Bullet Holes and Vacancy

In prior blogs we have discussed the high profit margins that landlords receive on renewal leases. The greater profit is due to the lower costs a landlord incurs in renewing a tenant compared with having to find a new one. The savings a landlord realizes in renewing a tenant fall into three main categories:

  1. Rent Abatements. Renewing tenants typically get little or no free rent; whereas new tenants can get as much as one month for every year of their term.
  2. TI Allowance. Renewing tenants typically get less than half the tenant improvement allowance offered to new tenants.
  3. No Rent Interruption. But by far the biggest savings to landlords in doing a renewal deal is avoiding the rent interruption caused by the time it will take to find a new tenant – in the industry the concept is referred to as “time on market”.

Before we can really understand the landlord’s exposure to “time on market”, we need to understand statistical biases. A great example can be found in World War II.

World War II was fought on many fronts, including the mathematical front. Unknown to most people even today was the super secret Statistical Research Group which included some of the best mathematical minds of the time. Their job was to use their brains, not their brawn, to win the war.

One of the greatest challenges of the war was to keep the Allies planes in the air. The attrition of planes and crew was appalling. During some campaigns, the chances of the plane and its crew returning were about the same as flipping a coin and calling heads.

Planes needed armament to survive, but where was the most effective place to put your limited armor? If you were given the statistics below on bullet holes in fighter planes returning from battle (these are actual numbers) where would you put the armor?

Section of Plane_001

Based on this data, the US Army reinforced the areas denominated “Fuselage” and “Rest of the Plane”, because that’s where the most bullet holes were. But losses increased. So the problem was sent to the Statistical Research Group, specifically Abraham Wald, a brilliant mathematician who had fled Austria shortly before the Nazi occupation.

Wald observed that holes in the returning aircraft actually represented areas where a plane could sustain damage and still return home safely. Areas that were relatively unscathed, on the other hand, represented critical portions of the aircraft which, if hit, would cause the plane to be lost.

Wald’s solution was the opposite of the U.S. Army’s original gambit. Wald said to put the armor where the bullet holes were not. This strategy was successful and more planes returned from battle.

Wald’s insight was so profound that it was given the name “Survivorship Bias”. In the case of the aircraft, only the survivors were in the sample.  No one saw where the lost planes had been hit. Survivorship bias is one of the many types of sampling biases found in analyzing statistical data.

Survivorship biases are around us every day, and they are particularly alive and well on Wall Street. For example, when looking at the historic performance of mutual funds investing in any segment of the market compared with the direct performance of that same market segment, the average performance of the funds is inflated by the survivorship bias. Mutual funds which failed or became so small as to be statistically insignificant are not reflected in the numbers. A comprehensive 2011 study published in Review of Finance examined the 10 year trailing returns of nearly 5,000 mutual funds. The study determined that survivorship bias increased the average mutual fund performance by nearly 20%.  The funds that failed during the 10-year study period were the planes that were lost on the way back.

So what does survivorship bias tell us about the “time on market” risk that landlords face if they cannot renew a tenant?

Data on time on market is even easier to get than counting bullet holes in a plane. Costar Group, headquartered in Washington DC, is the source that all commercial brokers turn to for market data. With the press of a few buttons I can tell you the distribution of Time on Market for the any submarket in the country. Let’s look at the 42 “Class A” buildings in the West Market Street submarket of Philadelphia:

Graph for blog_001

The Costar database can slice data every which way. So how much time would it likely take a landlord to re-tenant a space in the West Market Street Submarket?  The weighted average time for the above data is 28.4 months – a pretty remarkable number. But that number is wrong, materially wrong.

“Time on market” data suffers from a substantial survivorship bias. But fortunately, once recognized, it’s easy to correct for.  Let’s look at a hypothetical, overly simplified real estate market to illustrate the point. Assume that:

  1. exactly one space becomes available on the first of every month, and
  2. it always takes exactly 12 months to find a tenant for any space.

What would the distribution of “time on market” look like in this hypothetical market if you ran your Costar report on December 31? Well, there would be one space that had been on the market for 1 month (the space that was listed for rent on December 1st); there would be one space that had been on the market for 2 months (the space that was listed for rent on November 1st); there would be one space that had been on the market for three months (the space that was listed for rent on October 1st); and so on. The distribution would look like this:

hypo_001

In other words, the data would be evenly distributed. The weighted average of the above data is 6 months (1+2+3…+12 divided by 12). But wait, we specified in the hypothetical that it always takes 12 months to lease space. Why are we off by a factor of two?

The problem with the above data is obvious: the spaces haven’t been leased, they are in the process of being leased. It’s like counting bullet holes before the battle is over.

From the simplified model we can see that “time on market” statistics underreport the actual vacancy threat by a factor of two.  Landlords have more to lose if they lose you as a tenant than the data would suggest. The 28-month number for West Market Street that we discussed above – as surprising as that number may be – is in reality probably closer to 50 months. Of course more desirable spaces would generally be below that average and less desirable spaces generally above the average. But good, bad or ugly, the statistics grossly understate the landlord’s probable vacancy if they lose a tenant.

When negotiating a renewal with your landlord it’s important to understand the landlord’s expected costs if you move. Lost rent is one of the major costs a landlord incurs in losing a tenant and having to find a new one. Any broker can produce reams of data from CoStar about average “time on market”, just like any buck private can count bullet holes in a plane. What you need is somebody who cannot just report what the data is, but can tell you what the data means.

This entry was posted in Leases. Bookmark the permalink.

Leave a Reply

Your email address will not be published. Required fields are marked *