Telephone Demand and Economic Growth in Africa1 Kwabena Gyimah-Brempong Department of Economics University of South Florida 4202 East Fowler Avenue Tampa, FL 33620 email:
[email protected] Tel: (813) 974 6520 and John Agyei Karikari Center for Economics US Government Accountability Office Washington, DC, 20548 email:
[email protected] March 1, 2007
1
Paper to be presented at Annual CSAE Conference, St. Catherine College, Oxford University, Oxford, UK, March 19-20, 2007. Joe MacDougald provided outstanding research assistance. Any remaining errors, however are ours.
Abstract This paper uses panel data from African countries from 1992 to 2004 and a dynamic panel data (DPD) estimation method to investigate the demand for telephone services and the impact of telephone use on income growth in African countries. We find that telephone use has a positive and significant impact on income growth in African countries, all things equal. We also find that the demand for telephone is price and income inelastic in both the short and long runs. The demand for telephone services in Africa is strongly influenced by the size of the network. Finally, we find that mobile telephones are substitutes for fixed telephones. The results are robust to different specifications. Our results have telephone research as well as growth policy implications.
KEY WORDS: TELEPHONE DEMAND, CELLULAR, FIXED LINES, AFRICA, ELASTICITIES, NETWORK EFFECTS, SUBSTITUTION, INCOME GROWTH JEL: L96; O55; C33
1
Introduction
The growth of telephone use, especially mobile phone, in Africa over the last decade has been remarkable.1 Between 1992 and 2004, telephone use in Africa increased by a factor 7.5, fixed-line subscriptions per 100 people increased 227% (from 1.69 to 3.68 per 100 people) while internet use per 1000 people increased by 2,869% (from 1.13 to 32.60) during this period.2 This rapid increase has spawned several business possibilities. Indeed, The Economist argues that while the personal computer transformed the economies of developed and Asian countries, it is the mobile phone that is transforming the economies of Africa.3 Although the growth rates of telephone use differ across countries and regions of Africa (ranging from a low of .08% for Guinea to 75% for South Africa), the upward trend is clear and very strong. What are the determinants of this dramatic growth in the use of telephone services and what are the implications of this rapid increase in demand on incomes? In spite of this dramatic growth rate and its possible effects on the growth rate of income, there has been virtually no empirical studies that investigate the determinants of demand for telephone in Africa and how increased use of telephone affects the growth of income in African countries.4 Although there is anecdotal evidence and few empirical evidence to support the notion that telecommunications development has positive impacts on economic growth (Roller and Waverman: 2001, Cronin et al : 1993b, c) or effects sectoral productivity (Correa: 2006), there has been no attempt to investigate this issue for African countries. Yet, Africa is the region of the world where telphone use could be more important for its growth than any other region of the world. The low level of technological development, lack of infrastructure in rural areas (especially electricity) to power computers implies that telephones, especially, mobile phones will play the role that the personal computer played in the development of other regions of the world. It is therefore important to understand the effect telephone use has on economic growth and what determines the demand for it. This paper uses panel data to empirically investigate the determinants of the demand for telephones and the effect of telephone use on the growth rate of incomes. We do so by estimating a demand for telephone equation and income growth equation using a dynamic panel estimator that accounts for endogenous regressors and dynamic effects. We allow telephone demand to influence the growth rate of income. Although telecommunications emcompass a wide array of technologies, we focus
1
on telephone because it is the form of communications technology that is most relevant to Africa. This paper makes several contributions to the literature on the demand for telephone services in developing countries and the effect of telecommunications on income growth. First, unlike other research on the subject that relies on either time-series or cross national data, we use panel data and employ dynamic panel data estimation methodology that produces consistent results in the presence of dynamics. We also control for endogenous regressors, such as the price of telephone services and income in estimating the equations. We allow for possible substitution effects between mobile and fixed-line phones, control for capacity constraints, as well as allow for network effects in the demand for telephone services. Finally, this is the only study we are aware of that combines the investigation of telephone use and the effects of telephone use on income growth. Our results are briefly summarized as follows. We find that the demand for fixed telephone services is a lagged adjustment process with a very long adjustment time. The demand for the fixed line services is positively related to per capita income and the availability of public pay phones. In the short-run, the demand for fixed telephone subscription is price inelastic with an estimated elasticity of about 0.02. In the long run demand for telephone services is also price inelastic but the estimated elasticity increases to about 0.66. We also find that fixed telephone demand and mobile telephones are substitutes in consumption–an increase in the price of mobile phone services leads to an increase in the demand for fixed telephone lines. Capacity constraints, as measured by capacity of local exchanges and wait times to receive telephone connection are significantly correlated with the demand for fixed telephone lines. Finally, we find that telephone use has a positive and significant impact on the growth rate of per capita income in African countries, all things equal. Our results are robust to specification and estimation methods, and they have important policy implications for development policy. The rest of the paper is organized as follows: Section 2 reviews the literature and briefly describes the growth and telephone use equations that we estimate. Section 3 describes the data and estimation method while section 4 presents and discusses the statistical results. Section 5 concludes the paper.
