MOROCCO Selected Issues
Author: International Monetary Fund
Publication Date:
23 Nov 2005
eISBN:
9781451824773
ISBN:
9781451824773
Language:
English
Keywords:
ISCR; CR; capital stock; nominal GDP; foreign direct investment; contribution of total factor productivity; performance of a country; output gap; factors of production; Productivity; Production growth; growth performance of a country
Download PDF (2.6 MB)

This Selected Issues paper on Morocco analyzes Morocco’s growth performance over the past thirty-five years with a special focus on the more recent past. The paper presents stylized trends in growth; it first puts the overall growth performance of Morocco in an international perspective, including an analysis of the evolution of growth volatility, and then decomposes growth by major sectors to retrospectively identify patterns of structural transformation as well as the most dynamic sectors. The paper also analyzes accumulation of factors of production and total factor productivity growth through a growth accounting exercise.

Abstract

This Selected Issues paper on Morocco analyzes Morocco’s growth performance over the past thirty-five years with a special focus on the more recent past. The paper presents stylized trends in growth; it first puts the overall growth performance of Morocco in an international perspective, including an analysis of the evolution of growth volatility, and then decomposes growth by major sectors to retrospectively identify patterns of structural transformation as well as the most dynamic sectors. The paper also analyzes accumulation of factors of production and total factor productivity growth through a growth accounting exercise.

I. Morocco: Growth Performance1

A. Introduction

1. The macroeconomic environment in Morocco has been stable along many dimensions for more than a decade. Inflation has remained low, the balance of payments in surplus, international reserves high, and the public debt to GDP ratio has declined. Growth, however, has been weak and below that of comparable developing countries during that period. This chapter’s objective is to attempt a diagnostic of Morocco’s growth performance over the past thirty-five years with a special focus on the more recent past.

2. Section B presents stylized trends in growth. It first puts the overall growth performance of Morocco in international perspective, including an analysis of the evolution of growth volatility. It then decomposes growth by major sectors to retrospectively identify patterns of structural transformation as well as the most dynamic sectors. It continues with a decomposition of the demand side, highlighting in particular the evolving contributions of public expenditure and external demand over time. Finally, it analyzes flows of foreign direct investment.

3. Section C focuses on the accumulation of factors of production and total factor productivity growth through a growth accounting exercise. Following an estimation of a Cobb-Douglas production function for the Moroccan economy using time series techniques, overall growth is decomposed by factor and by period. Given the duality between the primary and nonprimary sectors in Morocco, a similar decomposition is done for the nonprimary sector before conclusions are drawn. The recent trends uncovered by the analysis are then used to construct three illustrative scenarios under which the Moroccan authorities’ growth and employment objectives in the medium term can be realized.

4. Based on a cross-country econometric analysis, Section D assesses whether the growth acceleration observed in the last few years in Morocco can be accounted for by main empirical determinants of economic growth, such as transitional convergence, cyclical reversion, structural policies, macro stability, and external conditions.

5. Conclusions from the analyses described above as well as policy implications are presented in Section E.

B. Stylized Trends in Growth

6. This section documents growth patterns in Morocco since the early 1970s. Growth has been lackluster and volatile, especially in the 1990s. The most recent years show some encouraging signs, such as a strong decline in the volatility of nonagricultural output, a recovery in investment and private saving as well as an increase in foreign direct investment flows. However, the performance of the economy still needs to improve to catch up with the recent trends of GDP and export growth observed in developing countries.

Overall growth

7. Morocco’s average growth2 performance since 1971 has been weaker than that of developing countries as a group.3 Indeed, over that period of three and a half decades, the real GDP growth differential with this group represented a little more than half a percentage point per annum (see Table 1).

Table 1.

GDP Growth, 1971–2004

(In percent)

article image
Sources: Moroccan National Accounts; WEO database; and IMF Staff estimates.

8. A breakdown into four periods4 reveals that this growth differential is the result of Morocco’s low growth since the early 1990s. Supported by large public investment projects, growth was strong in the 1970s. The growth rate then declined somewhat during the 1980s, but remained above that in the developing world, despite a balance of payments crisis in the first half of that decade. Affected by several severe droughts in a row, the economy’s performance then sharply deteriorated during the 1991–98 cycle and the negative differential with the average developing country reached two percentage points per annum. The economy has strengthened since 1998, but the differential has remained. Given the rapid expansion of Morocco’s population until recently,5 its growth performance looks relatively worse in per capita terms since the gap with the average developing country amounted to 0.8 percentage point per annum over the whole period.

