Methodology for long-term forecasts

The Economist Intelligence Unit has traditionally produced five-year forecasts. However, many companies make strategic business decisions over timeframes in excess of five years. Therefore, the Economist Intelligence Unit has developed a methodology for producing long-term economic forecasts, which has been applied to the 60 largest economies. Our long-term projections will provide information to facilitate such decisions made over these longer timeframes. Long-term forecasts and scenarios are also the key to understanding some of the big economic issues that will shape global business in the coming decades.

The methodology is distinct from that used to generate our five-year forecasts-that is based on a "demand side" forecasting framework which assumes that supply adjusts to meet demand either directly by changes in output, or by the drawing down (or building up) of inventories. Such a framework is appropriate for constructing short- and medium-term projections where output can deviate substantially (but temporarily) from its long-run sustainable level. But a demand side framework is not appropriate for forecasting over the long term. Instead, we utilise a supply side framework, in which output is determined by the availability of labour and capital equipment, and the growth in productivity.

The key output of our long-term model is a forecast of real GDP growth per capita, which can be combined with population growth forecasts to give a forecast for each country for real GDP growth. From this building block, we are then able to make projections for a series of market sizing variables important for long term business planning. These include GDP in US dollar terms and at PPP conversion rates, consumer spending, and exports and imports.

The Economist Intelligence Unit is well placed to produce long-term projections and scenarios-we have considerable experience in tracking and forecasting a series of economic and institutional factors which our analysis suggests are closely related to long-term growth prospects. These factors include the availability of an educated workforce, the openness of the economy to trade, the quality of institutions (including the legal framework and the quality of the bureaucracy), fiscal policy, the degree of government regulation, movements in the population of working age relative to the overall population, and the development of information and communication technology infrastructure. In addition, the income gap between each country and the global technological leader (the US) is important as this illustrates the potential for economic catch-up by importing ideas and techniques. Forecasts of GDP growth per capita can then be combined with demographic projections (taken mainly from the US Census Bureau) to give forecasts for overall GDP growth. This is explained in more detail below.

Growth projections
The main building blocks for the long-term forecasts of key market and macroeconomic variables are long-run real GDP growth projections. We have estimated growth regressions (based on cross-section, panel data for 86 countries for the 1970-2000 period) that link real growth in GDP per head to a large set of growth determinants. The sample is split into three decades: 1971-80, 1981-90 and 1991-2000. This gives a maximum of 258 observations (86 countries for each decade); given missing values for some countries and variables, the actual number of observations is 246. The estimation of the pooled, cross-section, panel data is conducted on the basis of a statistical technique called Seemingly Unrelated Regressions to allow for different error variances in each decade and for correlation of these errors over time.

The regressions, which have high explanatory power for growth, allow us to forecast the long-term growth of real GDP per head for sub-periods up to 2030, on the basis of demographic projections and assumptions about the evolution of policy variables and other drivers of long-term growth.

Definitions of variables
The dependent variable is GDPG: Average annual growth in real GDP per head, in the 1970s, 1980s and 1990s, measured at national constant prices.