2
2
Previous Studies and Estimating Models
In this section, we briefly review the relevant research on the telecommunications and economic growth as well as introduce the equations we estimate. The first subsection reviews the relevant literature while the second subsection introduces the growth and demand equations we estimate.
2.1
Previous Studies
The literature on the demand for telecommunications, especially the demand for telephone services have been growing rapidly. Although most of the studies have focused on the demand for long distance telephone services in developed countries or international telephone services between developed and developing countries, the principles developed in these studies are relevant for our study. We therefore mention a few of the studies that are relevant to our study. We first mention a few studies that are relevant to the telephone demand equation and then mention the studies that look at the relationship between telephone use and income growth. A large number of studies of the demand for telephone services have focused on estimating the price and income elasticities of demand for telephone services. Gyimah-Brempong and Karikari (2002, 2001) and Karikari and Gyimah-Brempong (1999) investigate the demand for international telephone calls between the US and Africa and find low price and income elasticities in both the short run and long run. Trotter (1996), Agiakloglou and Yannelis (2006), Lurdes and Martins (2003), and Garen-Munoz and Perez-Amaral (1998) investigate the demand for long distance phone calls and find that demand is influenced by price of telephone calls, income and network effects. The estimated price and income elasticities are generally less than unity. In recent years, there has been a large number of telephone demand studies that emphasize the substitution or complementarity between fixed and mobile telephone services. While some of these studies find substitution between mobile phones and fixed phones systems using consumer phone data (Rodini et al : 2003, Vagliasindi et al : 2006, Madden and Coble-Neal, Okada and Hatta: 1999, others find that mobile phones and fixed phones are moderate substitutes. Vagliasindi et al (2006) find that the lower the penetration rate of fixed phones, the stronger the substitutability between fixed and mobile phones, all things equal. This may be similar to the African situation since telephone penetration rates are low in Africa compared to other parts of the world. These studies find substantial cross-price elasticities. In addition to price, substitution between fixed and
3
mobile phones is also influenced by capacity constraints of fixed lines (Minges: 1999) as well as network externalities (Doganoglu and Grzybowski: 2007). Although studies of the relationship between telephone infrastructure and economic growth dates back to the 1970s, it is only recently that researchers have applied modern and powerful techniques to investigate the relationship.5 In a very early study, Cronin et al (1993b) use county data from the state of Pennsylvania and Granger causality methodology to investigate the causal relationship between telecommunications investment and income growth. They find a bi-directional causal relationship between the growth of income and telecommunications investment in Pennsylvania. In a companion study, Cronin et al (1993c) investigate the relationship between telecommunications investment and sectoral productivity in the US economy using a Granger causality approach. As in the earlier study, they find a bi-directional causality between telecommunications investment and sectoral productivity. They find that the effect of telecommunications investment on productivity growth varies across sectors. Correa (2006) combines input-output methods and econometric analysis to investigate the relationship between the diffusion of telecommunications technology on sectoral productivity and economic growth in the UK. Using input-output methods to isolate the inter industry relationship, she finds that in addition to rapid productivity growth in the telecommunications sector itself, increased productivity in the sector increases productivity growth in other sectors as well. The result is that the contribution of telecommunication investment and the diffusion of telecommunications technology to economic growth far exceeds investment and productivity growth in that sector. This suggests that telecommunication technology and its use should be included in traditional growth equations. Roller and Waverman (2001) uses panel data and a simultaneous equation model that endogenizes the demand for investment in telecommunications infrastructure to investigate the relationship between telecommunications investment and economic growth in OECD countries. Estimating the model with a variety of nonlinear estimators, they find that telephone penetration rate has strong positive effect on income growth while the demand for telephone infrastructure is dependent on income and telephone price. They estimate own-price and income elasticities far in excess of unity suggesting that telephone demand in OECD countries is highly price and income elastic, a result that is quite different from the results of most researchers. We note that the demand equation does not account for possible substitution or complementary effects, or network effects. In the same way, 4
the income growth equation does not account for institutions or the policy environment, factors that are deemed important in modern growth empirics. Most of the studies mentioned above focus on the effects of telecommunications investment on income growth. We feel however, that it is more the use of, rather than investment in telecommunications technology that impact economic growth, because these investments are generally low. We therefore focus on the use of telecommunications technology in this paper. None of these studies combines the demand for telephone services and the effects of telephone demand on income growth and none of the studies focuses on Africa where telephone is playing the role played by the personal computer in fostering economic growth in the developed world.
