Cellular-Only Substitution in the U.S. as Lifestyle Adoption:

Implications for Survey Research

John Ehlen AE West/Somewhat Retired

(Formerly GTE/Verizon)jehlen@yahoo.com

Patrick EhlenCSLI, Stanford University

ehlen@stanford.edu

Cellular substitution is a stable process of lifestyle adoption.

Consequently,• The cell-phone-only population will

increase substantially.• We can predict the size of key

demographic groups in this cell-only population.

• Those predictions can be used to adjust wireline surveys when responses are not independent of sample frame, develop sampling plans, adjust subpopulation results and dual frame surveys.

These conclusions flow from a model that describes and predicts a vector of behavioral momentum

that is driven by• The strength of current incentives (Xit)

• The strength of the tendency to respond to these incentives (Bi)

• The strength of the tendency to persist in established behavior (or rate of habit retention)

This concept is consistent with both economic and psychological theories of behavior

e.g. Nevin and Grace (2000), Koyck (1954) & Houthaaker and Taylor (1970)

The nature of the vector of behavioral momentum over time depends on the strength of the perceived incentives to adopt

(Biln(Xit)).

0%

10%

20%

30%

40%

50%

60%

70%

80%Weak Incentive Shift

New Adoption - Strong Incentive

New Adoption - Weak Incentive

All vectors exhibit the same rate of habit retention.

Equilibrium-seeking vectors given stable perceived

incentives

Aggregate perceived incentive and rate of

habit retention determine target.

The vector of behavioral momentum over time also

depends on the strength of the rate of habit retention ().

0%

10%

20%

30%

40%

50%

60%

70%

80%

New Adoption - Higher Rate of Habit Retention

New Adoption - Lower Rate of Habit Retention

Both vectors approach an 80% target. The higher rate of habit retention Implies greater resistance to change.

Both wireless lifestyle adoption and wireless-only adoption

reflect the same basic behavior, characterized by the same rate

of habit retention (). • This hypothesis could be tested based on the

limited data available from the U.S. National Health Interview Survey (NHIS) – 8 semi-annual observations

• Three phases in test process• All parameters estimated from independent

samples (not NHIS) in Phase I and Phase II (prior information)

• Perceived incentives (ln(K)=biln(Xi)). by age cohort derived from NHIS data in Phase III

Phase I: Estimate for wireless adoption.

US Cellular Adoption 1987 - 2006Graphed forecast uses lagged estimates.

Parameters estimated from semiannual data through June 2006.

0

100

200

300

400

500

600

700

800

900

1000

1989

1992

1995

1998

2001

2004

2007

Subs per 1000 Population

Forecasted Subs (Cellulars)

Price (P), income (DI) and rate of habit retention () are significant at p<.01. Second half seasonal impact (II) and the index of consumer confidence (CI) are significant at p<.05

Transformed Estimating Equation (R2=0.981):

(lnYt – lnYt-1 ) = B0 + B1lnDIt + B2Pt + B3lnCIt + B4II

Where lnYt is the logarithm of cellular lines per 1000

persons in the current period.

Phase II uses the fact that an aggregate perceived incentive, K, can be derived from

data on the proportion of survey respondents (T) who would elect wireless only.

• We can think of the the aggregate perceived incentive as some constant that summarizes the implications of all the individual incentives: ln(K) = biln(Xi).

• When adoption reaches the equilibrium target, T, it must be the case that ln(T) = ln(K) + ln(T)

• So that the aggregate perceived incentive that describes the dynamic approach path to T given can be computed as

ln(K) = ln(T) – ln(T)

Phase II: Is the rate of habit retention the same in the adoption of both

wireless and wireless only lifestyles?The aggregate perceived incentive (K) was derived from

the 2006 dual frame survey conducted by Keeter & Kennedy. Predictions use as estimated in Phase I.

2%

4%

6%

8%

10%

12%

H2 - 0

3

H1 - 0

4

H2 - 0

4

H1 - 0

5

H2 - 0

5

H1 - 0

6

H2 - 0

6

Forecasted Cell Only

Actual Cell Only

R2 = 0.945, F = 18.3, p<0.01

Sensitivity Analyses:

75% of Somewhats p<.01

25% of Somewhats p<.05

Forecast Equation: Yt = K(Yt-1) andK

is constant.

Lagged estimates used after 2nd half 2003.

Phase III: We estimate models by age cohort. Tests indicate forecasts should be accurate and forecast results for I

2007 track fairly well…

Yt = K(Yt-1) and Yt-1 is actual percent for 2nd

half of 2006.

