HomeMacro TrendsEquity versus fixed income: the predictive power of bank surveys

Equity versus fixed income: the predictive power of bank surveys


Jupyter Notebook

Bank lending surveys help predict the relative performance of equity and duration positions. Signals of strengthening credit demand and easing lending conditions favor a stronger economy and expanding leverage, benefiting equity positions. Signs of deteriorating credit demand and tightening credit supply bode for a weaker economy and more accommodative monetary policy, benefiting long-duration positions. Empirical evidence for developed markets strongly supports these propositions. Since 2000, bank survey scores have been a significant predictor of equity versus duration returns. They helped create uncorrelated returns in both asset classes, as well as for a relative asset class book.

The below post is based on proprietary research of Macrosynergy.

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.

This post ties in with this site’s summary of systematic trading strategies based on macro trends.

Equity and duration returns since 2000

In this post, we compare the risk-adjusted returns on equity and duration positions in developed market countries that publish bank lending surveys: the U.S., the euro area, Japan, the UK, and Canada (“DM5 basket”). As targets of the analysis, we choose derivatives positions in both asset classes for 10% annualized standard deviation targets. Positions are scaled to a 10% volatility target based on historical standard deviations for an exponential moving average with a half-life of 11 days. They are rebalanced at the end of each month with a constrained maximum leverage (notional to risk capital) of 20. Returns are expressed in % of the risk capital on positions scaled to the volatility target.

For the equity leg, we use local-currency returns on the front future of main equity indices, i.e., the Standard and Poor’s 500 Composite (USD), EURO STOXX 50 (EUR), Nikkei 225 Stock Average (JPY), FTSE 100 (GBP), and the Toronto Stock Exchange 60 Index (CAD). View full documentation here. For the duration side, we use returns on 5-year interest rate swap fixed receiver positions assuming monthly roll. View full documentation here. The country returns are combined into developed market basket returns with equal weights.

The chart below shows the evolution of the cumulative volatility-targeted returns for the two asset classes. Total returns since 2000 have been of similar magnitude for equity and duration exposure. However, their profit and loss seasons have been different. Equity returns outperformed notably from 2003 to 2007, 2016-2018, and from 2021. Duration returns outperformed in the early 2000s, around the great financial crisis of 2008-09, and towards the end of the 2010s. Both asset classes posted strong returns in tandem after the great financial crisis during the time of non-conventional monetary policy support, which featured large-scale asset purchase programs, long-term refinance operations, and interest rate forward guidance.

Bank lending surveys and related quantamental scores

In developed markets, five currency areas have been publishing bank lending surveys:

  • In the U.S. the Federal Reserve has produced the “Senior Loan Officer Opinion Survey on Bank Lending Practices” since 1967. This is a survey of up to eighty large domestic banks and twenty-four U.S. branches and agencies of foreign banks that is published quarterly. View the official reference here.
  • In the euro area, the European Central Bank has released the “Euro Area Bank Lending Survey” quarterly since 2003. The survey is addressed to senior loan officers at a representative sample of euro area banks. Currently, this includes around 158 banks representing all euro area countries. View the official reference here.
  • The Bank of Japan has published the quarterly “Senior Loan Officer Opinion Survey on Bank Lending Practices at Large Banks” since 2000. The survey takes responses from the 50 largest banks. The survey covers all lending activity to UK resident households and corporations. View the official reference here.
  • The Bank of England has produced the quarterly “Credit Conditions Survey” for the UK since 2007. The survey is intended to cover all lending activity to UK resident households and corporates conducted by lenders resident in the UK, which has been around 20 recently. View the official reference here.
  • The Bank of Canada has published the quarterly “Senior Loan Officer Survey of Lending Conditions” since 1999. It covers a select group of about 10 banks and, unlike the other surveys, only provides information on credit supply conditions, not on demand conditions. View the official reference here.

We focus on two key metrics of these surveys: changes in bank credit demand and changes in credit supply, where the latter is usually approximated by bank lending conditions.

