mobile phones and income inequality

52
The Effect of Cell Phone Penetration on Income Inequality in the Developing World

Upload: cmreilly

Post on 14-Oct-2014

1.078 views

Category:

Documents


2 download

DESCRIPTION

Undergrad thesis examining the effect of cell phone penetration into developing countries. Spring Semester, Georgetown University 2007.

TRANSCRIPT

Page 1: Mobile Phones and Income Inequality

The Effect of Cell Phone Penetration on

Income Inequality in the Developing World

Senior Thesis for B.A. Degree in Political Economy

By

Christopher M. [email protected]

April 26, 2007

Page 2: Mobile Phones and Income Inequality

PECO 401Profs. Olofsgard and Shambaugh

Presented to WGR 208Introduction to the Topic

Cellular phones are an integral part of the daily lives of Americans, providing us

with the ability to communicate and access information regardless of location or time.

According to statistics from the American wireless association CTIA, Americans as a

whole spend roughly 156 billion minutes on their cell phones and send over 3 billion text

messages per month.1 While cell phones spread quickly in America and most of the

developed world during the 1990s, in the developing world they remained, however, a

luxury good. This has changed. Over the better part of the past decade cell phone

penetration has been consistently increasing in the developing world. This is due in great

part to falling hardware and service prices and the spread of service availability.2

The spread of cellular technology into the developing world has raised debate

about its value as a tool for development. One aspect of this debate which has not

received sufficient study is the effect of cell phones on the distribution of income in the

developing world. The study of cell phones as a tool for general development is

valuable. However, the mere fact a country is developing and growing wealthier does not

necessarily mean that the welfare of its entire population is improving. In many

countries, growth exacerbates the already unequal distribution of wealth, making the

richest echelons tremendously wealthier while leaving the poor still burdened.3 I believe

it is necessary to peel back the skin of development to study whether the penetration of

1 “Wireless Quick Facts: December 2006.” CTIA: The Wireless Association. accessed at http://www.ctia.org/media/industry_info/index.cfm/AID/10323 2 Goodman, David N. “Used phones drive Third World wireless boom.” MSNBC. October 29, 2006. accessed at http://www.msnbc.msn.com/id/15434609/page/3/ 3 Galor, Oded and Daniel Tsiddon. “Income Distribution and Growth: The Kuznets Hypothesis Revisited.” Economica. Vol. 63. No. 250 (1996): S103.

CMR 2

Page 3: Mobile Phones and Income Inequality

cell phones in developing countries is having a positive or negative effect on income

inequality. This is the goal of this study.

Motivating Assumptions

There are a few basic assumptions which further motivate this study, and I will

expand upon these assumptions in the subsequent section. I believe that cell phones are a

viable tool to increase equality in developing countries because they are accessible,

provide multiple forms of economic empowerment and, due to the former reasons, create

the opportunity to “leapfrog” up income strata and close the income gap.

As cell phone production and service consumption have increased, the prices of

both cell phone hardware and cellular service have dropped dramatically, increasing the

accessibility of cell phones in poorer countries. Cell phone makers like Motorola have

begun to develop cheap, rugged phones specifically for use in the developing world. The

Motofone is capable of remaining on standby for 400 hours; this is especially useful in

more remote areas where electricity is difficult to come by.4 To further alleviate the

problem of scarce electricity, Motorola has developed a bicycle which has a holster

where a rider can charge their cell phone using peddle-power.5 Furthermore, used cell

phone resellers have emerged in many developed countries, purchasing old cell phones,

refurbishing them and then reselling for low prices in the developing world. One

company, ReCellular, refurbishes over 35,000 phones a week and sells nearly 60 percent

of these to countries outside of the United States for, on average, roughly $30 per phone.6

4 Kirkpatrick, David. “Tech targets the Third World.” Fortune on CNNMoney.com. December 22, 2006. accessed at http://money.cnn.com/2006/12/20/technology/fastforward_thirdworld.fortune/index.htm 5 Murph, Darren. “Motorola to roll out cellphone-charging bicycle in ‘emerging markets.’” Engadget. January 9, 2007. accessed at http://www.engadget.com/2007/01/09/motorola-to-roll-out-cellphone-charging-bicycle-in-emerging-mar/ 6 Goodman. Supra Note 2

CMR 3

Page 4: Mobile Phones and Income Inequality

The accessibility of cell phone technology in the developing world is further enhanced by

the widespread availability and relatively low cost of cellular service. According to

Michael Blumberg of consulting firm D.F. Blumberg Associates, around 80 percent of

people in the world now have access to a cellular signal.7 This number will only rise as

cellular providers invest more to bring their service to emerging markets. Finally, as

more private cellular service providers increase competition in developing countries, the

price of cellular service will decrease, making usage more affordable to those in lower

income brackets. Cell phone technology provides powerful services – voice

communication, text messaging and internet access – at a cost which is not prohibitive.

Cell phones provide numerous forms of economic empowerment to those in

developing countries.

Access to vast amounts of information, enabled by cell phones, creates

efficiencies in micro-businesses which allow citizens of developing countries to increase

profits and subsequently income. Cell phones allow individuals to utilize price

discovery, thereby maximizing profits. For example, fisherman in India call markets

while still at sea in order to acquire the best possible price for their catch.8 They are able

to maximize their profit while also minimizing the risk that the market be flooded with

too many fish upon their return from sea, reducing demand. Farmers in Kenya utilize

their cell phones in a similar matter, checking up to date crop prices on a service provided

by Safaricom, a local cellular service provider.9 Author Nicholas Sullivan estimates that

farmers and fisherman make an extra 10 to 20 percent in profits from their goods thanks

7 Ibid. 8 “In Vietnam -- and across developing world -- cell phones play vital role in fueling growth.” Technology Review. January 27, 2007. accessed athttp://www.technologyreview.com/Wire/18117/page2/ 9 Ibid.

