus the 5-year moving average (view documentation here). The adjustment for multi-year averages reflects that natural unemployment rates are very different across countries and time, and that metric provides a convenient presumed neutral zero level for market impact.
  • Excess employment growth: This concept is based on the main measure of employment growth of the country as % over a year ago, quarterly, or as a 3-month moving average (view documentation here). To derive excess growth, we subtract the past five years’ median of growth of the workforce (view documentation here).

The above set of indicators is not necessarily optimal. JPMaQS offers other consumer-related indicators, such as retail sales, consumer surveys, and credit growth. The above set is merely a plausible minimum for testing the proof of concept that point-in-time indicators of consumer strength have been negative predictors of subsequent equity returns.

To make the five macro indicators comparable in terms of variation, they have been sequentially normalized around their zero values, with standard deviation estimates being updated monthly, always considering only past values. The sample data for estimation is always the full panel, i.e., the data of all 16 countries. Also, the normalized values have been “winsorized” (capped or floored) at three standard deviations, meaning that larger absolute values have been set to 3.

Historically, excess real private consumption and nominal import growth scores have broadly tracked similar cycles. However, merchandise import growth has often posted more abrupt changes and mini-cycles, probably due to its closer relation to manufacturing inventory dynamics and commodity price swings. Note that not all indicators go back the full 30 years for all countries, somewhat reducing the value of analyses for the early years in the below analyses.

Labor market-related scores have historically been less volatile than spending scores. However, data quality and availability have been uneven. Quantamental indicators of wage growth were not often available before 2000. Also, wage and labor market indicators in some EM countries, particularly South Africa and Malaysia, seem to be unusually volatile and contradictory.

As a main equity trading signal, we compute a composite consumer strength score as an unweighted average of the five above scores. If one or more constituent scores are missing, consumer strength is measured by the average of the remaining ones. This leads to heterogeneity in the individual country indicators, which may be composites of different constituents in the early years. However, it is also a realistic replication of markets’ information states. For example, if no wage data are available in acceptable form, then wage growth cannot possibly be part of the assessment of consumer strength, and markets will have to form judgments based on the evidence of what they monitor. Once indicators become available, they should be added to the information set.

The cross-country correlation of consumer strength has almost universally been positive except for some relations between Brazil and Mexico and the Mexico-South Africa relation. This means that consumer strength does not create uncorrelated cross-market signals but tends to track a global trend with local specifics.

Evidence of predictive power

We test if there has been a significant relation between the composite consumer strength score and subsequent equity index future returns for the panel of 16 countries since 1995. Specifically, we postulate that stronger consumer-related data reduce subsequent returns in absolute terms and relative to a country with weaker data.

The critical test for this purpose is the Macrosynergy panel test (view post here), which uses the significance statistic of feature coefficients in a panel regression with period-specific random effects. This test draws on the experience of all currency areas while recognizing that country experiences are not independent. Simply pooling data can lead to “pseudo-replication” and overestimated significance of correlation. Panel regression models with period-specific random effects adjust targets and features of the predictive regression for common (global) influences. The stronger these global effects, the greater the weight of deviations from the period-mean in the regression.

We applied this test to the relationship between the consumer strength score at period end, and subsequent equity index returns at weekly, monthly, and quarterly frequencies. For either frequency, the predictive relation has been negative, with a probability of significance near 100%. The negative correlation was stronger for emerging countries than for developed markets but held for both groups.

The negative relationship also held for half samples, i.e., for the 1990s/2000s as well as for the 2010s/2020s, albeit significance was a bit lower for the second half of the sample.

Across the constituent scores of consumer strength, 4 of 5 indicators individually posted highly significant negative (for unemployment positive) relations with subsequent equity index returns, as measured by Pearson and Kendall (non-parametric) correlation coefficients. The exception was excess wage growth. The predictive correlation of wage growth was positive according to the Pearson correlation coefficient but near zero according to the Kendall statistic.

There has been a negative predictive correlation across all countries excluding South Africa, according to Pearson, and all excluding South Africa, Mexico, and Japan, according to Kendall. Almost half of the countries posted significant positive Pearson correlations just based on their own data.

Example of value generation through timing

To check the economic value of the predictive relation for a simple equity overlay strategy, we generate naïve PnLs for a simple equally weighted cross-market equity index futures position and one that is managed by the consumer strength score. For a managed long-biased strategy, we simply use the country scores, add one standard deviation for the long bias, and winsorize at three standard deviations. This means that strategy signals have an asymmetric range between -2 and 3, determining direction and leverage in accordance with recorded consumer strength. Positions are updated at the beginning of each month based on the signal at the end of the previous month and allow for a 1-day implementation delay for trading. The naïve PnL does not consider transaction costs or compounding because those depend on assets under management and institutional rules. Also, for presentational purposes, all PnLs have been scaled to 10% annualized volatility.

A simple equally weighted long exposure across all 16 country futures would have produced a Sharpe ratio of 0.54 and a Sortino ratio of 0.75 over the past 30 years. While the cross-country allocation has produced some diversification benefits, the correlation with the S&P500 benchmark is still very high at 62%. Managing the exposure with the consumer strength score would have lifted the Sharpe ratio to 0.8 and the Sortino ratio to 1.18 while lowering the correlation with the S&P500 to 44%. This would have been a material improvement.

However, the outperformance of the managed PnL has been seasonal and concentrated on times of serious economic fluctuations. There have been three noteworthy episodes: the early 2000s (often associated with the dot-com bubble and bust), the great financial crisis, and the Covid pandemic. The managed strategy mostly mitigated the drawdown into the equity market downturn and always outperformed during recovery. The drawdown mitigation worked in the 2000s when equity market downturns were preceded and aggravated by late-cycle economic dynamics with strong labor markets and household spending. It did not help at the outbreak of COVID-19, which had no relation to the business cycle. However, in all cases, the weak position of consumers and soft labor markets after the respective crises correctly heralded strong equity returns, presumably due to its impetus for monetary policy support and a turnaround in corporate profitability. The flip side of the outperformance in turbulent times has been gradual mild underperformance in quieter periods, particularly in the later stages of the business cycle, as conditions for equity markets were deteriorating, but markets were still holding up.

If the managed long had used the individual constituents of the consumer strength score, all would have outperformed the long-only, with Sharpe ratios of 0.56-0.79 and Sortino ratios of 0.85-1.15. The excess wage signal was the weakest and only tradable from 2000 onward.


Ralph Sueppel is managing director for research and trading strategies at Macrosynergy. He has worked in economics and finance since the early 1990s for investment banks, the European Central Bank, and leading hedge funds.