For a meaningful analysis of the predictive power of bank lending surveys for financial markets, one needs to look at available information in real-time. This is simple with point-in-time quantamental data, particularly the indicators of the J.P. Morgan Macrosynergy Quantamental System (“JPMaQS”). Generally, quantamental indicators are real-time information states of the market concerning an economic concept and are designed specifically for backtesting and operating macro trading strategies.

For this post, we use daily information states of real-time standardized bank lending survey measures of credit demand and supply conditions (view documentation here). Like all quantamental indicators, these are based on vintages, i.e., concurrent historical versions of standard time series. Moreover, survey indices vintages are sequentially standardized using historical means and standard deviations on the survey level. Such standardization is important to replicate the market’s information state on what was considered normal regarding the level and deviation of a survey metric and to make it comparable across countries.

The bank lending survey scores of credit demand and supply conditions are then aggregated over the five developed countries that produce such survey metrics, using the concurrent dollar values of their GDPs as weights (view documentation here). For credit demand, only four countries can be used since Canada does not take a loan demand survey.

The chart below shows the developed market-weighted average of credit demand and supply conditions according to bank lending surveys. A positive value means improving demand or supply conditions. Since 2000, the developed market survey scores show almost four credit cycles. Demand and supply conditions have been positively correlated. However, they have displayed different short-term dynamics, and supply conditions have, on average, been viewed as more restrictive than demand.


Why bank lending conditions matter for equity and duration returns

Improving demand for bank credit supports aggregate demand in the economy and the creation of leverage in the financial system. Both bode well for corporate profitability and forthcoming earnings reports. For financial companies, more credit directly translates into higher revenues. Hence, all other things equal, and as long as markets are not fully information efficient, signs of strengthening loan demand should positively predict equity returns. At the same time, increasing loan demand directly puts upward pressure on longer-dated yields and reduces the central bank’s need to be accommodative or even to tighten monetary conditions. This bodes ill for returns on low-risk fixed-income positions, such as swap receivers or high-grade government bonds.

The theoretical effects of easing credit supply conditions are similar, albeit with a slightly different emphasis. Improving credit supply conditions also supports the demand for goods and services in the economy and the build-up of leverage. Hence, credit easing should predict equity returns positively as well. However, the impact of more accommodative credit conditions on fixed-income markets is a little less clear. If such easing results from circumstances other than monetary policy or interest rates themselves, it should invite more restrictive monetary policy and be negative for low-risk fixed-income returns. However, if credit easing reflects a more supportive monetary policy itself, it may have no predictive power for the future.

Predicting equity returns

We empirically investigate the relationship between bank survey scores and subsequent equity index basket returns for the developed markets in aggregate.

Regression analysis of the relation between the bank survey demand score and subsequent monthly equity index future returns shows a clear positive relation with very high significance. Also, accuracy and balanced accuracy of the prediction of the direction of monthly returns have been above 56% since 2000. This is remarkable given that the credit demand signal posted a slight short bias, while equity indices posted positive returns in 60% of all months.

Predictive correlation has been even a bit stronger between bank lending conditions and subsequent monthly equity index returns. The relation has been highly significant both for parametric and non-parametric correlation coefficients. Accuracy and balanced accuracy of the prediction of the direction of monthly returns have been just above 56% since.

To assess the value generation of bank survey predictors over time, we calculate stylized “naïve PnLs”, i.e., dollar-based profit and loss developments over and above funding costs according to the standard rules of Macrosynergy research posts.

Positions are taken in vol-targeted equity index future positions across the five developed countries following the survey score signal for the GDP-weighted basket of the five countries. Here, we consider proportionate positioning, in accordance with normalized signal and a limit of 3 standard deviations, and binary positioning, i.e., unit long or short positions in accordance with the sign of the survey score change. Positions are re-calculated monthly at the end of the month and re-balanced at the beginning of the following month with a one-day slippage for trading time.

The long-term volatility of the PnL for positions across all currency areas has been set to 10% annualized. This is no proper vol-targeting but mainly a scaling that makes it easy to compare different types of PnLs in graphs. A naive PnL does not consider transaction costs or realistic risk management rules, which would depend on portfolio size and the institution that is trading. Its purpose is to inform merely on the value generation of the factor.