CMR 4

Page 5: Mobile Phones and Income Inequality

to cell phones.10 In South Africa, 62 percent of small businesses claimed higher earnings

thanks to the utilization of cell phones.11 Cell phones also reduce the travel costs for

business owners.12 It is far cheaper and less time consuming to call a parts dealer in

Brazzaville than to physically travel to the city. Finally, the information available via cell

phones can help businesses mitigate risk. For example, fishermen can check weather

conditions on their phone before setting out to sea, ensuring that conditions will be safe

and amenable to fishing.

In addition, cell phones provide opportunities to create new businesses based

around cell phone services. The concept of “village phones” has spread throughout

Bangladesh thanks to a partnership between Norwegian telecom company Telenor and

famous Bengali microfinance bank GrameenBank, known collectively as

GrameenPhone.13 GrameenBank makes microloans to women in small villages in

Bangladesh with the purpose of purchasing a cell phone and service plan from

GrameenPhone. These women then sell minutes of cell usage to fellow villagers in order

to pay back the microloan and eventually make a profit.14 The concept of a village phone

allows poor villagers to utilize cell phone technology without having to make the

investment in a phone or service plan. For example, if a villager must call his doctor in

Dar es Salaam to inquire about prescriptions once a month, but cannot afford a cell phone

himself, he can simply pay the village phone operator for usage once a month. Thus he is

able to acquire his needed services without absorbing the cost of a monthly cellular

10 Sullivan, Nicholas P. You Can Hear Me Now: How Microloans and Cell Phones are Connecting the World’s Poor to the Global Economy. San Francisco: Jossey-Bass, 2007. 152.11 Sullivan. 6.12 Hadingham, Wenona et al. “Mobile Communications in South Africa, Tanzania and Egypt: Resultsfrom Community and Business Surveys.” Africa: The Impact of Mobile Phones. Vodafone Policy Paper Series. March 2005. 50.13 Sullivan. xviii14 Sullivan. 40-42.

CMR 5

Page 6: Mobile Phones and Income Inequality

service plan. The village phone operators in Bangladesh – known as phone ladies – are

able to make from $750 to $1,200 a year, doubling the average per capita income in the

country.15 Cell phones allow village entrepreneurs to earn a significant profit while also

providing a social good to their hometowns.

Cell phones also make job hunting and advertising more efficient and expansive.

First, job seekers are able to save time and travel costs by using their phone to contact

potential employers.16 In addition, by lessening the need to travel, job seekers can search

for jobs in multiple cities and regions in a short amount of time. This allows seekers to

get the best job possible in the shortest amount of time with the lowest cost. Laborers

and craftsman can also utilize advertising and their cell phones to access a larger

customer base and consequently earn higher profits.17 For example, a carpenter can

advertise his services and cell phone number on a main rural thoroughfare. Interested

customers can then contact the carpenter and organize the business transaction without

having to spend the time and money to travel and physically meet. Cell phones increase

both the prospective employer base for job seekers and the customer base for laborers.

Furthermore, cell phones are able to be used for money transactions. A study in

The Economist showed that the poorer a country is, the fewer people that will own bank

accounts.18 However, companies like Smart Communications, Globe Telecom (both

Filipino companies) and Celtel (South Africa) have created systems by which can use

their cell phones as bank accounts.19 Users can load money onto their phones in the form

of minutes and use these minutes to pay for goods and services. For example, if I went to

15 Sullivan. 152.16 Hadingham. 50.17 “In Vietnam-“ Supra Note 9.18 Sullivan. 126.19 Sullivan. 128-131.

CMR 6

Page 7: Mobile Phones and Income Inequality

a market and wanted to purchase an apple from a grocer I would send a text message to

the grocer’s cell phone which would credit to his account the number of minutes which

equal the price of the apple. There are many advantages to cell phone banking. First,

maintaining funds in digital form is safer than carrying cash, which can be lost or stolen.

In addition, in owning a cell phone bank account some users would be able to earn

interest or perhaps even develop a line of credit.20 Microloans can even be deposited

directly into these cell phone accounts.21 Finally, cell phones can be used as cheap and

mobile credit card terminals.22 Thus, small business owners can receive payment from

customers in the former of credit card transactions, creating flexibility and the potential

for greater profit.

One final advantage of cell phones is that they can make the transfer of

remittances from expatriate family members and friends easier and cheaper. Remittances

are a huge source of income for residents of developing countries, totaling around $300

billion per year.23 However, organizing the transfer of funds can be difficult and transfer

fees can be high. This pitfall is being overcome with systems such as UAE Exchange’s

UAE Exchange Wallet. Using this system family members or friends abroad can wire

money directly into the cell phone bank accounts of those requiring the remittances.24

Because these cell phone transfers are cheaper than traditional wire transfers, this money

saved can be forwarded to the recipient of the remittances. As reliance upon remittances

as a supplemental income is high in developing countries, this efficient and cheap

delivery system is of great value to those in the lower classes.

20 Sullivan. 133.21 Sullivan. 139.22 Sullivan. 134.23 Sullivan. 135.24 Sullivan. 137-138.

CMR 7

Page 8: Mobile Phones and Income Inequality

The Hypothesis

It is these qualities of cell phone technology which brought me to my hypothesis.

