Monte Carlo Simulations
Out of curiosity, does anyone actually run Monte Carlo simulations when looking at deals? Debt or Equity side. If so, what asset class and firm size?
Trying to introduce another layer into underwriting but wanted the general Ape take.
I work in acquisitions, if an analyst did this at my firm we'd probably laugh at him. Not trying to be rude, always for improving data/underwriting/etc, but I see zero application to CRE, it's a simple industry, no need to overcomplicate it. Of course, if it helps you, by all means proceed.
That said, if you find an interesting and useful application, always happy to take a look and implement in my own models.
Don't have too much experience with this but intuitively:
The entire industry is heavily levered with various covenants.
Single value assumption models only show you the "median outcome" not the expected value, which rewards higher leverage. If you model in the scenarios where cap rates or interest rates rise, you will eventually get outcomes where you have exhausted all available additional equity you can scrape together from LPs, won't be able to further amortize the loan and will have to default and be forced to sell, which of course destroys your returns.
Also if the distribution of outcomes are asymmetric for a decision or clause:
For example a development loan that extends a couple years post completion. The agreement has max LTV AND LTC covenant after completion vs. just LTV, your median outcome or worse scenarios wont show much of a difference but all of your above median outcomes get significantly worse (less debt, more equity which you could have distributed already on completion or used on other developments).
If you calculate the weighted average of these outcomes you will get the expected return, which will be wildly different than your median outcome return.
Wouldn't this warrant at least for a sanity check that you're not missing anything major? What do you think?
I did it once is school. never in industry. Had a director ask about it to incorporate into modeling and nothing ever happened. It's a cool concept but not any real applicability. Probably because a sensitivity table gets the job done in a round-about way.
You can overcomplicate real estate as much as you want, but I don’t.
I’ll throw in a table that computes returns at different .25 bp exit cap variations if I’m feeling fancy, right up there with a table that computes how .25 bp interest rate jumps impact my interest reserves, but I try not to get too wrapped around the axel on financial predictions.
Ultimately a deal comes down to execution, not excel wizardry.
Quick explanations with everyone on the same page. No need to reinvent the wheel.
Thanks everyone!
Don’t get me wrong though - it’s a cool idea. Don’t ever stop trying to think of cool ideas or ways to improve processes.
I’m still hoping in a few years I can have AI underwrite a deal for me from start to finish.
If people have to google what that even is, it might not be very applicable. What is that anyways?
Running a simulation ~10,000 times based on statistical outcomes and then looking at the distributions of outcomes
Turns a bunch of inputs into a single output number. Each variable can be randomized according to its own distribution. Then the output is a single distribution.
eg. Cash flow is dependent on A, B, C, D, E. 20% chance A is low. 50% it’s mid. 30% its high. Then do same for BCDE with individual probabilities. Monte Carlo spits out a single answer like this: CF is too low 10%, CF is low, 20%, CF is mid 40%, CF is way above hurdle 30%. Then you don’t have to consider the chances individually and try to get a feeling. It will run thousands of times and spit out a quantity.
Essentially it tells you how sensitive you are to each risk. It’s especially useful in a business like oil where you are reliant on a lot of outside factors. Does Kamala win the election? How much oil is being produced? How are rates changing?
The only Monte Carlo simulations I run are deciding where to dock my super yacht on F1 race day.
Hahaha - amazing!
The responses scream to short real estate. You have an entire agreement across the board that getting a full picture of a distribution of scenarios and getting at least a more accurate picture of VaR is a waste of time?
What specifically would you do with that knowledge that you otherwise would be unable to?
Quantifying the possible risks. Showing your LP's the possible weakness areas and showing the area under the curve and the implied impact. Also stress testing multiple variables and incorporating correlation analysis in your simulation.
IMO, all this does is add another layer of complication and potential error into the decision-making process. If you're choosing something insanely simple, say between two types of mortgages on your home, would you build a Monte Carlo simulation for this? Of course not, not only is it overkill and a waste of time and energy, but it's another layer of decision-making that could cloud your judgment. Real estate is inherently a simple business, sometimes it's better just keep it that way. For every deal I have worked on personally, there are 3-5 key levers that can be easily tested with a few scenarios and sensitivity tables, this is inherently better than introducing new probability weighted assumptions about additional outcomes.
Now to address your comment about shorting real estate, if an industry is unsophisticated, that doesn't necessarily mean the asset class is overvalued. If I were you, i'd run a Monte Carlo to verify that assumption and see what the outcome distribution looks like.
I think you're misunderstanding something here. Just because people think a Monte Carlo simulation isn't the right modeling tool doesn't mean that they don't have a sufficient risk analysis. Real estate is not a complex business. There are basically 3-4 metrics that matter and they're all easily encompassed in a few data tables that are more than sufficient get comfortable with your risk. You arrive very quickly to the point where the juice is not worth the squeeze with things like a Monte Carlo simulation.
Did this once to "impress a pension fund client." Was interesting to run, but ultimately was just a bell curve around our projected returns..
Retweet.
I don’t see any value from a Monte Carlo simulation when it comes to a specific deal. I work for an LP, we are not a pension fund that allocates to LPs, just an actual LP that has SMAs and open and closed end vehicles. I can see value from a Monte Carlo simulation applied to a whole series of markets perhaps, but if that’s the starting point for an investment thesis, something is wrong
Libero ullam asperiores necessitatibus quia. Quam accusantium culpa et officiis enim laborum deserunt laboriosam. Quo dolorum consequatur natus laboriosam dolorem aperiam. Maiores ducimus et ratione repudiandae id sed ut.
Distinctio quae architecto sint ut. Veniam adipisci illo nostrum maiores cupiditate. Ex nihil asperiores non et qui ut facilis. Fugit laudantium aut vitae consequatur nulla dolor.
Magnam laboriosam accusantium fugit. Ea laudantium dolor adipisci rem. Sit impedit voluptas exercitationem quia culpa recusandae sit molestiae.
Similique beatae laboriosam eveniet et dicta. Aut dicta alias et quia ut consequatur temporibus occaecati. Omnis voluptatibus asperiores quis esse consectetur. Et quas quod est adipisci ullam. Maiores nesciunt dolorem possimus libero neque consectetur ab. Voluptatem quia nemo non dolores aut.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...
Qui incidunt neque et nulla distinctio nulla nihil. Et repellendus esse a nesciunt odit velit praesentium. Ipsam voluptates pariatur reprehenderit consequatur quisquam eius iure ea. Quidem rerum assumenda dolores beatae facere.
Ullam asperiores soluta ex enim sunt sit. Eos aut et sint quasi incidunt. Excepturi quo et voluptatum mollitia ex veniam cupiditate ratione. Sint non aliquid consequatur adipisci eaque numquam et. Mollitia est placeat voluptas magnam recusandae.