2.2
Model
In this section, we provide a brief introduction to the telephone demand and income growth equations we estimate. The model we use to investigate the relationship between telephone demand and income growth is similar to the model developed by Roller and Waverman (2001). We postulate a Barro-type income growth equation with telephone use as an additional explanatory variable. The income growth equation is given as y˙ = y(y ˙ 0 , k, xgow, X), where y˙ and xgow are growth rates of per capita income and exports respectively, k is capital growth rate, y0 is initial income, and X is a vector of institutional policy variables. We see the demand for telephone (T ELd ) as a function of income (y), the price of telephone services (p), the price of related goods (pr ), and a vector of demand shifters (Z). The supply of telephone services depend on p and a vector of supply shifters (W). Equating the supply and demand functions and solving for telephone, we obtain the desired telephone demand equilibrium as a function of prices, income, and Z and W. The desired telephone equation can then be written as: T EL = tel(p, pr , y, Z, W). To estimate the income and telephone demand equations, we need to define the variables and provide specific functional forms. The telephone demand equation we estimate for African countries is based on the demand equation estimated by earlier researchers (Roller and Waverman: 2001, Agiakloglou and Yennelis: 2006, Cronin et al : 1993a, Doganoglu and Grzybowski: 2007, Gyimah-Brempong and Karikari: 2001, Karikari and Gyimah-Brempong: 1999, Garin-Munoz and Perez-Amaral: 1998, Okada and Hatta: 1999, among others). We assume that the desired demand for telephone services (tel∗ ) is a function of the price of telephone services, the prices of related goods, income, and capacity constraints. For fixed telephone lines, the relevant related good is the mobile phone while the 5
reverse is true in the case of the demand for mobile telephone services. The utility one derives from telephone services depends on the ability to talk to another person on the telephone; the more people one can talk to on the telephone, the higher the utility one derives from subscribing to the telephone network, hence the demand for telephone services will increase with the size of the telephone network. We therefore include the size of the telephone network as a regresoor in the telephone demand equation. The desired telephone demand equation we estimate is written in the log linear form as: lntel∗ = α0 + α1 lnp + α2 lnpr + α3 lny + α4 lncap + α5 lnpubpay + α6 lninternet + α7 lnwaitlist + ε
(1)
where p is the price of telephone services, pr is the price of related goods, y is income, caplocal is the capacity of the local network, pubpay is the number of public pay phones, internet is the number of internet users, waitlist is the number of people waiting to be connected to the telephone network, ε is a stochastic error term, α s are coefficients to be estimated. We use cap and pubpay to proxy network effects while waitlist is used to measure capacity constraint in this paper. We have also included the number of internet users as another possible network effect. We expect the coefficients of y, cap, pubpay and internet to be positive while those of p and waitlist are expected to be negative, all things equal. We cannot sign pr a priori. The actual quantity of telephone services demanded in any period may deviate from the desired quantity as shown in (1) above. This is especially the case of African countries with severe capacity constraints. The actual quantity demanded will take time to adjust towards the desired amount. We posit a Stone-Nerlove adjustment process of the form: telt − telt∗ = λ(telt∗ − telt−1 ) where tel is actual demand, 0 < λ < 1 is the fraction of the deviation of actual telephone demand that corrected in each period, and all other variables are as defined above. Substituting (1) into the adjustment process and solving for telt , we obtain the demand for telephone service as: lntel = α0 + α1 lnp + α2 lnpr + α3 lny + α4 lncaplocal+ α5 lnpubpay + α6 lninternet+α7 lnwaitlist+λtelt−1 +ε(2)Equation 2 is the telephone demand equation we estimate. The short run elasticities are α1 , α2 and α3 for own-price, cross-price and income
6
respectively. The long run elasticities of demand are α1 /(1 − λ), α2 /(1 − λ), and α3 /(1 − λ) for own-price, cross-price, and income respectively. There are several possible ways in which telecommunications can affect economic performance. In this paper, we focus on the effect of telephone use on the growth of real per capita income in African countries. In the last two decades, productivity growth in the telecommunications sector has outpaced productivity growth in other sectors of the economy. If aggregate economic growth is partly a function of weighted productivity growth in all sectors of the economy, then the above average productivity growth in the telecommunications sector means that the sector contributes more to growth of the aggregate economy than its size. Second, improved telecommunications reduces transactions cost and facilitates better communications and exchange of ideas that could increase total factor productivity. Third, improved telecommunications can spawn new businesses as it happened with the personal computer. Finally, increased efficiency and use of telecommunications improve performance of other sectors of the economy (Corea: 2006, Roller and Waverman: 2001, and Cronin et al : 1993, a, b). In view of the important role telecommunications could play in the growth process, it should be included as a factor in an aggregate growth equation. The equation we use to investigate the growth impact of telecommunication services is a Barrotype growth equation with telephone use as an explicit growth factor. We envision the growth of per capita income as a function of initial income, investment in physical capital, export growth rate (following Feder: 1983), government consumption, external aid, and telephone use as an input. Since this growth equation is now standard we do not spend time here to develop it. We expect a positive relationship between income growth and physical capital formation as well as export growth. Consistent with the literature we expect a negative relationship between initial income, and government consumption on the one hand and economic growth on the other. The relationship between aid and income growth continues to be controversial. Given the inconclusive nature of the relationship found in the literature, we leave its sign as an empirical issue. Finally, given our discussion above, we expect a positive relationship between telephone use and income growth. The income growth equation we estimate is given as: y˙ = β0 + β1 y0 + β2 k + β3 xgrow + β4 aid + β5 tel + β6 govcon +
(3)
where y˙ is the growth rate of per capita income, y0 is initial income, k is investment rate, xgrow is the growth rate of real export earnings, aid is net inflow of external aid and govcon is government 7
consumption, is a stochastic error term, and betai s are coefficients to be estimated. While this equation is similar to the growth equations that have been estimated in the growth literature, we note that the inclusion of tel makes it different from those that have been estimated in the growth literature. We note also that there are several possibly endogenous regressors (e.g. xgrow, k, and tel) in this growth equation hence we need to use an estimator that can account for endogeniety.