I believe that cell phones create the opportunity for those at the lower classes -- who are

just now gaining access to cell phone technology – to greatly increase their incomes in

relation to the upper classes, thereby creating a more equal distribution of income. There

are three main competing hypotheses to this puzzle. First, the flow of cell phones may

still be going primarily to the wealthiest strata of society, thereby further empowering

them economically and exacerbating income inequality. A second possible competing

hypothesis is that, though cell phones are making it into the hands of the poor, they are

unable -- because of lack of education, illiteracy or a general apprehension to handling

new technology – to effectively utilize them for economic gain. Thus we would see no

change in income inequality caused by cellular phones. A final competing hypothesis is

that the positive economic effects of cell phones are merely an exaggeration and not

applicable to developing populations on the whole. It could be that the success stories I

have described in some countries are merely extraordinary. If this were the case we

would see no positive relationship between cell phone penetration and the reduction of

inequality.

In order to prove my hypothesis, I ran regressions which sought to measure the

effect on income inequality of cell phone penetration alongside other communication

technologies and traditional factors which affect income inequality. I included other

communication technologies so that the results could help steer debate about what

technologies aid in reducing inequality, and consequently, which technologies should

CMR 8

Page 9: Mobile Phones and Income Inequality

receive attention from non-government organizations and philanthropists alike. Before I

go into a detailed description of each independent and dependent variable, I will discuss

my data population and time frame.

The Data and the Variables

As can be determined from the examples used earlier in this study, success stories

involving cell phones span country and continent. With this in mind I decided to use a

population for this study which represented all developing countries. Specifically, I first

selected countries which were labeled as ‘developing’ by the World Bank. Then I

filtered these countries, keeping only those which received a low to medium Human

Development Index ranking by the United Nations. Thus I was left with a sample which

contained only those countries which were poor and unequal. This final dataset consisted

of 120 different countries. See Figure A, found in the Appendix, for a complete list of

countries used in this study. The time frame I selected for my data is 1995 until 2004. I

chose 1995 because when analyzing my data sets it was around 1995 that cell phone data

became available for my target countries. I selected 2004 as my end year as this was the

final year for which I could attain the data I needed. All of my data came from the World

Bank’s World Development Indicators database.

The dependent variable in this study is income inequality. However, reliable

measures for income inequality are extremely difficult to attain. One popular measure is

the GINI coefficient, which describes how unequal a country’s income distribution is on

a scale of 0 to 1 (0 being perfectly equal, 1 perfectly unequal).25 A second way to

25 “Gini Coefficient.” Wikipedia: The Free Encyclopedia. Accessed at http://en.wikipedia.org/wiki/Gini_coefficient

CMR 9

Page 10: Mobile Phones and Income Inequality

measure income inequality is to determine the ratio of how much of the national wealth

lies in the hands of the upper strata in a country (say, the wealthiest 20%) to the share of

the national wealth in the lowest strata (the poorest 20%). However, as I searched for

data for this study I discovered that data for these variables were computed infrequently

and erratically. Many countries had their GINI coefficient or income share computed

only once or twice over the ten year span of my time frame. To further exacerbate

problems with these data, those instances where the GINI or income share were

computed rarely occurred during the same years. While these variables were ideal for

measuring income inequality, the lack of data rendered them useless for running the

regressions necessary to this study.

In order to measure income inequality, I decided it was necessary to use proxy

variables. There are many competing ideas in the academic world for which variables

accurately represent changes in income inequality. They span the spectrum of types of

variables, from health related to labor to education. While attempting to acquire data for

several of these potential proxy variables I ran into many of these same issues I

experienced when searching for my original measures: lack of data, erratic computation.

However, I finally obtained data for two variables which were sufficient to run

regressions. These two proxy variables are the number of cars per thousand people

(carsperthou) and the number of women in the work force as a fraction of total women in

the country (femaleworkersperthoufemales). The number of cars per thousand people is

recommended by Ira S. Saltz in his article, “Income Distribution in the Third World: Its

Estimation via Proxy Data.” Saltz suggests that ownership of cars increases as the

CMR 10

Page 11: Mobile Phones and Income Inequality

income of the country is more widely distributed.26 This makes intuitive sense, as a

single household often has a maximum number of cars that it would purchase. Thus if

the total number of cars in the country is increasing, it is due to the fact that more

households are breaching the income threshold where they can afford a car. Thus, an

increase in the number of cars in a country represents a more equitable distribution of

income. The second proxy variable I used is used is the number of women in the work

force as a fraction of the total number of women in the country.27 As income is more

widely distributed, more households will be able to wean themselves off of subsistence

living – now having enough income to purchase food and household goods instead of

needing to produce these items themselves. The decreased amount of labor required in

the household affords the opportunity to the women of the household to enter the

workforce and earn an income of their own. I made this variable a fraction of total

women in the country to control for general growth in the female population. The

observation of increased women in the work force indicates a decrease in the level of

income inequality in the country. In order to provide quantitative proof of the

relationship between these two proxy variables and income inequality I ran correlations

between these variables and the limited number of GINI observations I acquired. The

correlation between cars per capita and the GINI is -.1767 and the correlation between

female laborers as a fraction of the female population and the GINI is -.1727. While

neither of these correlations are especially strong, they do represent a modest negative

correlation between both of my proxy variables and the GINI. This, as hypothesized,

26 Saltz, Ira S. “Income Distribution in the Third World: Its Estimation via Proxy Data.” American Journal of Economics and Sociology. Vol. 54. No. 1 (1995): 17-18.27 The suggestion of this variable was made by Prof. Shambaugh in a conversation during the earlier work of this study.

CMR 11

Page 12: Mobile Phones and Income Inequality

means that as the number of cars per capita and women in the labor force as a fraction of

the total female population increase the amount of inequality in the country decreases.

I used eleven explanatory variables in my regressions. These variables are

grouped into four categories: mobile phones, other communication technologies,

traditional explanatory variables, and regional identifiers.

My first variable, mobilephonesperthou, is the number of mobile phone

subscribers per thousand people. This explanatory variable is the focus of most of my

attention in this study.

