effectiveness of advertising on lottery ticket sales...x5 $17.34 bill proposal: do good to do well...

March 8, 2013

Effectiveness of Advertising on Lottery Ticket Sales OR: Auditing Lottery’s Oddities

Peter Heineccius, JLARC Staff

March 8, 2013 PNIAF: Effectiveness of Lottery Advertising in Washington 2

Video Clip

Sales

$500 M


Media Buys $6.4 M

Production & PR

Vendor Fees

Advertising

Revenue $124 M

Revenue $136 M

Revenue $136 M 36

30 25 20 15 05 00

Revenue

Admin & Operating

Costs $63 M

Advertising

Prizes $301 M

Lotteries Are Oddities in State Government

selling a product instead of

providing a service

$12.1 Million Advertising Budget

(FY 2011)

x5

$17.34

Bill Proposal: Do Good to Do Well


“Buy a ticket, get rich!” Advertises Greed

“Buy a ticket, send a kid to college!” Advertises Altruism

$89.24

Per Capita Revenue

WA GA

WA Beneficiary: GA Beneficiary: College Scholarships

Washington: Georgia:

School Construction College Scholarships

“Buy a ticket, send a kid to college!” Advertises Altruism


Video Clip

Lottery Sales Higher in Eastern U.S.


$17.34

$89.24 $-

$20

$40

$60

$80

$100

$120

$140

Per

Cap

ita

Sale

s

Longitude

Washington’s Lottery Beneficiaries


Revenue $124 M

School Construction 80%

Other 20%

Sports Stadiums Economic Development

Problem Gambling

Scholarships/Early Learning 80%

45% of players knew education was a beneficiary

A third of those (15% of all players) claimed the beneficiary influenced their decision to buy a ticket


Video Clip

Beneficiary Awareness


6% 36%

6% of Players Distinguished New Beneficiary from Old

36% of Players Claim Identity of Beneficiary

Impacts Their Purchase 2% of Players Both

Knew and Cared About the New Beneficiary

$480

$490

$500

$510

$520

$530

$540

2009 2010 2011 2012

Mill

ion

s

JLARC Directed to Study Advertising Impact


Beneficiary Change

Lottery spent $2.25 Million on new beneficiary ad campaign

Sales increased by $19 Million the next year

JLARC concluded that advertising had an 8.4 return on investment

Sale

s (

Mill

ion

s)

Linear Regression


$-

$20

$40

$60

$80

$100

$120

$140

Average Impact of Longitude on Sales

Per

Cap

ita

Sale

s

Longitude

JLARC’s Approach


Collected weekly ticket sales data for each of Lottery’s games

Collected weekly expenditure data for media buys (television, radio, internet, etc.) Since advertising is valued based on expected

audience, using expenditure dollars worked as a proxy value for reach

Only available for two years (104 observations)

No apparent relationship between weekly ticket sales and weekly advertising dollars

JLARC’s Approach


Rounded up other data on likely factors that might influence ticket sales in Washington: Jackpot sizes Employment Consumer Price Index Stock Market (DOW) Gas Prices Weather Holidays Seasons

Iterations After Iterations


Each variable added to the analysis can affect the significance of all the other variables

JLARC tried a myriad of different combinations to discover which variables were consistently significant, regardless of what other variables were in the analysis (i.e., stepwise regression the hard way)

Eventually came up with predictive models based on consistently significant variables

Model 0

Dependent variable: Sales from all games combined

Time period: 2 years; 4.5 years; 1 year Adjusted R2: 0.872; 0.754; 0.945 Advertising broken out into several variables,

tried in several combinations; none were significant EXCEPT: Advertising not related to the

beneficiary change showed a significant negative impact in the 1 year analysis

Too few observations to make a conclusion March 8, 2013 PNIAF: Effectiveness of Lottery Advertising in Washington 15

Model 1

Dependent variables: Sales from each game separately (since each game is driven by different incentives to play)

Advertising did not appear to be a significant factor in weekly ticket sales (even if lagged)

Instead, model was highly predictive of sales just based on jackpot amount, economic indicators, and holidays


Model 1


$-

$2

$4

$6

$8

$10

$12

$14

$16

Wee

kly

Sal

es (

Mill

ion

s)

Actual Sales Predicted Sales

Adjusted R2: 0.912 for all games recombined

Paranoia Sets In

Model 1 was predictive…maybe too predictive… However, it would not be able to recognize the

impact of advertising if it was gradual or irregular over time – it assumed that advertising would spur more sales in the week(s) following exposure What if the cumulative effect of advertising

slowly altered people’s behavior? What if a variable that more accurately captured

advertising’s impact was omitted?

Another model was needed! March 8, 2013 PNIAF: Effectiveness of Lottery Advertising in Washington 18

Model 2

Dependent variables: Sales from each game separately

Time Period: Developed from the years prior to FY 2011 to create a baseline expected value for the year after the new beneficiary

Found actual sales in FY2011 to be $19.3 Million higher than expected But on further analysis, increase was limited to

Scratch tickets – other games were as expected Determined that the increase was due to a new

Scratch game launched in FY 2011 March 8, 2013 PNIAF: Effectiveness of Lottery Advertising in Washington 19

Model 2


$-

$2

$4

$6

$8

$10

$12

$14

$16

We

ekl

y Sa

les

(Mill

ion

s)

Actual Sales Predicted Sales

Beneficiary Change

Adjusted R2: 0.893 for all games recombined

Natural Experiment: Advertising Hiatus


Lottery stopped almost all media advertising during a three month period

During this hiatus, ticket sales were slightly lower than the models would have predicted However, the drop in sales occurs during the

first half of the hiatus Sales in the second half of the hiatus are slightly

higher than expected

So even if drop in sales caused by hiatus, sales returned to “normal” a month before advertising resumed

Findings


Couldn’t find any direct impact weekly advertising might have had on weekly sales

Couldn’t find any gradual impact beneficiary advertising might have had on sales

Stopping advertising for three months didn’t appear to have any impact on sales

Limitations and Caveats


Could not conclude why we found no impact: Poor quality advertising? Non-responsive public? Not enough advertising? Too much advertising?

Could not definitively rule out that advertising had some impact: Possible that impact is diffuse and cumulative Possible that impact is “hidden” in other factors Possible that impact needs more time before

becoming noticeable

Contact Information

Peter Heineccius, Project Lead [email protected] 360-786-5123

Stephanie Hoffman, Research Analyst [email protected] 360-786-5297

John Woolley, Project Supervisor [email protected] 360-786-5184

www.jlarc.leg.wa.gov

http://www.jlarc.leg.wa.gov/

effectiveness of advertising on lottery ticket sales...x5 $17.34 bill proposal: do good to do well...

Documents