3
Data and Estimation Method
3.1
Data
The dependent variables in our study are the growth rate of per capita income and the penetration rates of telephone usage. We measure the growth rate of income gdpcapgr as the annual growth rate of real per capita GDP with 2000 as the base year. There are several possible ways to measure telephone use, such as the number of subscription and airtime use. One could also made a distinction between fixed phone demand and the demand for cellular phones. The data are generally not comprehensive or adequate. For instance, where the data have been split into air time use for fixed and mobile lines, the data are extremely spare. Due to lack of data on air time use, we measure our demand variable as the number of fixed line subscription per 100 people (f ixedsubs) in a country in a year. In essence, our demand variable measures the penetration of fixed telephone lines rather the demand for air time. The explanatory variables in the study include the lagged value of GDP per capita y1 . We follow earlier researchers and measure k as the gross fixed capital formation/GDP ratio while we measure xgrow as the growth rate of real export earnings. This growth rate reflects both quantity and price changes of exports, hence we cannot ascribe the growth rate to volume or price changes. We measure external aid aidgni, as the ratio of net external aid disbursement to GNI and government consumption govcon, is measured as the total government consumption expenditure/GDP ratio. We measure initial income y0 as the one period lag of real per capita income. We measured f ixedprice as the monthly charge for residential fixed telephone subscription in a country, internet1000 is measured as the number of internet users in a country per 1000 people, while mobileprice is the monthly subscription rate for mobile telephone in a country in a year. pubpay is the number of public pay phones per 1000 people in a country in a year while caplocal is the capacity of local exchanges to handle phone calls. Finally, waitlist is measured as the percentage
8
of the population on a waiting list for a fixed line telephone. All nominal variables (except income) were converted to real values with 1995 as the base using the national CPI. Data for the telephone demand variables (fixedsubs, f ixedprice, internet, pubpay, caplocal, and waitlist) were obtained from International Telecommunications Union (ITU) World Telecommunications Indicators, 2006, Online edition, while data for the income growth variables (gdpcapgr, y, xgrow, k, govcon) were obtained from the World Bank’s World Development Indicators, 2005. Data for aidgni were obtained from the Development Assistance Countries (DAC) International Development Statistics Online, 2006 version. The data are for 46 countries over the 1992-2004 period.6 However because we did not have observations for all variables for all years, we had an unbalanced sample of 522 observations for the income growth rate equation and 412 for the telephone demand equation. Summary statistics of the sample data are presented in table 1. It is clear from the summary statistics of the data that telephone use varies significantly in the sample as indicated by the standard deviation relative to the mean. While a large part of the variation is due to variation across countries, the variance in the sample is due mainly to dramatic changes in telephone use over time. Accompanying this increase in use is the falling prices of both fixed line and mobile telephone services over time. Despite the rapid growth, the summary statistics suggests that penetration rates are very low while there is a serious binding capacity constraint as evidenced by the large waitlist. The data also suggest that average per capita income and its growth rate tends to be very low although they exhibit high variation across countries and through time as indicated by the large standard deviations relative to the sample mean.