My next two variables -- telephonelinesperthou and internetusersperthou -- are

representative of other communications technologies. Telephonelinesperthou is the

number of telephone land lines per thousand people. Internetusersperthou is the number

of internet users per thousand people. I would have also like to have data for the number

of personal computers per thousand people. Unfortunately, there was not enough data for

this variable to make running regressions possible. I am interested in measuring the

effect of these variables on inequality as I would like to see the effect on inequality of

communication technologies in general. Additionally, any variance in significance

between these technologies and cell phones will help guide philanthropic and NGO

development efforts.

My next three variables -- gpdpercap, urbanpoppct and roadsperkmsq – represent

other traditional factors which affect income inequality. Gpdpercap is the Gross

Domestic Product per capita in constant year 2000 U.S. dollars. This variable represents

the general economic growth of a country over time. This variable should have a highly

significant affect on income inequality. Urbanpoppct is the percentage of the population

CMR 12

Page 13: Mobile Phones and Income Inequality

which lives in an urban environment. Urbanization should exacerbate inequality over

time, as cities are areas where there exist many low wage jobs and few high paying

executive and managerial jobs. Roadsperkmsq is the number of kilometers of roads per

square kilometer of land in a county. I made this variable a fraction of square kilometers

of land so to control for the varying sizes and levels of road development of countries in

this study. It is intuitive that the more expansive the road network in a country, the easier

it is to travel. With easier travel, laborers have greater access to income opportunities. I

am interested to observe the significance of these more widely supported explanatory

variables in relation to the cell phone and other communication technologies variables.

My final five variables – africad, middleeastd, asiaoceania, europed and

latinamercaribd – are regional dummy variables. Each variable takes either a 1 or 0

based upon whether a specific country is located in these general regions or not. Africad

represents countries in Africa, middleeastd those in the Middle East, asiaoceania those in

Asia or the Oceania region, europed those in Europe and latinamercaribd those in Latin

America and the Caribbean. These regional dummy variables help to alleviate the

omitted variable bias. There are regional cultural and religious variables, among others,

which either cannot be accurately measured or otherwise could not be included in this

study which may have an effect on income inequality. Including these regional dummy

variables will help to minimize the bias created by the omission of these variables while

also providing interesting insight on which regions are especially unequal.

The Initial Results

CMR 13

Page 14: Mobile Phones and Income Inequality

Using this data I ran two ordinary least squares regressions – one using cars per

thousand people and one using women in the labor force as a fraction of the total female

population – using the aforementioned independent variables. The results of these

regressions were as follows:

***Please See Figure 1 on Next Page***

Dependent Variable femaleworkersper~s carsperthou

Number of Obs 551 311

cons 442.364 3.849551

(19.06)* (0.65)

gdppercap -.013547 .0131956

CMR 14

Page 15: Mobile Phones and Income Inequality

(-2.72)* (2.96)*

mobilephonesperthou -.0320784 -.0393965

(-0.34) (-1.17)

telephonelinesperthou .2192299 .2720967

(2.38)* (6.04)*

internetusersperthou .542912 -.0478008

(1.54) (-0.38)

urbanpoppct -.7036817 .0194672

(-2.48)* (0.10)

roadsperkmsq -12.60565 -8.940365

(-1.22) (-1.62)

middleeastd -205.0752 21.24525

(-7.30)* (1.13)

asiaoceaniad -39.24138 -3.472976

(-1.78) (-0.45)

europed 3.027514 36.62372

(0.12) (4.34)*

latinamercaribd -73.32578 -4.438535

(-3.07)* (-0.42)

africad -69.76991 -1.229439

(-3.13)* (-0.20)

Figure 1

The first result of interest is that cell phones have no significance on inequality in

either of these regressions. This does not mean that this project is for naught. I will

address how these regressions still support my hypothesis later in this section and in the

following “Sensitivity Analysis” section.

CMR 15

Page 16: Mobile Phones and Income Inequality

Two other unexpected results must also be mentioned. First, in the females

regression, GDP has a negative relationship with women in the work force and, by proxy,

with equality. According to this regression, as GDP increases, so does inequality.

However, according to the second car regression, as GDP increases, inequality decreases.

This result required further analysis and regression and is explained in the “Sensitivity

Analysis” section. The final strange result in this regression relates to roadsperkmsq. It

is completely counterintuitive that as the number of roads in a country increases,

inequality should decrease. More roads should make travel easier and should thus make

it easier to find better jobs and buy cheaper goods. Though this variable does not have

significance in the 95% confidence interval in this regression, the result still concerns me

and is resolved in the “Sensitivity Analysis” section.

One of the most significant variables in these regressions is the number of

telephone lines per thousand people. Telephone lines are positively and strongly related

to an increase in equality. Telephone land lines offer many of the same benefits that I

discussed in my earlier section on cell phones. Job seekers are able to search for more

jobs if they can call potential employers in many areas of the country. Business owners

can make sales and organize the purchase of goods via telephone. In addition, telephone

service aids in price discovery, decreasing costs and increasing profits. The increased

profits acquired via the use of telephone service aids in closing the income gap. The

importance of land line telephone service helps to support the importance of cell phones

for the reduction of inequality. Both land line phones and cell phones offer the same

service, voice communication. However, cell phones have two advantages over land

lines systems. First, cell phones are mobile and therefore one can utilize voice

CMR 16

Page 17: Mobile Phones and Income Inequality

communication anywhere, instead of being bound to a land line connection (of which

there may only be one in some remote villages). Secondly, and more importantly, cell

phone systems require much less infrastructure and maintenance and, consequently, cost

less to install and run.28 Instead of having to run miles of cable companies need only to

construct a few cellular signal towers to cover many square miles of territory. Less

hardware means less that can break or be damaged by the elements and therefore less to

fix. I believe that cell phones were not significant in these regressions because they have

not been around very long, and there numbers are still relatively low. Thus, not enough

time has passed for them to have a strong effect on inequality. I do believe, however, that

because they offer the same services as land lines and for a lower cost, cell phones are an

important and viable infrastructure that will, in time, reduce inequality.