9. Morocco’s performance also looks lackluster when compared to other nonoil and politically stable countries in the Middle East and North Africa region. In terms of GDP per capita growth during 1971–2004 (Figure 1a), Morocco did better than Jordan only, when that country had to absorb inflows of refugees in the early 1970s and early 1990s. Since 1991, Morocco is the weakest performer of the group, although there has been some improvement in the latest cycle as a consequence of an acceleration in GDP growth and a slowdown in population growth (Figure 1b).

Figure 1a.
Figure 1a.

Real GDP Per Capita, 1971–2004

(1971 = 100)

Citation: IMF Staff Country Reports 2005, 419; 10.5089/9781451824773.002.A001

Sources: Moroccan National Accounts; WEO database; and IMF staff estimates.
Figure 1b.
Figure 1b.

Real GDP Per Capita, 1991–2004

(1991 = 100)

Citation: IMF Staff Country Reports 2005, 419; 10.5089/9781451824773.002.A001

Volatility of growth

10. The volatility of output has been a constant feature of the Moroccan economy (see Figure 2a). The path of GDP has been subject to the volatility of the price of phosphate in the 1970s and to the vagaries of rainfalls since the early 1980s. These latter shocks have had a first-order impact on the fluctuations of agricultural production and of total output as a result. Given the negative correlation between volatility and growth observed in cross-country studies (see Ramey and Ramey (1995) for example), a high volatility has serious negative implications for welfare through both its direct and indirect effects.

Figure 2a.
Figure 2a.

Real GDP Growth Rate

(in percent)

Citation: IMF Staff Country Reports 2005, 419; 10.5089/9781451824773.002.A001

Sources: Moroccan National Accounts; and IMF staff estimates.

11. Nonagricultural GDP growth has also been volatile, although there is clear evidence that this volatility has been on a declining trend and has substantially diminished during the past ten years (see Figure 2b). The share of cereal production in the total nominal value of primary sector production fell from one third in 1980, the base year used for GDP calculations, to less than one fifth in 2003, while the shares of other less volatile components such as fruit and especially livestock have expanded. This evolution suggests that the direct spillover from the volatility of real agricultural production to that of real nonagricultural output could have declined over time. Moreover, greater stability of food prices has been observed in the last decade and the government has implemented some countercyclical policies in rural areas during the last two droughts (1999 and 2000). The combination of these factors may explain why the impact of real fluctuations in the primary sector on the rest of the economy has become more moderate.

Figure 2b.
Figure 2b.

Nonagricultural Real GDP Growth Rate

(in percent)

Citation: IMF Staff Country Reports 2005, 419; 10.5089/9781451824773.002.A001

12. When put in international perspective, the volatility6 of real GDP growth in Morocco has been higher than that of the average for developing countries since the 1980s, and among the highest in the group of comparable economies in the region (see Figure 3). However, the volatility of nonagricultural GDP growth over the past ten years looks quite low by international standards.

Figure 3.
Figure 3.

Volatility of Real GDP Growth

(in percent)

Citation: IMF Staff Country Reports 2005, 419; 10.5089/9781451824773.002.A001

Sources: Moroccan National Accounts; and IMF staff estimates.

The supply side: sectoral composition of growth

13. The sectoral composition of nominal GDP has not changed substantially over the past twenty-five years (see Figure 4). The share of the primary sector declined in the 1970s while that of the government sector increased. Since the early 1980s, the primary sector accounts for about 16 percent of total nominal output, with a share fluctuating from year to year, reflecting intermittent droughts and subsequent recoveries as described above.

Figure 4.
Figure 4.

Sectoral Shares in Nominal GDP

Citation: IMF Staff Country Reports 2005, 419; 10.5089/9781451824773.002.A001

Sources: Moroccan National Accounts; and IMF staff estimates.

14. The share of the industrial sector has not significantly expanded over time. The adjustment programs in the 1980s aimed at diversifying the tradable goods sector and reducing the share of commodities—subject to exogenous price shocks—in that sector (see Nsouli and others (1995)). The share of mining did indeed decline gradually over time—it now represents 1.6 percent of GDP and has more than halved in twenty years. However the share of the manufacturing sector in total GDP has remained in the same remarkably narrow range as in the earlier period, fluctuating between 16 percent and 19 percent. Within manufacturing, all subsectors grew at approximately the same rate over the period. For a long time, the strongest performer was the export-oriented subsector of textiles and leather, but its growth has slowed markedly since 1995.

