NYU Stern Marketing Sitting through commercials: How commercial break timing and duration affect viewership Bryan Bollinger and Wenbo Wang January 01, 2012
Motivation Television advertising in Q4 increased from $4.38 in 2011 to $4.90 billion in 2012. Despite the continued growth, increased ease of recording or streaming content make actual viewership of ads a concern. Effect of Commercial Breaks 5/30/13 2
Motivation Effect of Commercial Breaks 5/30/13 3
Motivation Effect of Commercial Breaks 5/30/13 4
Motivation Effect of Commercial Breaks 5/30/13 5
In this paper A natural experiment in China commercial bans on within-serial commercials beginning January 1, 2012. Minute-by-minute ratings data Dynamic model of channel viewing Advertising slot costs Effect of Commercial Breaks 5/30/13 6
The effect of the ban Effect of Commercial Breaks 5/30/13 7
Example of viewership change Effect of Commercial Breaks 5/30/13 8
Serial viewership Effect of Commercial Breaks 5/30/13 9
Aggregate effect on ratings Effect of Commercial Breaks 5/30/13 10
Model y cpt = α cpt y cpt 1 + X cpt β +ξ t +ε cpt! The alpha coefficient on the lag rating captures the rate at which viewers switch away from the channel. X capture explanatory variables such as indicators for the start of a show, show and channel dummies, time dummies, etc. Effect of Commercial Breaks 5/30/13 11
Results from 15 serials marginal viewers Coefficients S.E t Sig. α 0.969.004 220.392.000 α drama.005.030.181.856 α romance -.070.028-2.545.011 α horror -.027.028 -.962.336 α mystery -.013.031 -.437.662 α crime -.023.005-4.199.000 α ancient costume.042.014 3.071.002 α comedy.036.031 1.164.245 α fantasy.009.022.405.686 γ within -.142.002-65.070.000 weekend -.007.003-2.543.011! Effect of Commercial Breaks 5/30/13 12
Results from 15 serials all viewers Coefficients S.E t Sig. α 0 1.003.002 654.126.000 α drama.040.016 2.504.012 α romance -.089.016-5.680.000 α horror -.124.018-6.862.000 α mystery -.039.016-2.396.017 α crime -.017.002-8.908.000 α ancient costume.053.006 9.126.000 α comedy.014.016.856.392 α fantasy.027.009 3.196.001 γ within -.140.002-62.912.000 weekend -.002.003 -.858.391! Effect of Commercial Breaks 5/30/13 13
Channel-level model The rate of switching away from the channel is a function of: Show genre Whether the show is in a commercial break Whether the channel is in a commercial break between shows Whether the channel just switched to commercials Show type(s) Whether we are in the post-ban period Effect of Commercial Breaks 5/30/13 14
Channel-level model results: Before the ban Base switching rate is 0.984 (varies by genre). Switching between shows is significantly lower (6.8 percentage points) if the next show is a serial. Switching during within-show commercials is larger for the nonserials. Channel switching is larger by 16.1 percentage points for the first minute of the within-show commercial break, and 7.5 percentage points after the first minute. The rate of switching is higher for the first minute of the betweenshow commercials, 25.4 points in the first minute with a base rate of 12.3 percentage points. Effect of Commercial Breaks 5/30/13 15
Channel-level model results: The effect of the ban Switching increases during shows of either type. Switching during within-show commercial breaks declines for the non-serials. Throughout the between-show break, viewers are less likely to switch channels (o.9 points). Switching is higher if the last show was a serial (1.3 points) Switching is lower if the next show is a serial (1.3 points) Effect of Commercial Breaks 5/30/13 16
Do people even begin watching? Ban Effect of Commercial Breaks 5/30/13 17
Full model ' y cpt = α cpt y cpt 1 + ( y ct ' y ct 1 )Δ ct +X cpt β +ξ t +ε cpt Δ ct =δ c +W cpt λ Δ describes channel switching to the focal channel δ includes dummy variables for the other channels W includes: Whether the other channels are in commercial Whether the other channels just went to commercial Whether the other channels are showing the same genre show Effect of Commercial Breaks 5/30/13 18
Full Model Findings Viewers are more likely to switch from other channels when: They are in commercial They have nearby channel numbers They are showing a show of the same genre On average, due to the ban: Switching during non-serial, within-show commercials declines Switching during commercial breaks between non-serials declines Switching during commercial breaks between serials have no significant effect Effect of Commercial Breaks 5/30/13 19
What are the implications? We expect the marketing clearing prices of ad spots will change. Effect of Commercial Breaks 5/30/13 20
What are the implications? Percentage change in price of within-non TV series ad break! Mean.2114 Median.1202 Std. Deviation.2926 Minimum -.11 Maximum.81 Percentiles 25.0270 50.1202 75.3427 Effect of Commercial Breaks 5/30/13 21
Price of advertising slots Percentage change in price of the four types of between-show ad break! after non TV series, before non TV series after non TV series, before TV series after TV series, before non TV series after TV series, before TV series Mean.2348.1908.3149.2890 Median.1176.2244.2727.2213 Std. Deviation.2206.2237.1349.3910 Minimum.01 -.19.22 -.33 Maximum.80.56.61.99 Percentiles 25.0526.0628.2239.0110 50 75.1176.2244.2727.2213.3810.3333.3000.6232 Presentation Title Goes Here 5/30/13 22
Price of Advertising Slots Percentage change in price of the four types of between-show ad break! after non TV series, before non TV series after non TV series, before TV series after TV series, before non TV series after TV series, before TV series Mean.2348.1908.3149.2890 Median.1176.2244.2727.2213 Std. Deviation.2206.2237.1349.3910 Minimum.01 -.19.22 -.33 Maximum.80.56.61.99 Percentiles 25.0526.0628.2239.0110 50 75.1176.2244.2727.2213.3810.3333.3000.6232 Presentation Title Goes Here 5/30/13 23
In conclusion Viewers have more ways to avoid commercials than ever. We use a natural experiment and a dynamic model of viewership to assess the effect of commercials on TV ratings. Higher switching rates during commercials mean more viewers for serials. Relative prices for ad spots change as a result of the commercial ban. Effect of Commercial Breaks 5/30/13 24
Thank you Effect of Commercial Breaks 5/30/13 25
Channel- level model α cpt = α 0 cp +α g +α post serials cp Post t S serial +α post nonserials cp Post t S cpt nonserials +......+ (γ within serials p +γ within serials FM within serials p FM cpt +γ post within serials FM ct Post t +γ post p FM cpt Post t )Cw serials +... cpt cpt... + (γ within nonserials p +γ within nonserials FM within nonserials p FM cpt +γ post within nonserials FM ct Post t +γ post p FM cpt Post t )Cw nonserials cpt +... where!...+ (γ btw btw c +γ post ct Post t )Cb ct γ btw ct = γ btw-base c +γ btw-fm c FM cpt + (γ btw-last c +γ btw-fm-last c FM cpt )Last ct + (γ btw-next c +γ btw-fm-next c FM cpt )Next ct! and γ btw-post ct = γ btw-base-post btw-fm ct +γ -post ct FM cpt + (γ btw-last-post btw-fm c +γ -last-post ct FM cpt )Last ct + (γ btw-post-next btw-fm c +γ -next-post ct FM cpt )Nex 5/30/13 Effect of Commercial Breaks 26