Commercial real estate history changes. What was once true can change over time. This is not due to politically motivated conspiracies, or even neglect, but due to the nature of the raw data. Most statistics are derived from a set of properties, sampled on a regular basis, to produce a meaningful trend. Sometimes properties are added to the sample to keep up with changes in the structure of the market. Sometimes, these changes are substantial as in the addition of a large geographical area in a growing market. In such a case, the history of those properties must be feathered into the statistics, requiring a revision of the prior aggregations. While this is cumbersome for users of the data to incorporate, it is necessary in ever growing markets. The best sources of data will account for the changes clearly and sometimes even produce parallel sets of statistics for a few editions to allow the users to adapt to the revisions. You need to check yourself to detect many of the changes. A few simple tests in your collected data will expose some of the most common failings. One of my favorites is this:

1. Create a time series of total inventory and construction completions for a market.
2. Calculate the change in inventory for each period and compare it to construction completions. It should match.
3. Differences in any period may indicate a serious shift and render the data highly suspicious.

Another common problem with time series data is the inclusion of increasing numbers of markets in national statistics. As the brokerage firms grew in size, they added the new markets to their statistics. This means that the definition of the nation changed over time. As a result, reported levels can not be used in variables such as construction completions, inventory, net absorption, and vacant space. It also means that ratios such as vacancy rates may be skewed as markets like Atlanta, with systemically high vacancy rates, are added. Worse are markets like Los Angeles which changed dramatically over the decades and have huge volumes of inventory. Unless you know which markets were added and when, it is impossible to repair the trend lines. Comparing erroneous time series to independent variables such as population, employment, or gross domestic product, leads to incorrect conclusions.

Finally, there are simple errors that you can make in compiling data from different sources. Pay special attention to the units of measure, and scrutinize the reports as they commonly include mistakes in this area. Other common issues include:

• California markets usually report rents on a monthly basis while the rest of the U.S. reports annually.
• The components included in the rent vary widely between gross, net, and hybrids of the two.
• Almost all reported rents are asking rents which tend to differ from actual negotiated rents and those differences shift through the cycles.
• Inclusion or exclusion of sublet space can affect rent and vacancy levels.
• The shifting of large properties between single and multi-tenancy can skew the data.

Now the internet age is fully upon us. We can look up information on every conceivable subject from anywhere. Search engines like Google use sophisticated algorithms to deliver what we want, even if we are poor at describing it. But search engines are designed to provide us content to read and enjoy. How do we turn those results into data fit for a database? Services such as Fetch, Wolfram Alpha and Seaglex do an excellent job there, but don't think your work is over at that point. You still need to spend time training these tools, and be very careful on how you use the results. You still need to understand what you are putting into your models. That said, the power of databases and the internet come together in these tools, and you won't get a paper cut flipping through old publications.