The independent variables include:
LnGDPPL: The natural logarithm of GDP (adjusted for purchasing power parity—PPP) per worker (that is, per population aged 15-65) in constant 1980 US dollars at the start of each decade. Expressed as an index, US=1.
LnSCHOOL: The natural logarithm of the mean years of schooling of the population aged over 15 at the start of each decade. Missing values for some countries are filled in by estimating mean years of schooling on the basis of an equation relating mean years of schooling (where available) to gross primary school enrolment ten years previously, and to secondary and tertiary enrolment ratios five years previously.
LnLIFEEXP: The natural logarithm of life expectancy at birth at the start of each decade. This variable also enters the equation in squared form, reflecting diminishing returns to growth of increases in life expectancy at high levels.
OPEN: Updated Sachs-Warner index of openness—the fraction of years during each decade in which a country is rated as an open economy according to the following four criteria: (1) average tariff rates below 40%; (2) average quota and licensing coverage of imports of less than 40%; (3) a black-market exchange-rate premium that averaged less than 20%; and (4) no extreme controls (taxes, quotas, state monopolies) on exports.
INST: Index of institutional quality (on a scale of 1-10) that is an average of five sub-indices of measures of the rule of law, quality of the bureaucracy, corruption, the risk of expropriation and the risk of government repudiation of contracts. Forecast values are based on corresponding indicators from our business environment rankings.
LABPOP: The difference between the growth rate of the working-age population (aged 15-65) and the growth rate of the total population in each decade in the 1970-2000 period.
TOT: The average annual rate of change of the terms of trade in a given decade.
GOVSAV: The average government savings ratio in each decade (current government revenue minus current government expenditure) expressed as a share of GDP.
TRADESH: The average share of trade (exports and imports of goods and services) in GDP, lagged by one decade to deal with the endogeneity of growth and trade.
GOVREG: An index on a scale of 1-10 of regulation of product, credit and labour markets. For forecast periods, the composite index is based on seven indicators from three categories of our business environment rankings model—from Policy towards private enterprise (ease of setting up new businesses, freedom to compete, price controls); from Financing (openness of the banking system, financial market distortions) and from Labour markets (restrictiveness of labour laws, wage regulation).
LnICT: The natural logarithm of an index, on a scale of 1-10, of the development of information and communications infrastructure. ICT development is found to influence growth significantly only from the 1990s, with little or no impact in previous decades. For 1990 the index is measured simply on the basis of fixed telephone lines per 1,000 population. From 2000 a more sophisticated measure is constructed, reflecting the very rapid development of ICT. The composite ICT index is based on ten indicators. Six indicators are quantitative and rely on our forecasts of fixed-line telephone penetration (lines per 100 population); mobile telephone penetration (subscribers per 100 population); the stock of personal computers (PCs per 100 population); Internet users (per 100 population); the number of Internet servers (per million population); and broadband penetration (per 1,000 population). In addition, there are four qualitative indicators from our “e-readiness” model. These ! include the quality of Internet connections, the development of e-business, the development of online commerce and the exposure of the population to the Internet (“Internet literacy”). Each of the ten indicators is transformed into an index scaled 1-10. The composite ICT infrastructure/use index, on a 1-10 scale, is an average of the ten component indices.
Control variables include PRIMARY: Share of the exports of primary products in GDP at the start of a decade; TROPIC: Percentage of the land area within a country that has a tropical climate; COLONY: History of independent statehood—a dummy variable taking the value of 1 if a country was a colony before 1945; and, in some specifications, regional dummy variables.

Summary of findings
As in other studies, income per head and human capital are found to be important determinants of growth, with the coefficient on the logarithm of GDP per worker suggesting a relatively modest pace of convergence. The measures of institutional quality and of government regulation enter significantly in all specifications. We found a strong positive impact on growth of government savings and openness in all specifications. The criteria for classifying countries as open are quite permissive. The crucial aspect of trade policy captured by the measure is that it is a high level of distortion, rather than modest levels, that is deleterious for growth. The trade share variable is also moderately significant. The openness index (which is more of a true measure of policy) is hardly affected by the inclusion of trade/GDP shares in the equation. The correlation of the two measures is only .26. Although a tropical climate is highly significant, as is the share of primary exports in GDP,! other geographic indicators—such as access to the sea, distance from major growth centres and the proportion of the population residing near coastlines—were not significant. A colonial past (pre-1945) is found to have a significant negative impact on growth, even in the 1970-2000 period.

Productivity growth
The forecasts of GDP growth, of capital stock growth (based on estimated investment shares and assumed depreciation rates) and of growth in labour supply (based on projections of working-age population and assumptions on labour force participation) yield labour productivity growth and total factor productivity growth forecasts. The latter utilise the growth accounting identity, GY=b*GK+c*GL+A, where GY is growth of real GDP, GK growth of the capital stock and GL growth of human capital (the labour force adjusted for changes in skills). “A” stands for growth in total factor productivity; “b” and “c” are the shares of capital and labour in income.

Trade values are forecast on the basis of simple import (function of GDP and relative prices) and export functions. Forecast market exchange rates (that is, the differential between PPP and market exchange rates) depend on the differential in labour productivity growth between a country and the US.

Back to forecasting and methodology