3.2
Estimation Method
The income growth and telephone demand equations we estimate have endogenous regressors (investment, aidgni, fixedprice, among others) as well as country heterogeneity. It is well known that in such cases, the fixed effect (FE) and the random effects (RE) estimators are not consistent. Under these circumstances, researchers have either used an instrumental variable (IV) or general method of moments (GMM) estimators to consistently estimate the growth equations. A consistent estimator that has recently been used by researchers to estimate cross-country growth regressions in a panel format is Arellano and Bond’s Dynamic Panel Data (DPD) estimator (Arellano and Bond: 1991). This estimator is a GMM estimator that uses lagged levels of endogenous and predetermined 9
regressors as instruments in a difference equation. Arellano and Bond proposed a one-step and two-step estimators—with the two step estimator being the optimal estimator. The difference between the two estimators is the weighting matrix. The one-step estimator is obtained when the weighting matrix is the average covariance matrix of P Z v¯i given by AN = (N −1 i Zi0 HZi )−1 where H is a T − 2 square matrix with 2s in the main diagonal, -1s in the first sub-diagonal, and 0s everywhere else. The optimal two step estimator replaces the H matrix with an estimated variance-covariance matrix formed from the residuals of a preliminary consistent estimate of θ. The optimal choice of AN for the two step estimator is given P as: AN = VˆN = N −1 i Zi0 vˆ¯i vˆ¯i Zi where vˆi are the residuals obtained from a preliminary consistent estimate of θ. The two estimators will be asymptotically equivalent if the error terms are spherical. There is no reason to believe that the error terms are spherical, hence we use Arellano and Bond’s two-step estimator to estimate the equations in this study. The DPD estimator consistently estimates dynamic panel data equations and has been used in some recent panel data studies of the demand for telephone services (e.g. Gyimah-Brempong and Karikari: 2002, 2001, and Karikari and Gyimah-Brempong: 1999). However when the series are persistent, lagged levels of endogenous and predetermined regressors tend to be weakly correlated to their subsequent first differences, thus leading to biased estimates on account of weak instruments. Blundell and Bond (1998) have introduced the “systems DPD” estimator to correct this problem. The “systems estimator” adds a levels equation with lagged values of first differences of endogenous and predetermined regressors as instruments to the difference equation and jointly estimate the two equations as a system. We use the systems estimator to estimate both the growth and investment equations. In our estimation, we base all statistical tests on small sample statistics. In the presence of regressors that are correlated with the error terms, the FE estimator produce inconsistent estimates while the DPD estimator produces consistent estimates. On the other hand, if all regressors are exogenous, the DPD estimator produces consistent but inefficient estimates while the FE estimator produces both consistent and efficient estimates. We therefore use a Hausman test to test for the exogeneity of regressors. We also test for the presence of second order serial correlation since the validity of the DPD estimates depend crucially on the absence of autocorrelated errors. Although the FE estimator is inconsistent in the presence of dynamics, we nevertheless provide FE estimates of the equation for comparison with the DPD estimates.
10
4
Results
The statistical results of our study are presented in tables 2 and 3. Table 2 presents the estimates of the income growth equation while table 3 presents the estimates for the telephone demand equation. As indicated above, we did not have diasgreggated data to be able to estimate the demand for both fixed and mobile phone use, hence the estimates here refer to subscription to fixed phone lines. Column 2 of table 2 presents a fixed effects estimates of the growth equation, column 3 presents the DPD estimates without a constant term, column 4 presents DPD estimates of the growth equation with a constant term while cloumn 5 presents the DPD estimates with time dummies. In general, the growth equation fits the data reasonably well. In particular, there is no second order autocorrelation, and the Hansen test does not reject the overidentifying restriction. We also reject the null hypothesis that all slope coefficients are jointly equal to zero at any reasonable level of confidence. Finally, the Hausman test suggests that not all regressors are exogenous, hence the DPD estimator is the appropriate estimator for this equation.
In column 2, the coefficients of k and xgrow are positive and significant while those of y1 estimatesarequalitatively .05 or better suggesting that telephone use has a significantly positive impact on income growth, all things equal. However, given that the FE estimator may be inappropriate for estimating the growth equation, we discuss the DPD estimates in the succeeding paragraphs. In columns 3-5,the estimates of k, and xgrow are positive and significant at α = .05 or better, suggesting that there is a positive and significant relationship between the growth rate of per capita income and physical capital investment and export growth on the other. The coefficients of y1 and govcon are negative and significant, indicating that there is a negative relationship between initial income and government consumption on the one hand and income growth on the other. The coefficients of y1 and govcon are negative and significant, indicating that there is a negative relationship between initial income and government consumption on the one hand and income growth on the other. These estimated coefficients are consistent with our expectations and similar to the results obtained by other researchers. The coefficient of aidgni is negative and significantly different from zero at α = .05 or better in all specifications, an estimate that is similar in spirit to the results of research that find negative relationship between external aid and income growth in developing countries. Our major concern in this section of the paper is the relationship between telephone use and
11
