English
German
Bank market valuations have been largely disappointing for the last twenty years, particularly in Europe. However, there have been some welcome exceptions. To determine those factors that have the greatest impact on this variation in valuation, zeb carried out an in-depth analysis of the top 100 global banks.
While the results we unearthed are certainly complex, our main takeaway is that bank leaders seeking to maximize their institution’s price-to-book value should first and foremost concentrate on balance sheet management, while ensuring that risks are kept under control.
The economic repercussions of the COVID-19 pandemic have had a further adverse effect on banks’ already underwhelming market valuation. In recent times, European banks in particular have tended to lag the market by a significant margin. Their price-to-book (P/B) ratios have consistently been below one, indicating that investors have lost confidence in their ability to generate shareholder value.
Interestingly, however, banks in the United States, as well as some European institutions, have been able to buck the trend and reach reasonable valuations. Our study seeks to establish whether there are certain drivers or market trends that explain this variation. Armed with this knowledge, bank leaders could potentially manage their institution’s valuation in a more systematic way.
Global 100 study of bank market valuation: methods and key conclusions
Our approach was to analyze the relationship between the P/B ratio of the 100 largest global banks and a number of specific variables. We also took into account boom and bust periods and whether the banks were in developed or emerging markets.
The first step in our study was to use traditional panel data statistics. The selected variables, based on current academic research as well as our own expertise, comprise several banking key performance indicators (KPIs) in addition to macroeconomic and market indicators from 2003 to 2020. The KPIs include return on equity (ROE), return on assets (ROA) and on risk-weighted assets (RORWA), cost-income ratio (CIR), non-performing loan (NPL) ratio, leverage, growth in total assets or revenues, and risk-weighted assets (RWA) density. The macroeconomic variables include GDP growth in the relevant country, GDP per capita, number of banks per country (as a proxy for the banks’ competitive environment) or inflation (as a proxy for interest rate movements).
Some notable initial conclusions emerged from these traditional regression models. Although the results differed according to the individual periods, certain fundamental banking KPIs – ROA, NPL ratio and RWA density – were particularly significant determinants of banks’ P/B ratios over time. However, the picture was far from uniform. Other variables became more relevant when certain economic conditions prevailed. ROE or leverage helped to explain bank valuations in a boom cycle (2003-2006), while a Tier 1 ratio played a more consequential role during a crisis period (2007-2013), and total asset growth was a useful indicator in the recent post-crisis period (2014-2020). Another finding was that the usefulness of our model increased over time. The model can explain only 22% of the total variation in P/B ratios for the 2003-2006 period, but 43% for 2007 to 2013, and nearly 60% for 2014 to 2020.
Regional differences were also apparent. In our study, we focused primarily on banks from developed countries. For the emerging market banks in our sample, certain factors, such as ROE and real GDP per capita, assume a more decisive role. Despite leading us to invaluable conclusions, there are also some disadvantages to the traditional regression models we used in the first instance. Calibrating such models is time-consuming, involves a degree of subjectivity and requires several robustness checks to test the validity of statements.
As a second step therefore, we used state-of-the-art supervised machine learning techniques in an attempt to overcome some of these drawbacks, while at the same time searching for any further explanations of bank valuation that had yet to be brought to light. This method confirmed some of our initial results. Over time and across regions, the ROA, the NPL ratio and the ROE were consistently strong predictors of banks’ P/B ratios. Meanwhile, inflation and real GDP growth were generally the most important macroeconomic variables.
However, this machine learning step also revealed other variables, such as RORWA, CIR, Tier 1 ratio, total asset growth or liquidity (taking into account deposits, long-term funding and equity to total assets) to be relatively important factors, although only for certain periods and/or different regions. For example, total asset growth and RORWA went some way to explaining P/B ratios during the crisis and post-crisis periods, but were not relevant for the 2003-2006 boom period. The Tier 1 ratio was also shown to be relatively salient during the crisis period, while liquidity grew in importance during the post-crisis period.
We used the same bank KPIs and economic variables to derive sample predictions of the P/B ratio and test the validity of the machine learning exercise. The results were striking, with a predictive accuracy on average of between 80% to 90%, clearly demonstrating the effectiveness of the machine-learning approach.
We can derive several pertinent conclusions from this research. First, data models and statistics are clearly helpful tools for explaining or predicting banks’ P/B ratios. Furthermore, the explanatory or predictive power of such models is very high, emphasizing the usefulness of data analytics in bank valuation processes.
The highlighted statistical relationships, however, are complex and not predefined. Machine learning methods to interpret the results objectively, and build efficient models, are therefore advisable. The available technology, as well as the abundance of data at our disposal, mean that we can run such analysis on a continual basis. Data-driven procedures should therefore be obligatory for anyone who really wants to understand stock performance in the financial services industry.
Despite the variation in the results, bank leaders can themselves derive some vital conclusions when it comes to managing their institution’s valuation. Balance sheet management - increasing asset or RWA productivity – should be their focus. At the same time, risks should be kept under control. Banks need to select the right deals and counterparties.
Another important conclusion is that cost efficiency can only partially explain the P/B ratio and needs to be combined with overall higher profitability.
It is important to bear in mind too that GDP growth and inflation, a common proxy for potential interest movements, are key macroeconomic variables. Indeed, most bank stocks appear to be akin to a call option on rising yields and follow the performance of the economy in which the bank operates.
However, leaders should always be aware that the identified relationships may fluctuate over time and according to the relevant region. Therefore, the analysis should be regular, wide-ranging and specific to the bank. In this way, banks can understand how all trends influence valuation and perhaps adjust steering or reporting mechanisms accordingly.
The results of our study are not clear-cut, and in no way provide us with all the answers to the puzzle of bank valuation. But we nevertheless believe that such analytics can lead to more informed and ultimately better management decisions, and all at minimal cost and effort.