The percent of the population which lives in an urban environment is significant

and negatively related to equality in the first regression. This makes intuitive sense as

urban centers tend to be areas where there exists a job dichotomy between a small

amount of well paid executives and managers and a multitude of low wage labor jobs.

Cities also tend to be centers of industry which require many low wage laborers. Thus

the result we received in the first regression makes good sense. However, in the second

regression, urbanization has no effect on the number of cars, and consequently,

inequality. This could be because if all other variables are held constant and urbanization

increases by one point -- meaning one percent more of the population moves to the city –

this new population may not necessarily be able to afford a car. Thus, urbanization has

an insignificant effect on cars.

28 “In Vietnam-“ Supra Note 9.

CMR 17

Page 18: Mobile Phones and Income Inequality

The number of internet users per thousand people had no significance at the 95%

confidence level in this regression. I believe this result is due to reasons similar to those

related to cell phones. Internet technology has not been around for very long and thus is

insignificant in the limited time frame of this regression. In addition, if there are few

computers, which are an expensive good, then access to the internet would be limited.

Though this variable is insignificant, in the first regression it does have significance at the

88% confidence interval. While not significant enough here, it is high for such a new

technology. In addition, the relationship between internet access and equality is positive.

This is an important connection on which I will expand later.

Finally, in the first regression, three out of the five regional dummy variables

were significant, one was relatively significant and one was not significant at all. For all

regions except Europe, omitted regional factors negatively affected the number of women

in the workforce and exacerbated inequality. These omitted factors could be religion – in

which restrictions on women would affect their ability to work outside the home – or

cultural factors which restrict individuals from leaving subsistence lifestyles. Europe

may have no significance in the first regression because countries in question (e.g.

Russia, Albania, Macedonia, Georgia et al.) are so socially and culturally different that

the fact that they are all in Europe has no impact on the amount of women in their

workforce and therefore on our proxy measurement of inequality.

In the second regression, Europe was ironically the only region to have

significance. This could be because Europe generally has more cars than any of the other

regions and thus being in Europe increases the possibility of an individual having a car.

In regard to the other regions, countries within these regions may be so differently

CMR 18

Page 19: Mobile Phones and Income Inequality

developed (within the category of developing) that merely being in the region has no

effect, one way or the other, on the number of cars per person. For example, Brazil has

far more many cars than Haiti, so they offset each other.29 Without a significant effect on

cars to observe, we cannot determine – using this regression – the regional effect on

inequality. We can only observe that being in Europe reduces inequality.

Sensitivity Analysis

For the first part of my sensitivity analysis, I would like to address three small

issues. First, when testing for heteroskedacity, I learned that my second regression (cars)

was heteroskedastic. I resolved this problem by regressing with robust standard errors. I

corrected for heteroskedacity prior to writing this paper, so the results above are accurate.

A second issue I had when performing this study was that I could not attain

enough data regarding the number of personal computers in my target countries. I

believe that in having this data I would have observed interesting results. Computers,

though an information technology, are not necessarily a communication technology and

therefore might not necessarily help alleviate inequality. In an early regression for this

project, using a somewhat different dataset and focusing on African countries I was able

to test for the effect of PCs on inequality. Interestingly, PCs had the effect of

exacerbating inequality. I believe this is due to the fact that not all PCs are connected to

the internet and that PCs have a relatively high learning curve, especially for the

illiterate.30 I believe that PCs do not provide the same developmental power as cell

29 Data taken from WDI dataset. 30 Maney, Kevin. “Gates sees cellphones as a way to help Third World.” USA Today Online. January 31, 2006. accessed at http://www.usatoday.com/money/industries/technology/maney/2006-01-31-gates_x.htm

CMR 19

Page 20: Mobile Phones and Income Inequality

phones, due to their lack of accessibility, mobility and normally high cost. However, due

to the lack PC data for my population, I cannot prove this point using my regressions.

This issue with PC data is representative of a great problem I had when producing

this study: the lack of reliable data for developing countries. Ideally, for the number of

countries and years in my study, I would have had around 1200 observations in my

regressions. However, due to a lack of data much was dropped from my regressions and

the most observations I have is 551, well under half of my ideal amount. I believe that

with more data the explanatory value of my entire regression would have been greater,

along with clearer relationships between data within the regression. As better data for

this subject becomes available, the regression above should be repeated and its results

compared with mine.

When I ran a correlation between all of my variables in the above regressions, I

noticed an especially high correlation between the number of cell phones and internet

users in countries. The two variables were positively correlated with a value of .6685.

This raised the concern that there were multicollinearity issues with my regression. In

order to test for multicollinearity I ran auxiliary regressions for internet users and cell

phone users. For these auxiliary regressions I set the dependent variable to be either

internet users or cell phone users and kept the independent variables the same, excluding

whichever variable I had selected to be the dependent variable. From these regressions I

observed that internet users and cell phone users had high explanatory value for each

other. The full results of these auxiliary regressions can be viewed in Figure B, found in

the Appendix. In order to further observe the effects of these two variables upon each

other, I ran regressions in which I had removed one or the other variable. In these

CMR 20

Page 21: Mobile Phones and Income Inequality

regression I observed that if I removed, for example, internet users from the regression,

the significance of mobile would increase, and vice versa. The results of these

regressions can be found in Figures C and D, located in the Appendix. From these results

I determined that the effect of cell phones on inequality could also be found in using the

internet. Cell phones provide mobile access to the internet. Thus, in having both

variables in my regression their effect was to produce a bias. Because of this bias, and

because this study is primarily focused on the effects of cell phones on income inequality,

I decided to remove the internet variable from my study (See Figure C). This result is

interesting because, as I mentioned before, cell phones are a cheap, mobile way to access

voice communication. The analysis above reveals that cell phones also provide the

positive effects of traditional internet access. Traditional internet access occurs through

land lines. As I discussed earlier, the infrastructure for a land line network is more

complex and costly than that of a cellular network. The ability of a cell phone to provide

both voice communications and internet in a cheap, mobile fashion renders it a standout

candidate as a tool for development and income inequality reduction.

