LeatherSagiKnowledgebase

Cap Rate: Please Explain!

3 min read

Problem

Practitioners use the term “cap rate” loosely, but there are at least six different definitions in the NCREIF query tool alone, each producing different numerical values for the same underlying market. This note explains the definitional differences and their implications for data users.

Key idea

A cap rate is the ratio of NOI to value, but the definition depends on:

  1. Which NOI: current quarter annualized (x4), trailing four quarters, or forward-looking pro-forma.
  2. Which value: appraisal value (if unsold) or transaction price (if sold).
  3. Weighting: equal-weighted (statistically generalizable) vs. value-weighted (portfolio-representative).

NCREIF provides six cap rate series: {equal-weighted, value-weighted} x {appraisal, transaction} x {single quarter x4, trailing 4 quarters}. Other providers (RCA, CBRE, caprate.net) use forward-looking or stabilized definitions.

Method

Descriptive comparison of the six NCREIF cap rate definitions with data visualizations showing their co-movement and differences. Key empirical observations:

  • Value-weighted cap rates are ~40 bp lower than equal-weighted (larger properties in major metros have lower cap rates).
  • Transaction cap rates are ~40 bp higher than appraisal cap rates.
  • Four-quarter trailing NOI has lower volatility than single-quarter annualized.
  • Filters applied: appraisal cap rates outside [-30%, 30%] excluded; transaction cap rates outside [2%, 14%] excluded.

Results

  • All six series show the same cyclical patterns but differ in level by up to 80 bp (value-weighted appraisal being lowest, equal-weighted transaction being highest).
  • The value-weighted appraisal series has the lowest standard deviation.
  • The gap between equal- and value-weighted series shows cyclicality depending on the CRE cycle position.
  • Cap rates are distinct from yields: yields account for capital expenditures, tenant improvements, and leasing commissions not included in NOI, making yields lower than cap rates.

Limitations

  • NCREIF cap rates use trailing (historical) NOI, while the theoretically correct definition is forward-looking expected NOI.
  • At the metro or property-type level, smaller samples introduce noise that may require 4-quarter moving averages.
  • Negative quarterly NOI (an accounting artifact) creates occasional negative cap rates, which are filtered out.

Open questions

  • How much does the trailing vs. forward-looking NOI difference matter for empirical asset pricing models that assume forward-looking cap rates (like the dynamic Gordon model)?
  • Are the systematic 40 bp differences between appraisal and transaction cap rates driven by appraisal smoothing, transaction selection bias, or both?

My take

Essential reference for anyone working with NCREIF cap rate data. For the CRE asset pricing project, the key takeaway is that the ~40 bp appraisal-transaction spread and the value-weighted vs. equal-weighted difference are systematic and should be accounted for when comparing model-implied cap rates to different data sources. The project uses NCREIF-derived data (capData), so understanding these definitional issues is important for interpreting the crisis_scale adjustment and the overall data scaling conventions.

Graph View

Backlinks