AIDS, longevity and long-run income

Abstract

This article studies the long-run impact of HIV/AIDS on per capita income and education. We focus on the disincentive to human capital accumulation given by shorter life span. We work with a continuous time overlapping generations model with education and saving decisions, calibrated for a cross-section of countries. The simulations predict that the most affected countries in Sub-Saharan Africa will be in future, on average, 20% poorer than they would be without AIDS. Schooling will decline in some cases such as Botswana, South Africa and Zambia by more than 40%. The impact of population decline was found to be irrelevant.

Keywords:

• HIV/AIDS
• schooling
• long-run output
• Africa

JEL Classification::

• O11
• O40
• I10

Acknowledgements

I gratefully acknowledge the comments and input of Samuel Pessôa and two anonymous referees and the financial support of CNPq and INCT/FAPERJ.

Notes

¹ Kalemli-Ozcan et al. (2000), Boucekkine et al. (2002), Soares (2005), Ferreira and Pessôa (2007), and many others, study the effect of longevity on income and human capital.

² See Hamoudi and Birdsall (2004) and Soares (2006) for econometric evidence on the positive effect of longevity on schooling.

³ For instance: in a model without distortion, the simulated schooling is five to four times larger than observed values. Once we introduce τ _K and τ _H in the model, in most cases (especially for those countries in which schooling is above 2 years) the model almost matches observed values.

⁴ Of course we are not claiming that every country, particularly in Africa, will grow at this rate every year. However, we do claim that for the long run this is an adequate assumption. Although we have observed growth disasters, those are the exceptions, growth being the rule when we examine long periods of time, especially in modern history (Parente and Prescott, 2002). In contrast, as leapfrogging is also rarely observed among countries, it is not reasonable to assume that some nations could grow forever at rates above the leaders.

⁵ For instance, in the last 10 years, according to the Survey of Current Business, published by the US Department of Commerce, the average capital share in income of the educational services sector was only 6% in the United States. Most likely, the figure for Africa is even smaller.

⁶ This last assumption is necessary for a balanced growth path in which tuition increases at a rate equal to technological change.

⁷ We normalize variables as qη and for calibration purposes.

⁸ This is a long-run average for the investment/capital ratio, as given by NIPA, both evaluated at market prices.

⁹ We estimated a trend line for the variable RGDPW of the Summers and Heston database from 1960 to 1992.

¹⁰ From Equation 6 and the equilibrium expressions for N _C , N _S and N _W it is clear that we do not need to calibrate the parameter m.

¹¹ Note that in countries where the HIV/AIDS epidemic is under control, the model predicts that income will be above its 1985 trend, as is the case of the five countries at the bottom of Table 1.

¹² This contrasts with Young (2005), where saving rates are given. In this case, the reduction of population has a positive impact on per worker capital and wages and offsets the human-capital losses. In our model, everything else being constant, reductions in the population are followed by reductions in capital stock, so that k does not change much. It does decrease, however, because its return falls with education.

¹³ Actually, we changed the population growth rate for the sake of coherence. This, however, had no impact on the results.

¹⁴ In other words, in Table 1 it was shown that per capita output in Botswana decreased by 35% due to the life-expectancy effect. With the addition of the experience channel it now falls by 38.11%.

¹⁵ Of course if we had a more precise calibration of the impact of AIDS on each cohort, this result could change. As the epidemic evolves, the birth rate will decrease and so will the weight of younger cohorts in the labour force. This is why Haacker finds the experience effect to be higher in the medium than in the long run.

¹⁶ Labour supply in these models and in our study is exogenous and inelastic. This being so, the impact on output of absenteeism and poor physical performance are observationally equivalent, although the loss in output in the first case is due to fewer hours and in the second on account of lower productivity.

¹⁷ The median span between infection and death is estimated to be around 8 to 10 years in South Africa (Arndt and Lewis, 2000) but the direct symptoms of AIDS are mostly felt during the last 2 years of life of an HIV-positive person. Assuming that those are stationary figures, in a given year the AIDS to HIV prevalence ratio will be around 20% to 25%.