After observing the results of my earlier regressions I was also concerned about

the relationship between roads and inequality. Intuitively, as more roads are constructed

in a country the effect should be to decrease inequality. This effect should occur because

as a network of roads spreads, travel becomes easier and this enables workers to search

for more jobs, business owners to attract more customers and acquire goods more

inexpensively and generally improves commerce. However, in my results roads are

observed to have a negative relationship with equality (this observation made despite the

fact that roads are not significant on a 95% confidence interval). In trying to figure out

CMR 21

Page 22: Mobile Phones and Income Inequality

why these results did not match my hypothesis I determined that perhaps the number of

roads in a country did not have a linear relationship with inequality. I determined that as

a road network is built, from its start up until a certain threshold it would actually

contribute to inequality. The location of the first roads to be built would likely be

determined by policymakers within the government of the developing country. These

elites would likely have roads built in areas that would be advantageous to their

government and to any business interests. Thus this young road network would serve the

few and therefore increase inequality. However, after a certain threshold there would be

enough roads to connect the masses, helping to reduce inequality. In order to capture this

characteristic of the road variable I inserted an additional term, roadssq, which is the

number of kilometers of roads per square kilometer of land, squared. When I re-ran the

regression, this time excluding internet users (for the aforementioned reasons) and

including the roadssq term, the results appeared as follows:

***See Figure 2 on Next Page***

Dependent Variable femaleworkersper~s carsperthou

Number of Obs 113 79

cons 434.8558 26.58343

(10.54)* (1.42)

CMR 22

Page 23: Mobile Phones and Income Inequality

gdppercap -.0143101 .0134745

(-1.23) (2.32)*

mobilephonesperthou .151779 -.1133557

(0.95) (-2.23)*

telephonelinesperthou .4978692 .3263002

(2.67)* (5.28)*

urbanpoppct -.8007361 -.6195638

(-1.14) (-1.23)

roadsperkmsq -154.8368 -7.617844

(-1.99)* (-0.31)

roadssq 79.61476 -9.335696

(1.71) (-0.58)

middleeastd -243.7206 55.641111

(-3.81)* (2.00)*

asiaoceaniad -21.63203 9.649287

(-0.57) (0.63)

europed -14.99768 47.92073

(-0.34) (2.98)*

latinamercaribd -70.52162 4.670904

(-1.57) (0.29)

africad -76.55603 2.76391

(-1.84) (0.24)

Figure 2

These results are especially interesting. Though at first glance the results of

regression one vis-à-vis roads may seem not to have changed significantly, one must

analyze the effect of an increase in one unit of road (one km) on women in the labor force

and, by proxy, inequality. First, we must pull the equation for the road variables effects

CMR 23

Page 24: Mobile Phones and Income Inequality

on the dependent variable out of the general equation. The equation is:

femaleworkersper~s= β1 roadsperkmsq + β2 (roadsperkmsq)2 .31 From our regression, we

know the values of β1 and β2 . They are -154.8368 and 79.61476, respectively. In order

to determine the change in femaleworksper~s, we much calculate the first partial

derivative of female workers with respect to roads. The result we achieve is

Δfemaleworkersper~s = β1 + 2 β2. Plugging in the coefficients we get

Δfemaleworkersper~s = -154.8368 + 2 (79.61476) = -154.8368 + 159.22952 = 4.39272.

Thus an increase in the number of roads leads to an increase in the number of female

workers and, by proxy, a decrease in inequality. Though roadssq does not have

significance in the 95% confidence interval, I still believe these results have high

explanatory value, backed by intuitiveness.

In the second regression, neither of these road variables have any significance.

This makes logical sense because, if one holds all else constant, an increase in the

number of roads will not necessarily have an impact on the number of cars. An increase

in income via GDP would enable more cars to be purchased, the existence of roads,

however, does not create the opportunity to purchase cars. By adding the roadssq term

into my regressions – and thereby adjusting for the nonlinear relationship between roads

and inequality -- my results are now far more sensible.

The final issue with my regressions is that the effect of GDP per capita on my

inequality proxies is negative in my first regression and positive in the second. In

analyzing my newest and most complete regression I have discovered a logical

explanation for this difference in vector. First, throughout this entire study, the two

variables that have been most significant in decreasing inequality are telephones per

31 To clarify: roadssq = (roadsperkmsq)2

CMR 24

Page 25: Mobile Phones and Income Inequality

thousand people and roads. These two represent key infrastructure – one in

communication and the other in transportation – that enable individuals to find better

jobs, become more efficient and earn more income, consequently narrowing the income

gap. It is the existence of this infrastructure that is key to effecting positive change on

inequality.

In order to understand the results above, we must look at two situations, one

where this key infrastructure exists and the other where it does not. When this

infrastructure exists, individuals are able to leave their villages to acquire new and better

jobs (regression one- roads and telecoms have positive effect). This increase in jobs and

productivity increases the GDP per capita in the country and with this new wealth more

people are able to afford cars, so cars per thousand people increases (regression two).