income growth in Africa. In column 2 of table 2, the coefficient estimate of tel is positive and significant, suggesting a positive relationship between income growth and telephone use. Given that some of the regressors are endogenous, we cannot put much credence in this estimated coefficient. In column 3, the etimated coefficient of tel is positive, relatively large and significantly different from zero at α = .01. When we estimate the growth equation with a constant term as shown in column 4, the coefficient of tel remains positive, relatively large and significantly different from zero at α = .01. Finally, estimating the growth equation with time dummies neither changes the coefficient of tel nor its statistical significance. The estimates of the coefficient of tel in columns 2-5 in table 2 suggests that there is a positive relationship between telephone use and income growth in African countries. Given that telephone demand is lagged one period, we conclude that the direction of causation is from telephone use to income growth rather than the other way round. The estimated relationship between telephone use and income growth is robust to different specification. We note however that the FE estimate of tel is significantly different from and less precisely estimated compared to the DPD estimates. We do not investigate the mechanism through which telephone use affects income growth in Africa in this paper. However, it is possible that telephone use affects growth through reductions in transactions cost, increasing the efficiency of business interaction as well as contributing to increased productivity of other sectors of the economy (Correa: 2006, Cronin et al : 1993b, c). Our results have both research and growth policy implications. Given the strong positive relationship between telephone use and income growth we find, researchers should include the use of telephone (or its spread) as a regressor when investigating income growth rate in LDCs. This is especially true for Africa where the telephone is playing the role of a technology change agent. Failure to do so may imply a misspecification of the growth equation. From a policy point of view, our results suggest that policy makers should make reasonable efforts to improve the telephone infrastructure as a means to spur economic growth. The estimates of the telephone demand equation are presented in table 3. We estimated this equation in double-log form, hence the estimated coefficients can be interpreted as elasticities. Columns 3 to 5 present the DPD estimates of the telephone demand equation. Column 3 presents the estimates without a constant term, column 4 presents the estimates for the equation that include a constant term while column 5 presents the estimates for the equation that include time dummies. Column 2 presents the FE estimates for the purposes of comparison. Because of the constantly 12
falling prices over time due technical progress in telephony, adding the time dummy variables in column 5 washes out the effects of price and internet use. The estimates in column 5 may therefore not be appropriate for making inference about the demand for telephone in Africa. Generally, the equation fits the data reasonably well as indicated by the regression statistics. We reject the null hypothesis that all slope coefficients are jointly equal to zero and the Hausman statistic indicate that not all regressors are exogenous, hence the DPD estimator is the appropriate estimator for this equation. The coefficient of the lagged dependent variable in columns 3 to 5 is positive, relatively large, and significanty different from zero at α = .01 or better. This indicates a dynamic effects in the demand for telephone with a very long adjustment adjustment period. This is not surprising, given the inability of African teleocommunications providers to meet demand for telephone services.7 The estimate suggest that less than 5% of excess demand is satisfied in each period. The coefficient of y is positive and significant in columns 3-5 suggesting that telephone service in African countries is a superior good. This implies demand is likely to increase as income grows, all things equal. We note however, that the income elasticity of demand for telephones in Africa is very low. The coefficient of capacity of local exchanges is positive and generally significantly different from zero at convntional levels. The coefficient of pubpay is positive and significantly different from zero in the columns 3 and 5. The combination of the coefficients of caplocal and pub, our measures of network size suggests that there is a network effect in the demand for telephone use in African countries, all things equa. This results is similar to the results of research that finds strong network effects on the demand for telephone services elsewhere. This result suggests that as telephone networks expand in Africa, the demand for telephone use will increase faster than will be dictated by income increases or price reduction. The coeffiicient of the price of telephone is negative, relatively small but significantly different from zero at α = .01 in columns 3 and 4. The estimated coefficient implies short-run own-price elasticity of demand of less than .04. The implied long-run own price elasticity of demand ranges between 0.28 and 0.44. This implies that telephone demand in Africa is highly price inelastic, hence it makes very little economic sense to subsidize the price of telephones in Africa in order to stimulate demand. A better way may be to expand capacity and increase the size of the network. The inelastic price effect we estimate here is similar to the results of research that has investigated the demand for telephone services on both developed and develpong countries. It is, however, different from the 13
results obtained by Roller and Waverman (2001) who estimated very large own-price elasticities of demand in OECD countries. The ceoffient of the price of mobile phones (mobileprice) is positive and significant. The short run cross-price elasticity of demand is 0.01 while the long run cross-price elasticity is 0.4. This suggest that mobile phones are substitutes for fixed phones in African countries. For African countries where there is virtually no waiting time and no credit requirements for subscription to cell phones but there are long waiting lines for fixed lines, this result is not surprising. The estimated coefficient of mobileprice combined with the low own-price elasticities of demand for fixed phones suggest that increasing use of fixed telephones may not be achieved through decreasing the price of fixed phones. Perhaps African countries may be better off skipping this technology altogether and concentrate on building their cell phone networks.
5
Conclusion
This paper uses panel data from a large number of African countries over the 1992-2004 period and a dynamic panel data estimator to investigate the the effect of telephone use on the growth rate of income as well as investigate the determinants of telephone use. Using a two equation model, we find that telephone use has a positive and significant impact on the growth rate of income in African countries. We also find that income, price of telephone, as well as the size of the network have strong effects on the demand for fixed telephones. Finally, we find that mobile phones and fixed telephones are substitues in use in Africa. Our results are robust to different specifications and have both research and policy implications.