15. Construction and public works grew at a low rate over the period—the sector even stagnated in real terms between 1975 and 1996. Arguably, this is due to unsustainable rates of public investment in the mid-1970s, but the analysis of the evolution of the components of investment (see Figure 5), in particular public works, suggests that infrastructure may not have expanded fast enough to meet the growing needs of the economy.

Figure 5.
Figure 5.

Components of Fixed Gross Capital Formation

(as a percentage of nominal GDP)

Citation: IMF Staff Country Reports 2005, 419; 10.5089/9781451824773.002.A001

Sources: Moroccan National Accounts; and IMF staff estimates.

16. The share of the tertiary sector has also remained relatively stable, around 38 percent of GDP, while the share of international tourism has risen steadily since the early 1980s. National accounts indicate that the most dynamic subsectors have been transportation, communications, and financial intermediation.

17. The sectoral composition of real GDP growth seems to have become more balanced over the past cycle. When analyzing the real growth of each sector through the four periods (Figure 6), it appears that the manufacturing and tertiary sectors have a dynamics which is very close to that of the economy in the aggregate, and that the growth of the government sector has been much stronger than overall growth except in the last period.7 This cycle is characterized by balanced contributions of all major sectors, and by an acceleration of the growth of the tertiary sector as a result of the dynamism of most of its components.

Figure 6.
Figure 6.

Real Sectoral Growth

(in percent)

Citation: IMF Staff Country Reports 2005, 419; 10.5089/9781451824773.002.A001

Sources: Moroccan National Accounts; and IMF staff estimates.

The demand side: external and domestic factors

18. The past few years have witnessed a revival of domestic demand. Strong growth in the 1970s was supported by the spectacular expansion of public expenditure (see Figures 7 and 9). The slowdown of government demand in the 1980s was counteracted by a remarkable (but volatile) expansion of exports of goods and services (see Figure 8) in a context of continuous depreciation of the real effective exchange rate. This rapid expansion came to a halt in 1991 and Morocco has not yet caught up with the high rhythm of export growth observed in the average developing country since 1990. As private domestic demand failed to regain vigor, following several consecutive droughts, the 1990s were a decade of weak growth. The moderate rebound which took place during the last cycle was helped by government demand, nongovernment investment, and an accelerating recovery of resident consumption. The export performance, however, remained tepid. The ongoing process of structural reforms, in particular the recently signed free trade agreements, should help exports regain some strength over the medium term, despite the negative shock to the textiles and clothing sector following the worldwide abolition of quotas on January 1, 2005.

Figure 7.
Figure 7.

Growth of Components of Real Demand

Citation: IMF Staff Country Reports 2005, 419; 10.5089/9781451824773.002.A001

Sources: Moroccan National Accounts; WDI database; and IMF staff estimates.
Figure 8.
Figure 8.

Real Growth of Exports of Goods and Services

Citation: IMF Staff Country Reports 2005, 419; 10.5089/9781451824773.002.A001

Sources: WEO database; WDI database; and IMF staff estimates.
Figure 9.
Figure 9.

Government and Nongovernment Investment

(in percent of nominal GDP)

Citation: IMF Staff Country Reports 2005, 419; 10.5089/9781451824773.002.A001

Sources: Moroccan Authorities; and IMF staff estimates.

19. Investment8 has recently regained vigor. It peaked in 1976 at above 30 percent of nominal GDP as a result of large government-supported infrastructure projects and then declined gradually until 1996, a year in which it fell below 20 percent of GDP (see Figure 9). It has recovered slightly since then and stood at 24.6 percent in 2004. Since the peak of the mid-1970s, a radical shift has occurred in the composition of investment as general government investment fell steadily from 11 percent to 2.4 percent over the past thirty years. At the same time, the share accounted for by the nongovernment sector9 fluctuated around a slightly upward trend.10

20. A cross-country comparison of investment to nominal GDP ratios (see Figure 10) reveals that Morocco has fallen behind the average developing country since the mid-1980s and has invested on average about 4 percent of GDP less than Tunisia—the best performer in the region—since 1970. However, this gap has disappeared in the past two years. Limited data availability and comparability on central government investment in many developing countries unfortunately preclude a complete cross-country comparison of nongovernment investment.

Figure 10.
Figure 10.

Investment

(in percent of nominal GDP)

Citation: IMF Staff Country Reports 2005, 419; 10.5089/9781451824773.002.A001

Sources: Moroccan National Accounts; WEO database; and IMF staff estimates.