The new amount of cars is not only a sign of increased equality but the fact that people

have cars allows them to travel even more efficiently and saves more time and money,

increasing income. Thus, inequality decreases with the existence of communication and

transportation infrastructure, while cars increases as a result of wealth being more

equitably distributed.

On the other hand, if there is no key infrastructure, many new jobs will not be

discovered and/or created. Thus in order for an increase in GDP per capita to occur those

people already with jobs must become more productive. With this increased wealth

flowing only into this smaller group of workers, inequality is exacerbated. The number

of cars in the country goes up (regression two) but they are only being purchased by these

wealthy individuals and provide no advantage to the poorer classes. In addition, because

no infrastructure exists and consequently no job discovery is occurring, this increase in

CMR 25

Page 26: Mobile Phones and Income Inequality

GDP per capita has no effect on the number of women in the work force (regression one-

gdppercap is insignificant on a 95% confidence interval). These two new regressions

have a much higher explanatory value than previous regressions and logical explanations

anchor these results.

Finally, in the second regression above, cell phones are significant and negatively

related to the number of cars per thousand people. I believe this is because when there

are fewer cell phones in a country, traveling by car is a non instantaneous means of

communication. However, as cell phones spread the need for each individual to have a

car decreases, due to the fact that cell phones are cheaper to use for communication and

travel by car can be arranged without having to own one personally. I do not think the

negative coefficient of cell phones in this regression is evidence of cell phones causing

inequality. Instead, I believe this coefficient is the result of the relationship between cars

and cell phones specifically.

Policy Options and Conclusion

The results of my work suggest some clear policy directions. First, philanthropic

efforts should be directed towards the development of communications and transportation

infrastructure in these developing countries. Though I only looked at roads as a

transportation infrastructure in this study, I believe the general development of roads, rail

and air travel would help to increase growth and decrease inequality. Furthermore, this

transportation infrastructure must be coupled with an effective communication

infrastructure. If one has a means by which to travel outside of the village, but does not

have access to communications, then it will be difficult to fully take advantage of this

CMR 26

Page 27: Mobile Phones and Income Inequality

transportation infrastructure as a means to get to new jobs and increase profits for

existing businesses. Though cell phones were mostly insignificant throughout the

regression results, I do believe that they are a key communication technology for

development as they provide voice, text messaging and internet service all for a low cost,

with mobility and without the need for a complex network infrastructure. If these key

infrastructures are in place I believe these target countries will not only grow but their

new wealth will be more equitably distributed.

Bill Gates has shifted his attention – and considerable financial weight – away

from programs which support sending computers to developing countries32 and towards

programs which seek to increase cell phone penetration.33 As cell phones are an

accessible and viable tool not only for development but for the creation of equality more

philanthropist and NGOs should shift their sights to cell phones a-la Gates. In addition,

cell phone equipment makers should follow Motorola’s suit and begin to develop cheap,

rugged phones especially for the developing world. While profit margins from these

phones may not be high, the prospect of tapping the multi-billion-person market in the

Global South will undoubtedly be attractive to forward thinking CEOs. Finally, the

public should support the efforts of resellers like ReCellular, who take old phones off of

people’s hands while delivering them refurbished and cheaply into the hands of those in

the developing world.

Lastly, the governments of these developing countries should decrease barriers to

cell phone development such as high taxes or tariffs. In addition, incentives should be

offered to urge more cellular service providers to enter these countries, thereby increasing 32 Per the following source: Nicholas Negroponte has developed a $100 laptop specially made for developing countries. He has received backing from some big names, notably Google, AMD and the United Nations.33 Maney.--

CMR 27

Page 28: Mobile Phones and Income Inequality

competition and lowering prices. The cheaper and better quality both hardware and

cellular service becomes, the faster it will spread in the developing world and the more

significant effect it will have on development and inequality reduction.

More research must be done on this topic as more and better quality data becomes

available. In addition, more empirical work should be done on a broad level to determine

how different people in developing countries utilize their cell phones. A report which

utilizes more expansive proxy data or one that can gain access to income shares at

different strata would be valuable, as its results would surely be robust. Despite the

pitfalls which data gave me in this study, the results of this work have revealed the

importance of communications technology and specifically the viability of cell phone

technology as a tool for the reduction of inequality.

Appendix

Albania Egypt, Arab Rep. Liberia Serbia and Montenegro

Algeria El Salvador Libya Sierra Leone

AngolaEquatorial Guinea Macedonia, FYR Solomon Islands

Armenia Eritrea Madagascar South Africa

Azerbaijan Ethiopia Malawi Sri Lanka

Bangladesh Fiji Maldives St. Kitts and Nevis

Belarus Gabon Mali St. Lucia

CMR 28

Page 29: Mobile Phones and Income Inequality

Belize Gambia, The Marshall Islands St. Vincent and the Grenadines

Benin Georgia Mauritania Sudan

Bhutan Ghana Micronesia, Fed. Sts. Suriname

Bolivia Grenada Moldova Swaziland

Botswana Guatemala Mongolia Syrian Arab Republic

Brazil Guinea Morocco Tajikistan

Burkina Faso Guinea-Bissau Mozambique Tanzania

Burundi Guyana Namibia Thailand

Cambodia Haiti Nepal Timor-Leste

Cameroon Honduras Nicaragua Togo

Cape Verde India Niger Tunisia

Central African Republic Indonesia Nigeria Turkey

Chad Iran, Islamic Rep. Pakistan Turkmenistan

China Iraq Palau Uganda

Colombia Jamaica Papua New Guinea Ukraine

Comoros Jordan Paraguay Uzbekistan

Congo, Dem. Rep. Kazakhstan Peru Vanuatu

Congo, Rep. Kenya Philippines Venezuela, RB

Cote d'Ivoire Kiribati Russian Federation Vietnam

Djibouti Kyrgyz Republic Rwanda West Bank and Gaza

Dominica Lao PDR Samoa Yemen, Rep.