14
6
Notes
1. In this paper, we use mobile phone, cellular phone, and wireless phone interchangeably to refer to the same telephony technology. 2. This growth is calculated from ITU data and includes both fixed-line and mobile usuage. 3. “Buy, Cell Hold: Mobile Phones in Africa”, The Economist, January 27, 2007, p. 48. 4. Gyimah-Brempong and Karikari (2002, 2001) and Karikari and Gyimah-Brempong (1999) investigate the demand for international telephone calls between the US and Africa but do not investigate the demand for domestic telephone services. They did not investigate the effects of telephone demand on economic growth. The rapid growth of the telecommunications sector and its impact on other sectors of the economy is an issue that has not been well investigated within the African context. However, given the lack of data on this sector we focus on telephone use. 5. See for example, E. Bebee and E. T. Gilling, “Telecommunications and Economic Development: A Model for Planning and Policy Making”, Telecommunications Journal, August 1976, pp. 537543, and P. D. Shapiro, “Telecommunications and Industrial Development”, IEEE Transactions on Communications, March 1976. 6. The countries in the sample are: Algeria, Angola, Benin, Burkina Faso, Burundi, Cameroon, Cape Verde, Central African Republic, Chad, Comoros, Congo Republic, Cote d’Ivoire, Djibouti, Egypt, Equatorial Guinea, Ethiopia, Gabon, Gambia, Ghana, Guinea, Guinea-Bissau, Kenya, Lesotho, Libya, Madagascar, Malawi, Mali, Mauritania, Mauritius, Morocco, Namibia, Niger, Nigeria, Senegal, Seychelles, Sierra Leone, Somalia, South Africa, Sudan, Swaziland, Tanzania, Togo, Tunisia, Uganda, Zambia, and Zimbabwe.
15
7
References
1. Agiakloglou, C. and D. Yannelis (2006), “Estimation of Price Elasticities for International Telecommunications Demand”, International Advances in Economic Research, 12, 131-137. 2. Banerjee, A. and A. Ros (2004), “Patterns in Global Fixed and Mobile Telecommunicatins Development: A Cluster Analysis”, Telecommunications Policy, 28, 107-132. 3. Correa, L. (2006), “The Economic Impact of Telecommunications Diffusion on UK Productivity Growth”, Information Economics and Policy, 18 385-404. 4. Cronin, F. J., M. Gold, P. Hebert, and S. Lewitzky (1993a), “Factor Prices, Factor Substitution, and the Relative Deamnd for Telecommunications Across Across US Industries”, Information Economics and Policy, 5, 73-85. 5. Cronin, F. J., E. Colleran, P. Herbert, and P. Lewitzky (1993b), “Telecommunications and Growth: The Contributions of Telecommunications Infrastructure Investment to Aggregate and Sectoral Productivity”, Telecommunications Polivy, 17, 677-690. 6. Cronin F. J, E. B. Parker, E. Colleran, and M. Gold (1993c), “Telecommunications Infrastructure Investment and Economic Development”, Telecommunications Policy, 17 (6), 415-430. 7. Doganoglu,T. and L. Grzybowski (2007), “Estimating Network Effects in Mobile Telephony in Germany”, Information Economics and Policy, 19, 65-79. 8. Garin-Munoz, T. and T. Perez-Amaral (1998), “Economic Modelling of Spanish Very Long Distance International Calling”, Information Economics and Policy, 10, 237-252. 9. Gyimah-Brempong, K. and J. A. Karikari (2002), “Cost Shifting in Demand for International Telephone Services Between the US and Africa, Journal of Development Economics, 68, pp. 455477. 10. Gyimah-Brempong, K. and J. A. Karikari (2001), “The Effects of Capacity Constraints on US-African Telephone Traffic, Information Economics and Policy, bf 13, 1-18. 11. Karikari, J. A. and K. Gyimah-Brempong (1999), “Demand for International Telephone Services Between Africa and the US”, Information Economics and Policy, 11, 407-435. 12. Lurdes, M. and C. Martins (2003), “International Differences in Telecommunications Demand”, Information Economics and Policy, 15, 291-303. 13. Madden, G. and G. Coble-Neal (2004), “Economic Determinants of Global Mobile Telephony Growth”, Information Economics and Policy, 16, 519-534.
16
14.
Minges, M. (1999), “Mobile Cellular Communications in the Southern African Region”,
Telecommunications Policy, 23, 585-593. 15. Okada, Y. and K. Hatta (1999), “The Interdependent Telecommunications Demand and Efficient Price Structure”, Journal of the Japanese and International Economics, 13, 311-335. 16. Rodini, M., M. Ward, and G. Woroch (2003), “Going Mobile: Substitutability Between Fixed and Mobile Access”, Telecommunications Policy, 27, 457-476. 17. Roller, L. H. and L. Waverman (2001), “Telecommunications Infrastructure and Economic Development: A Simultanous Approach”, American Economic Review, 91 (4), 909-923. 18. Trotter, S. (1996), “The Demand for Telephone Services”, Applied Economics, 28, 175-184. 19. Vagliasindi, M., I. Guney, and C. Taubman (2006), “Fixed and Mobile Competition in Transition Economies”, Telecommunications Policy, 30, 349-367.
