Colloquium: The Willingness to Pay for Mobile Telephony Services Among Low Income Households in Six Countries of Asia

Posted on April 2, 2010  /  3 Comments

This colloquium was presented by Sangamitra Ramachander, PhD student, Oxford university. This is an early draft, and the paper may change significantly.

This is based on findings from the Teleuse@BOP3 project.

The context is that in a variety of sectors, private sector do not have prior experience in serving low income and rural markets; however, now the entering. They have to decide on an appropriate price such that it is still profitable. How should one set the price? One can be through actually asking customers how much they would willing to pay. Useful when they don’t know much about what the product is about. However, respondents think that the ultimate price may be higher than what they state, hence they could be biased in their responses. However, such a method remains popular.

There has been a huge increase in people using such methods, even despite criticisms.

Four major types
Simple question: ‘how much would you pay?’
Single bound: one price point used – yes/ no response
Double bound method/ bidding game
Payment ladder – respondents choose from a list of prices

W.r.t. the Teleuse@BOP study,  respondents were given a payment ladder to mobile users, in the form of a diagram, which denote different prices, and asking how many minutes they would use at each level (not only mobile use, but any type of phone). This paper looked at only mobile owners, not non-owning mobile users.  A limitation is that every reduction in circle represents a 20% reduction, however respondents may not know that and they could perceive it in any way that they want.

Advantages: Respondents are existing mobile users; non-numerical form facilitates comparison across countries; data is of a continuous form since the commodity is divisible (in terms of number of mobile minutes used at different prices)

Limitation: ‘Meaning’ of each diagram (strictly 20%) could have been interpreted differently across respondents. Assumes respondents calculate new price (20% cut) and estimate new corresponding consumption independently. Arguably a strong assumption given profile of respondents – BOP population

The problem: If interpretations are different across respondents on magnitude of price change, responses not comparable.

How we construct this:

  • Consider the first two diagrams only (a ‘small change’ from the current price)
  • If respondents change consumption – they are elastic (coded 1), and if not, they are inelastic (coded 0)

Why is a binary outcome variable better?

  • Assumes only that respondents understood direction of price change (not magnitude)
  • Responses based on ‘small change’ in price is likely to increase precision

Clean the data: eliminating responses reflecting a lack of understanding

All those who didn’t understand the question were moved from the analysis.
A two-stage model was used to determine determinants of price elasticity. If we asked this only owners, demand would be more elastic, compared with those who were non-owners, however this argument can be contested.
Rohan – there is no point in asking this paper about those who don’t own phones because non-owner users generally found per minute costs affordable; the problem was the cost of getting connected, so there is a qualitative break between owners and non-owners; hence it would be good to treat both groups separately. For non-owners, we could ask them about the cost of ownership as well.
Sriganesh – are you taking only mobile owners?
Sangamitra – two parts of the model – the first looks at whether you are an adoptor not, the second looks at elasticity of use in response to change in price.
Heckman probit model has two stages:
  • The outcome equation: ‘Is demand elastic to price change?’; this outcome is observable only if respondents own a mobile phone (binary outcome variable)
  • The selection equation: ‘adoption’ equation (DeSilva et al.) – Modeling likelihood of mobile ownership (also a probit)
  • The model also allows more theoretical exploration by separating adoption and use variables
    • Same variable might affect the two equations differently (example income)
    • There might be variables relevant to one and not the other
The determinants of mobile adoption and use can be different, there could be others that are same, although in varying extents.
Mobile adoption is a function of:
  • Demographic characteristics (age, gender, education, income)
  • Perceived benefits of mobile adoption (emergency, social and economic)
  • Top five other individuals most frequently contacted who own a mobile (‘social influence’)
  • Ownership of other electronic goods in the household (tv and radio)
  • Access to electricity
  • Remoteness of location
  • Country dummies
Price elasticity of demand is a function of:
  • Cost minimizing techniques (use of mobile for missed calls or incoming calls)
  • Loyalty to service provider – those who say they are unlikely to switch to other service providers
  • Ownership of multiple SIM cards
  • Level of existing use (last mobile top up amount)
  • Purposes of existing use (services in the areas of health, govt, agri/fishing, general info, competitions)
  • Top five other individuals most frequently contacted who own a mobile
  • Perceived benefits of mobile adoption (emergency, social and economic) (based on deSilva mobile adoption paper)
  • Demographic characteristics (age, gender, education, income)
  • Access to electricity
  • Remoteness of location
  • Country dummies

Rohan – you refer to “savvy” users.  who are they? you refer to them as those who have more contacts (top 5) with mobile phones.  they are not necessarily those more sophisticated, they are just people who are more electronically connected. replace this word with someone more plausible.