Dominican Republic Lebanon Sao Tome and Principe Zambia

Ecuador Lesotho Senegal ZimbabweFigure A – Country List

Dependent Variable mobilephoneperthou internetusersperthou

Number of Obs 569 569

cons -15.93055 1.526856

(-2.83)* (0.63)

gdppercap .0027973 -.0008309

(0.96) (-1.38)

mobilephonesperthou --- .1702284

--- (8.40)*

CMR 29

Page 30: Mobile Phones and Income Inequality

telephonelinesperthou -.0776676 .0741605

(-1.19) (4.88)*

internetusersperthou 2.48966 ---

(6.23)* ---

urbanpoppct .5121992 -.0375679

(4.44)* (-1.30)

roadsperkmsq 4.740015 -2.209289

(0.81) (-1.98)*

middleeastd -8.502736 3.647001

(-0.98) (1.02)

asiaoceaniad 7.806412 -1.039871

(1.55) (-0.40)

europed -5.127645 -3.803266

(-0.37) (-0.94)

latinamercaribd -11.33904 3.030834

(-1.53) (1.04)

africad -1.03709 .4818684

(-0.18) (0.19)

Figure B – Auxiliary regressions

Dependent Variable femaleworkersper~s carsperthou

Number of Obs 618 359

cons 422.2903 3.35993

(19.34)* (0.70)

gdppercap -.0135628 .012339

(-2.84)* (2.99)*

mobilephonesperthou .074696 -.0468462

(1.07) (-1.73)

CMR 30

Page 31: Mobile Phones and Income Inequality

telephonelinesperthou .2578993 .2790329

(2.99)* (0.000)

internetusersperthou --- ---

--- ---

urbanpoppct -.7104331 .0118871

(-2.67)* (0.08)

roadsperkmsq -12.16138 -8.0313907

(-1.24) (-1.73)

middleeastd -210.6998 21.23149

(-7.77)* (1.20)

asiaoceaniad -46.30768 -2.5564006

(-2.14)* (-0.42)

europed -1.74892 38.11273

(-0.07) (5.30)*

latinamercaribd -74.27385 -3.563659

(-3.17)* (-0.39)

africad -67.25812 -1.128347

(-3.06)* (-0.22)

Figure C – Original regression without internet users per thousand

Dependent Variable femaleworkersper~s carsperthou

Number of Obs 552 311

cons 422.7613 4.223622

(19.15)* (0.71)

gdppercap -.0137419 .0128371

(-2.78)* (2.89)*

mobilephonesperthou --- ---

--- ---

CMR 31

Page 32: Mobile Phones and Income Inequality

telephonelinesperthou .2187356 .2700433

(2.38)* (0.000)

internetusersperthou .4635008 -.1338457

(1.74) (-1.23)

urbanpoppct -.7091357 .0194055

(-2.51)* (0.10)

roadsperkmsq -12.63913 -9.139631

(-1.22) (-1.68)

middleeastd -204.9268 21.27162

(-7.31)* (1.13)

asiaoceaniad -39.59707 -4.148859

(-1.81) (-0.53)

europed 3.248475 36.99437

(0.13) (4.30)*

latinamercaribd -73.13388 -4.203659

(-3.07)* (-0.40)

africad -69.83582 -1.466359

(-3.14)* (-0.24)

Figure D – Original regressions without mobile phone subscribers per thousand

Bibliography

Galor, Oded and Daniel Tsiddon. “Income Distribution and Growth: The Kuznets Hypothesis Revisited.” Economica. Vol. 63. No. 250 (1996): S103.

“Gini Coefficient.” Wikipedia: The Free Encyclopedia. Accessed at http://en.wikipedia.org/wiki/Gini_coefficient

Goodman, David N. “Used phones drive Third World wireless boom.” MSNBC. October 29, 2006. accessed at http://www.msnbc.msn.com/id/15434609/page/3/

CMR 32

Page 33: Mobile Phones and Income Inequality

Hadingham, Wenona et al. “Mobile Communications in South Africa, Tanzania and Egypt: Resultsfrom Community and Business Surveys.” Africa: The Impact of Mobile Phones. Vodafone Policy Paper Series. March 2005. 50.

“In Vietnam -- and across developing world -- cell phones play vital role in fueling growth.” Technology Review. January 27, 2007. accessed at http://www.technologyreview.com/Wire/18117/page2/

Kirkpatrick, David. “Tech targets the Third World.” Fortune on CNNMoney.com. December 22, 2006. accessed at

http://money.cnn.com/2006/12/20/technology/fastforward_thirdworld.fortune/index.htm

Maney, Kevin. “Gates sees cellphones as a way to help Third World.” USA Today Online. January 31, 2006. accessed at

http://www.usatoday.com/money/industries/technology/maney/2006-01-31-gates_x.htm

Murph, Darren. “Motorola to roll out cellphone-charging bicycle in ‘emerging markets.’” Engadget. January 9, 2007. accessed at

http://www.engadget.com/2007/01/09/motorola-to-roll-out-cellphone- charging- bicycle-in- emerging-mar/

Saltz, Ira S. “Income Distribution in the Third World: Its Estimation via Proxy Data.” American Journal of Economics and Sociology. Vol. 54. No. 1 (1995): 17-18.

Sullivan, Nicholas P. You Can Hear Me Now: How Microloans and Cell Phones are Connecting the World’s Poor to the Global Economy. San Francisco: Jossey-Bass, 2007.

“Wireless Quick Facts: December 2006.” CTIA: The Wireless Association. accessed at http://www.ctia.org/media/industry_info/index.cfm/AID/10323

CMR 33