Page 148 - 35-2

The Effect of the Fair Value Reporting Model on Analyst Forecast Properties: Evidence from Real Estate
Firms

YEAR_0506, YEAR_0708, YEAR_0910, YEAR_1112, and YEAR_1314. For example,
YEAR_0506 equals one for the period 2005-2006 and zero otherwise. The other time-
period indicator variables are defined accordingly. More precisely, we estimate the
following regression:

DISPERSION = γ + γ UK + γ YEAR_0506 + γ YEAR_0708 + γ YEAR_0910 +
2
it
it
0
it
1
3
t
it
4
γ YEAR_1112 + γ YEAR_1314 + γ UK × YEAR_0506 + γ UK ×
8
it
5
it
7
it
it
it
6
YEAR_0708 + γ UK × YEAR_0910 + γ UK × YEAR_1112 +
10
it
it
it
9
it
it
γ UK × YEAR_1314 + γ SIZE + γ LEV + γ BM + γ OTHER_
14
it
13
it
it
12
it
it
11
15
A + γ OTHER_L + γ EPS_CHANGE + γ GROWTH + γ REIT it
17
19
it
it
18
16
it
it
+ γ DIVERSIFIED + γ STD_RET + γ PINT + γ FOLLOW +
it
it
23
22
20
21
it
it
γ HORIZON + YEAR + ε. (2)
t
it
24
Finally, we winsorize all continuous variables at the 1% and 99% levels to mitigate
the effect of outliers and cluster standard errors at the firm level.
4. Sample Selection and Descriptive Statistics
We construct the sample using six data sources: (1) real estate information from
the SNL database; (2) financial statement information for US firms from Compustat; (3)
financial statement information for UK firms from Worldscope; (4) analyst data from I/
B/E/S; (5) price information for US firms from CRSP; and (6) price information for UK
firms from DataStream. We consider publicly traded investment property firms from 2002
to 2014 that are domiciled in the US and UK for which we have both the data necessary to
7
derive control variables and at least three analyst forecasts. Table 1 presents the number
of firm-year observations across our testing samples for each year. We have 1161 firm-year
observations, of which 849 are from the US and 312 are from the UK.
Table 2 presents the descriptive statistics of the variables for US and UK firms. We
find that, relative to US firms, UK firms are smaller and have lower leverage, higher book-
7 To calculate the standard deviation of forecasts, we additionally require at least three analyst forecasts
to be consistent with the literature (Barron, Byard, Kile, and Riedl, 2002). Our results are robust to
the more stringent requirement of having at least 4 analyst forecasts.
140