The long-term Sharpe ratios of naïve survey-based PnLs would have been between 0.3 and 0.4 with no correlation to equity index benchmarks. Equity market value generation has focused mainly on the financial crises of the early 2000s and late 2000s. Lending demand scores have also created some value in the second half of the 2010s. On the other hand, bank lending surveys did not help manage equity during the COVID-19 pandemic.

This illustrates the natural seasonality of bank lending scores as an isolated trading signal: for credit conditions to have an impact, they need to change noticeably. Put simply, bank lending scores matter in financial system downturns (or crises) and recoveries, less so in quiet times.

Since survey-based PnLs are uncorrelated with the broad equity market, using naïve bank survey scores to enhance long-biased equity portfolio management would have added significant value. The below PnLs approximately add up returns of long-only and survey-based positions in equal weights to produce long-biased portfolios. Whilst the long-term Sharpe ratio of a long-only portfolio would have been 0.47, the Sharpe ratios of the combined positions for either credit supply or demand signals would have been 0.62-0.72.


Predicting fixed-income returns

Next, we investigate the relationship between bank survey scores and subsequent 5-year IRS fixed receiver returns for the developed markets in aggregate.

As expected, bank survey scores have negatively predicted fixed receiver returns in the interest rate swap market. However, the relation has been less strong and significant than with equity index performance. The negative predictive correlation of credit demand scores with subsequent monthly swap receiver returns has still been quite clear, with 95% probability of significance. Also, the monthly accuracy and balance accuracy of the prediction of return direction has been above 53%.

Meanwhile, the relationship between bank lending supply conditions and fixed receiver returns has been tenuous and insignificant. This fits the theoretical argument that improving supply conditions may themselves reflect easing monetary policy and declining yield trends and, hence, are unreliable as an indicator of future policy and market responses.

While empirical relations for fixed income are less strong than for equity, they confirm the fundamental proposition of this post: bank lending conditions drive a wedge between equity and fixed-income performance.

Naïve PnLs for IRS receiver positions based on survey scores suggest that bank loan demand scores would have added significant value to a fixed-income portfolio. Sharpe ratios of credit demand-based strategies would have been around 0.4 since 2000, with around 10-15% correlation with U.S. treasury returns. Bank lending supply scores have not been a convincing value generator. With long-term Sharpe ratios of 0.1-0.4, their PnL contribution has been less robust to different signal versions.

Predicting relative returns

Finally, we look at the relation between bank survey scores and volatility-targeted equity versus duration returns for the developed market basket.

Bank loan demand scores have displayed a strong and significant positive correlation with subsequent monthly relative returns, with a  slightly higher coefficient than for the individual asset class predictions. Accuracy and balanced accuracy of monthly return predictions have also been a bit higher at 57%.

Bank loan supply scores have likewise been positive predictors of equity versus duration returns, albeit their probability of significance has been a bit less than 99%. Also, monthly accuracy measures have been a little lower at 56%.

The long-term value generation of an equity versus duration portfolio has been material but quite seasonal. For most naïve PnLs, value generation focused on the financial market crises in the early and late 2010s. Quieter periods and the pandemic-induced cycles of the 2020s have seen little or no positive PnL generation. This confirms the natural seasonality of bank survey scores: they are less interesting without pronounced financial sector cycles. This argues for combining the factor with other signals more suitable for “normal times”.

One can combine the bank survey signal with a long equity versus duration portfolio. Giving equal weight to both, one gets a long-biased book that is normally long equity index futures and pays rates, extracting the equity premium without systematic exposure to inflation risk and monetary tightening. That position can be managed by the bank survey scores: in times of strong credit demand and supply, the equity long and duration short would be particularly large. In times of weakening credit, the equity long would be taken down or even turned into a short, and the fixed income exposure adjusted conversely.

A long equity-biased equity portfolio of this type looks particularly attractive in times of higher inflation and low real interest rates. Bank survey-based management would have added significant value. Since 2000, the Sharpe ratio of the managed portfolio would have been between 0.3 and 0.5 versus just 0.1 for the static portfolio.

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.