17
Table 1: Summary Statistics of Sample Data
Variable
Mean∗
Standard Deviation
Minimum
Maximum
f ixedsubs
2.8800
5.0253
0.0716
28.6949
telprice (US$)
0.0492
0.0529
0.0001
0.3685
mobileprice (US$)
0.1281
0.1855
0.0001
1.5029
xgrow (%)
6.5275
14.5164
-43.7045
123.8283
gdpcapgr (%)
1.4295
3.8451
-19.4642
25.1583
gdp2000 (PPP2000 )
2699.959
3105.544
466.197
18850.57
k (%)
21.7205
7.9829
0.2105
61.8511
internet
11.4217
23.6896
1.00
240.1117
aidgni
11.4324
10.887
-0.2775
78.892
mainlines
28.5212
49.8997
0.6533
286.6624
pubpay
0.0005
0.0009
0.00
0.0047
caplocal
517475.60
1369984.0
2515.00
12000000.00
waitlist (days)
93499.53
226422.10
3.00
1362758.00
N
412 ∗ these are unweighted averages.
18
Table 2: DPD Estimates of Growth Equation Variable
Coefficient
Estimates
k
0.2298∗∗∗ (7.13)
0.2107∗∗∗ (7.79)
0.2513∗∗∗ (6.82)
xgrow
0.0170∗∗∗ (3.80)
0.0123∗∗ (2.36)
0.0131∗∗ (2.59)
y1
-0.3057∗∗∗ (5.27)
-0.6414∗∗∗ (6.83)
-0.7233∗∗∗ (6.22)
f ixedsubs
0.0879∗∗∗ (5.90)
0.0482∗∗∗ (9.86)
0.0380∗∗∗ (4.26)
time
0.0018∗∗∗ (2.96)
0.0024∗∗∗ (3.12) -0.0427∗∗∗ (4.30)
aidgni
N
522
F
216.17
546.96
112.09
1st ord. ser. cor. 2nd ord. ser. cor. Hansen test Hausman m
-2.54 -1.36 28.76 [29] 68.27 [5]
-2.55 -1.36 29.92 [30] 74.89 [4]
-2.48 -1.27 27.15 [29] 69.87 [6]
+ absolute value of “t” statistics in parentheses. ∗ 2-tail significance at α = 0.10 ∗∗ 2-tail significance at α = 0.05 ∗∗∗ 2 tail significance at α = 0.01
19
Table 3: Estimates of Telephone Demand Equation Variable
Coefficient
Estimates
lag1.f ixedsubs
0.9521∗∗∗ (70.20)
0.9701∗∗∗ (59.12)
0.9207∗∗∗ (36.82)
1.0101∗∗∗ (7.01)
f ixedprice
-0.0132∗∗ (1.97)
-0.0198∗∗∗ (3.21)
-0.0351∗∗∗ (4.29)
-0.1075 (1.05)
mobileprice
0.0111∗∗ (2.53)
0.0123∗∗∗ (3.54)
0.0073 (1.42)
0.0032 (0.56)
internet
0.0028 (0.72)
-0.0084∗ (1.71)
-0.0040 (0.57)
0.432 (0.39)
pubpay
0.0022 (0.38)
0.0151∗∗ (2.02)
0.0048 (0.48)
0.2428∗∗∗ (4.78)
waitlist
-0.0003 (0.15)
-0.0021 (1.55)
0.0054 (0.26)
0.0259 (0.68)
caplocal
0.0025 (0.58)
0.0324∗∗∗ (3.32)
0.1246∗∗∗ (3.66)
0.0907 (1.47)
y
0.0681∗∗∗ (4.31)
0.0258∗∗∗ (2.65)
0.9010∗∗∗ (23.77)
1.0142∗∗∗ (9.11)
constant
-0.4592∗∗∗ (3.22)
-1.0229∗∗∗ (2.75)
-6.7340∗∗∗ (8.94)
no
no
yes
59349.88 -1.81 0.24 33.46[52] 88.17[8]
27282.76 -0.32 1.08 25.25 [50] 83.78 [8]
455.60 0.27 -0.21 12.40 [59] 78.92 [8]
time dummies N
Wald F 1st ord. ser. cor. 2nd ord. ser. cor. Hansen test Hausman m
412
12488 [8]
+ absolute value of “t” statistics in parentheses. ∗ 2-tail significance at α = 0.10 ∗∗ 2-tail significance at α = 0.05 ∗∗∗ 2 tail significance at α = 0.01 20
