Understanding Survivorship Bias In Banking

A more profitable loan portfolio

It is 1942,  you are in Air Force command, and you want to keep our flyers safe. Our planes are being shot down at an alarming rate, and your solution is to install armor. But the armor makes the plane heavier, and heavier planes are slower, less maneuverable and use more fuel.  Realizing you cannot armor plate the entire plane, the question arises what is the optimal amount of armor and where should it be placed to give our soldiers the best chance to complete their mission alive? The methodology employed by this effort is an interesting example of the “survivorship bias” and makes for a valuable banking lesson.


What is the Survivorship Bias?


The Survivorship Bias is an error in logic caused by the concentration of on outcomes that made it past some selection process and some outcomes that do not make the selection process.  Because of the lack of visibility, false conclusions are drawn.  This has been happening at some banks for decades because of one very critical missing tool that has not allowed banks to have greater visibility.


Example of Survivorship Bias


Abraham Wald was on the Statistical Research Group (SRG) in World War II.  The SRG was a collection of best American statisticians and was something like the Manhattan Project, except the weapons being developed were math and logic, and not bombs.  The SRG notably included Norbert Wiener (creator of cybernetics), and Milton Friedman (Nobel prize winner in economics), but the smartest person in the room was Abraham Wald.


The military supplied the SRG scientists with data on the number and pattern of bullet holes on planes that came back from engagements.  The damage was not uniformly distributed across the aircraft.  There were more bullet holes in the fuselage and wings, and fewer bullet holes per square inch on the engine (below).  





The officers in the military started installing armor on the places with the greatest need – where bullets were most pervasive (fuselage and wings).  But they wanted the SRG to solve for the correct amount of armor given the weight and strength of the armor.  That is where Wald provided an answer that the military did not expect.  The armor, said Wald, does not go where the bullet holes are. It goes where the bullet holes aren’t: on the engines.


Wald reasoned that the planes that were coming back with fewer hits to the engine were the ones that survived and that the planes that got hit in the engine were not coming back for the military to observe.  The planes returning to base with bullet holes in the fuselage and wings were strong evidence that hits to the fuselage and wings can (and therefore should) be tolerated.  The military fell for a classic example of survivorship bias leading to the wrong conclusions. 


Survivorship Bias in Banking


A similar phenomenon has been at play in banking for decades.  The return on equity (ROE) for banks that are under $1B in total asset size and $1B to $3B in assets are currently 9.36% and 10.91%, respectively.  Many bank owners, board members, and managers are questioning the return on investment and demanding higher ROE.  One way to increase returns is on earning assets and, more specifically, loans.  If the current loans in the portfolio are not generating the required ROE, then different loans must be considered.  But which loans should bankers prioritize?


The problem for many banks is that the general performance of all loans is measured quarterly on financial statements, but the ROE of individual loans going into the portfolio is not measured at inception or through the life of the credit.  The loans that might garner the bank a 15% ROE are not in the portfolio because if bankers measure loan yield only, paradoxically those loans collectively return the bank nine to ten percent ROE.  Managers are then challenged to increase the bank’s ROE, and choose loans with even higher yield, but, unfortunately, lower collective ROE.


There are a few common loan factors that drive ROE that is not tied to yield, and, in fact, are inversely correlated to yield.  These factors are as follows:


  1. Diversification comes at a price. Banks may need to sacrifice pricing in exchange for greater diversification so that the loan portfolio can perform better over time. Banks with high commercial real estate concentrations need to price to gain credit exposure to healthcare, consumer products, energy, agriculture and other low correlative sectors.
  2. Higher credit quality loans will generate lower yield.  Almost every term loan a bank books today will probably endure a recession.  Now is the right time to give up yield for lower credit default risk for the next few years. While many loans perform the same when times are good, better credits greatly outdistance weaker credits in times of economic stress.
  3. Size of the loan does matter.  For most banks, and most types of credits, commercial loans below $1mm in size are less profitable than larger credits.  Certainly, few banks can make commercial loans below $250k profitable.  Higher loan balances are priced with a lower yield.
  4. Relationship matters more than yield.  Loans must provide cross-sell to deposits, referral opportunities, fee business and/or treasury management services. Banks can tolerate a lower yield for a higher ROE relationship.
  5. Lifetime value of a relationship is key.  In the banking industry, with the high initial cost of customer acquisition, a client becomes profitable over longer periods of revenue generation.  Bankers must focus on sticky relationships with longer retention rates.


Banks that quantify loan yield, but nothing more, are puzzled why their portfolio of loans is not more profitable, but it is a problem with survivorship bias.  The profitable loans never make it to a bank’s balance sheet because management only vets for loan yield.




The risk facing bankers is not always straightforward and easy to solve.  But with some logic, industry insight and a good risk-adjusted return on capital (RAROC) relationship pricing model, bankers should be able to avoid survivorship bias and book loans that have an impact on what matters – ROE during good economic times and bad.