HeckProb estimation of price elasticity of demand: significant variables – switching providers (those not likely to switch are more inelastic), number of sims, greater use in times of emergency (more elastic), top 5 contacts and competition.

Findings: Demand is elastic among more savvy users and those with more limited use:

    • Savvy users: Those who report a greater number of close contacts who own a mobile phone and those who use the mobile phone to participate in ‘competitions’ (the most popular among the list of services)
    • Limited users: Those who attribute a higher value to the emergency uses of a mobile
Finding: Loyalty to service provider appears to make demand less elastic:
  • Those who say they would not switch to other service providers are 22% more likely to have demand that is inelastic to price change, holding all other covariates at their mean values
  • A unit increase in the number of SIM cards owned increases the likelihood of demand that is price elastic by 11%, holding all other factors constant at their mean values

Concluding thoughts:

  • Latent demand exists among more savvy/active users and more limited (contingency) users since both have demand that is price elastic
  • However, users can easily switch between service providers
    • Competing on price can therefore lead to price wars, which could threaten survival particularly since profit margins are likely to be low in BOP markets
    • Loyalty to a service provider on the other hand appears to make demand relatively inelastic
    • Non price strategies therefore might be more effective in the long run for customer retention and survival

Rohan – the last slide suggests that we are interested in how companies can make more or less money; rather our focus is getting more at the BOP connected.

Sangamitra – i think if we are moving to private sector delivery of services of BOP useres, they have to survive, and hence how does one make sure that those who enter this market remain profitable. Then should this sector be regulated? These businesses should not resort to price cutting in such a way that drives companies out of business.

Christoph – i have conceptual issues.

Sangamitra – we don’t have panel data. With that limitation, we have use such models (contingent evaluation) to arrive at conclusions. w e are not looking at the magnitude of change, only whether it is elastic and inelastic. since we don’t know whether they understood the question

christoph – we can only say whether they would be willing to change or not, not elasticity.

Sangamitra – that’s usually an limitation of binary models

christoph – conclusions get falsified if one cannot talk about degree of elasticity. for those who don’t change use in response to price change, how come? it could either be because they didn’t answer the question, maybe were they not so poor.

christoph – it could be that the model is not relevant. also, if there was anything that looked nonsensical, i would check if there were outliers

sangamitra – i am less surprised by these findings; but it would be good to examine the profiles of those elastic versus those non-elastic.

sangamitra – what would be the regulator’s response to such findings? how would this relate to them?

nirmali – could setting some form a minimum price be a possible solution?

helani – yes, but situation not so simple.

ayesh – some form of bundling or price loyalty?

sangamitra –  yes.

ayesh – does it have implications on mobile number portability?

sangamitra – yes;

helani -it would be interesting to see if findings differ between pakistan where you can port numbers and those who don’t.

helani – however pakistan also has some of the highest multi-SIM ownership

ayesh – but qualitatives also showed that different in network coverage was one of the reasons why some of the multi SIMS.

helani – has anyone done analysis among those who don’t own mobile phones?

sangamitra – there is an estimate which shows the degree of correlation between unobservable factors in the model – we found that it was significantly negative; however, one must be cautious about making statements about such findings.

sangamitra – a lot of recent articles cite the Heckman model.


  1. while the slides referred to “savvy” mobile owners, in the actual presentation Sangamitra changed it to mobile-connected owners. I think it’s also important to clarify that the entire discussion is limited to mobile owner-users, not non-owner users. Therefore, quite a lot of the references in the slides and notes need to be replaced by owner-users, or owners.

  2. Sangamitra Ramachander

    Thanks to everybody who could attend for very useful feedback and discussions on the potential policy implications.