Real Estate Public Relations
What people ARE SAYING about US
"... astute, thoughtful and strategic... in other words, she gets it."
"It is refreshing to work with such a consummate professional."
"... helped define a vision for our business in California."
"... knows how to listen and get the most out of the media."
"Your advice and counsel has been insightful and your work product of the highest quality."
"Your connections and relationships in the building industry are fabulous."
"Timely and precise...creative and insightful...very easy to work with throughout."
"You get results...That's what it's all about."
Real Estate Public Relations Southern California - Anton Communications
 
 

Bank Technology News: Computing Real Estate Value

March 10th, 2008

Computing Real Estate Value

By Barry Gross and Martin Kulli

 

With the housing market in decline, and financial institutions tighter now about where to put their money, developers are often shaky about how to bring about the best value for their real estate. Especially on land obtained through foreclosures, where the original housing intent wound up in failure. They need a jolt of computer reality: The old-fashioned ways of doing business simply won’t produce the best results. What’s needed is an entire computer-general financial model that can track all the variables. And that’s going to mean a greater chance at profit than any gut-feeling approaches. How much greater? In most cases, 20% more profit.

 

Traditional real estate development is mathematically unsophisticated.  Deals are worked out on the back of napkins using rule of thumb analyses to estimate profit, Land Residual Value (LRV).  However, now that every development dollar counts and the new price of land is anyone’s guess, a rough analysis of a property is unacceptable. Developers simply must invest in computer-generated models that can show with far greater precision what product mix might be the most profitable. It gives the developer a far better chance to explore a wide range of more realistic options. It helps replace the guesswork with more precise evaluations of criteria such as potential profit, environmental impact, and financial risk.

 

The traditional method, used by most developers, involves making an educated guess of the best product mix and then intuitively improving LRV by changing the product mix. But the challenge in land development is to decide exactly what to build from an almost limitless set of options – options many developers simply don’t explore with enough depth. They need computer-generated help. There are hundreds of variables and relationships to be evaluated in determining LRV (i.e. unit count, absorption rate, grading requirements, sewer system length and size.)

 

Optimization of LRV means identifying the best product mix using projected market parameters and projected costs — something you can’t put together on the back of a napkin. The solution involves evaluating many variables that interact based upon established relationships and known constraints.  For example, a market research analysis may identify many acceptable product types for a given property.  A land improvement cost analysis combined with the recommended product mix is developed to establish a baseline LRV from which an improved product mix may be evaluated.  Once a baseline is established, a different set of inputs, generally product mix, is tested to determine if these inputs generate a greater LRV. Additional alternatives are generated by the computer after establishing a sensitivity analysis between the new product mix and the LRV. Exploring a wider range of options based on reliable computer data brings new avenues of idea for how to best develop a piece of real estate. Some developers might be amazed how such an approach will make them see their property in a whole new light.

 

In theory, optimizing LRV requires thousands of alternatives be individually generated and reviewed to determine the optimal solution. You can’t produce this just with meetings and rule-of-thumb observations based on previous experiences. Because of the high number of variables involved, this can only be done with computer-based approaches.

 

Existing computer models which optimize LRV primarily use linear regression modeling.  Linear regression relies on discrete data points of known information to generate a continuous linear function used to get results.  Here’s how complicated that can get: When there are hundreds of variables, linear modeling solutions require that each variable be measured against every other variable to determine correlation to LRV.  Once these correlations are understood, a second educated guess of product mix would be made and tested to determine if the LRV had increased.

 

But even this is not enough. The use of linear modeling reduces complex data to a linear function, but it does not adequately address step functions, discontinuities, and non-smooth relationships, which are inherent in development choices.  These “optimization methods” primarily deal with revenues. But developers need answers beyond revenues, a method that coordinates revenues and costs.  Ignoring these nonlinear relationships, current models are limited to finding local maximum solutions and unable to find the global maximum.

 

Too often the older, traditional approaches fail to consider significant non-linear (“Step Function”) byproducts of development. For example, consider the case of a development’s impact on local schools. The developer wants to put in 4,000 units. The local school board policy states that if a project generates more than 1,000 high school students, the developer must construct a new high school. That means an additional $50 million in costs. However, if the project generated less than 1,000 high school students, the developer would be required to pay a $2.50 school mitigation fee per square foot of house.  The average house in the subdivision was about 2,500 square feet, therefore, if the development generated only 990 high school students they would be required to pay $6,250 per unit. Compare that $25 million cost to that $50 million for a new high school.  The inference is that the generation of an additional 10 high school students (about 40 homes) saddled the developer with an additional $25 million in cost. Optimization financial modeling lays that out for developers, saving them millions they might have spend using more traditional approaches.

 

Since the function describing profitability is non-linear, genetic optimization modeling is needed to determine the most effective product mix and maximize the land residual value.  Using mathematical models, genetic optimization algorithms, regression analysis, and solid financial accounting practices to compare development options can save tens of millions of dollars and increase LRV.  To develop this system, one needs to design a financial model coordinating revenues and costs, while providing the flexibility to simulate different scenarios.

 

Then using a Visual Basic for Applications (VBA) program, adjust the variables with a genetic mutation algorithm.  The genetic mutation algorithm follows the following principles:

 

Once a base land residual value is determined, the program generates 99 different alternatives.  All 100 scenarios are ranked by order of increasing LRV.  The program then keeps the most profitable 10 alternatives.  The program then begins a second iteration where the 10 solutions crossbreed and mutate generating an additional 90 scenarios.  The 100 scenarios in the second group of are ranked and the program then keeps the new best 10 alternatives.  Each time the Genetic Algorithm ranks various scenarios, it is learning which inputs have the greatest effect on increasing LRV.  In effect, it is “learning” the sensitivities of each Input variable.  In a typical project, this  will generate approximately 500,000 various alternatives which will compete with one another, until they “evolve” into the best solution.  At the end of the exercise, the most profitable solution that meets all of the constraints is chosen for the product mix.  As a result, the program will find optimal solutions to virtually any type of problem, from the simple to the most complex.  Try doing that on the back of a napkin.

The resulting program can be huge success improving projects, increasing land values over 20 percent on average.  One case optimization recommendations increased profitability from $300 million to over $480 million (60%).

 

About the Authors

Barry Gross is founder and president of Developers Research, a real estate consulting firm based in Irvine, Calif., that evaluates the value and risk of land acquisition, disposition and development for all types of property. Martin Kulli is a project director who specializes in cash flow, profitability and optimization analyses. Through genetic algorithms and regression analysis, Developers Research is able to produce a detailed comparison of multiple development and land residual value scenarios. For more information, go to www.dev-res.com or call 949-861-3300.

 
 

Anton Communications |