0%

5%

10%

15%

20%

25%

30%

35%

40%

18-24 25-29 30-44 45-64 65+ Total

Forecast Actual

Cohort-specific values of K derived from NHIS data on cell-only adoption by age cohort. Estimate for is from Phase I.

… and results for II 2007 also track fairly well.

Yt = K(Yt-1) and Yt-1 is forecasted percent for

2nd half of 2006.

0%

5%

10%

15%

20%

25%

30%

35%

40%

18-24 25-29 30-44 45-64 65+ Total

Forecast Actual

Forecast assume that the aggregate perceived incentive to adopt (K) is constant.

10.0%

12.0%

14.0%

16.0%

18.0%

20.0%

II 2006 I 2007 II 2007 I 2008 II 2008 I 2009

Old Data

Base Forecast

25% Incentive Shift

New Research

• Wireline-Only Model

• The Impact of Aging on Communications Lifestyle

jehlen@yahoo.com

NHIS changes that improved reporting of wireline status

decreased the data available for estimating models of “Wireline

Only.”

• Efforts to use the earlier data (2003-06) did not succeed.• Consequently, 2007 quarterly data were used to derive

perceived incentives by age cohort.

Report Period

Wireline Only

Wireline and Cell

Wireline with Cell Unknown

II 2006 27.3% 48.1% 10.5%

I 2007 20.8% 63.3% 1.7%

Model implies that two years from now only about 5% of adults under

30 and 13% of all adults will be “Wireline Only”

I2008

II2008

I2009

II2009

I2010

II2010

0%

5%

10%

15%

20%

25%

30%

35%

40%

18-24

25-29

30-44

Adults

45-64

65+

65+ drops to 30% by 2nd half of 2010

Getting Older: one might conclude from cross section data that the tendency to adopt the cell-only

lifestyle peaks at age 20-21.Among those under 30, the statistics for those not living with parents better reflect the choices of that age group. In the 2nd half of 2007 the adoption rate for the 49% of those 18-24 not living with parents was 53% compared to 10% for those 18-24 who were living with parents.

0.0%

15.0%

30.0%

45.0%

60.0%

Wireless Only Adoption 2003 - 2007 by Age Cohort(18-31 Not Living with Parents)

2003

2005

2007

However, the picture over time suggests a different story: the perceived

incentive to adopt is strongest among the youngest and is retained given

aging.

0.0%

15.0%

30.0%

45.0%

60.0%

Wireless-Only Adoption 2003 - 2007 by 2003 Age Cohort ( 18-31 Not living with Parents)

2003

2005

2007

A comparison of adoption levels at age cohort 18-19 for those who were under 18 in

2003 also suggests that enthusiasm increases among younger teenagers.

Vesa Kuusela reports that Finnish cell-only families with children grew from 2% to 40% from 1996 to 2005, a statistic which suggests that Finns retain the preference for cell only as they age. The short history available in the U.S. also implies that the lifestyle is retained as younger adults mature and as teenagers become adults.

Adoption at 18-19 by Age Cohort in 2003 (Not living with parents at age 18-19)

0%

10%

20%

30%

40%

50%

60%

Adoption at 18-19

18-19 in

2003

16-17 in

2003

14-15 in

2003

The aging effect explains a substantial portion of the increasing incentive to adopt the

wireless lifestyle.

• The inclusion of aging in the model explaining wireless adoption substantially reduces the impact of price and income changes.

• The rate of habit retention changes only slightly.

• The role of incentive shifts in wireless-only adoption is smaller than we thought, and we may well be able to predict a substantial part of these target shifts based on the aging of those now in younger age cohorts.

0.00

0.25

0.50

0.75

1.00

Comparison of Absolute Coefficient Values (using indices)

W/O AgingVariable

With Aging

Confirming and using the aging effect: implications and research

planImplications• Negligible impact on forecasts through 2009 • Substantial impact on what we would expect over

the next ten years• Restructured model to forecast adoption of

baseline age cohorts over time

Confirming the Aging Affect• Collaboration with Vesa Kuusela on the Finnish

experience• Develop data set on wireless adoption by age

cohort from the Consumer Expenditure Interview Survey starting from mid 1990’s

Thank you for the invitation and thanks to:

• Paul Lavrakas and three unknown reviewers for their helpful comments and suggestions

• Scott Keeter, Courtney Kennedy and the Pew Center for their 2006 dual frame survey, The Cell Phone Challenge to Survey Research, a study that provided us key information for model testing

• Stephen Blumberg and Julian Luke for their patience with all my questions as I worked to develop a detailed summary data set that reconciled to their published data

Questions and suggestions welcome: jehlen@yahoo.com