21. Supported by strong remittances flows, national savings have also recently picked up. An important goal of the adjustment strategy at the end of the 1980s was to improve the overall savings performance, so as to reduce reliance on foreign savings. One can see on Figure 11 that this objective has been achieved only lately. Indeed national savings actually fell in the first part of the 1990s and bottomed at 17.3 percent of GDP in 1995. They have been recovering gradually since then, helped by strong inflows of remittances over the past few years. National savings stood at 27.3 percent of GDP in 2004.

Figure 11.
Figure 11.

Savings

(in percent of nominal GDP)

Citation: IMF Staff Country Reports 2005, 419; 10.5089/9781451824773.002.A001

Sources: Moroccan Authorities; and IMF staff estimates.

22. Reflecting these movements, the overall national savings-investment gap, which persisted above 5 percent of GDP until 1987 narrowed in a series of three steps, falling first to an average of 2.5 percent in the first half of the 1990s to almost zero in the second half, and then actually transformed into a significant surplus since 2001, an indication that there is currently room to increase investment.

Foreign direct investment

Figure 12.
Figure 12.

Foreign Direct Investment

(in percent of nominal GDP, three-year moving average)

Citation: IMF Staff Country Reports 2005, 419; 10.5089/9781451824773.002.A001

Sources: Moroccan authorities; IFS database; and IMF staff estimates.

23. Morocco’s attractiveness to foreign direct investment has gradually increased over the past twenty years. Following a period of restrictions on foreign ownership which lasted from the mid-1970s to the early 1980s, Morocco has attracted more and more foreign direct investment to reach a level of 1.55 percent of GDP on average in the 1990s. Large privatization deals have made FDI flows look abundant in the past few years and have enabled Morocco to overtake the average developing country and even the fastest growing countries of that group. Excluding privatization receipts, annual FDI flows amounted to US660 million on average during 2000–04, equivalent to 1.7 percent of GDP and 7.2 percent of total investment.

C. Sources of Long-Term Growth: Estimates for Underlying Growth

24. This section analyzes the contributions of capital, labor and total factor productivity to the long-term growth of Morocco’s output. It is found that over the past thirty-five years, most of Morocco’s growth performance can be attributed to factor accumulation, and that total factor productivity has not significantly contributed to the overall growth of the economy. However, the contribution of total factor productivity to the growth of the nonagricultural sector has significantly improved in the most recent period analyzed.

Growth accounting methodology

25. The following analysis of the long-term growth of Morocco’s output uses the growth accounting methodology. This technique decomposes the growth of output over a certain period into the contribution of capital, labor, and a residual, which is interpreted as the contribution of total factor productivity growth. This growth decomposition is done for the whole period 1971–2004 and four subperiods,11 both for the whole economy and the nonagricultural sector.12 Restricting the growth accounting to the nonagricultural sector makes it possible to strip out most of the impact of the year-to-year volatility of cereal output. Data sources are presented in Appendix I.

26. Before computing the contributions of each factor and of the residual, an aggregate production function is empirically estimated. It is assumed that the production process can be modeled with a Cobb-Douglas function. That is if Yt, Kt, Lt, and At are respectively output, capital, employment and total factor productivity in period t, the production process can be represented by the following equation:

Yt=AtKtαLt1−α(1)

where α is the elasticity of output to capital. For lack of data, the production function of the nonagricultural sector cannot be estimated. The elasticity of output to capital is therefore assumed to be identical in the nonagricultural sector and the whole economy.

After taking logarithms on both sides, equation (1) is then transformed into the following equation:

Ln(YtLt)=a+gt+αLn(KtLt)+εt(2)

which is the one that is estimated.13 The coefficients a and g are constants and εt is an error term.

Using the fact that for any variable Xt, Ln(Xt)—Ln(Xt-1) is “approximately” the growth rate Xt−Xt−1Xt−1 of the variable Xt, equation (1) is used to decompose the growth rate of output Yt−Yt−1Yt−1 into the growth rate of total factor productivity At−At−1At−1, the growth rate of capital Kt−Kt−1Kt−1, and the growth rate of employment Lt−Lt−1Lt−1. It then follows that

(Yt−Yt−1Yt−1)=(At−At−1At−1)+α(Kt−Kt−1Kt−1)+(1−α)(Lt−Lt−1Lt−1)(3)

Equation (3) is the “growth accounting framework” that is used in this analysis. The growth rate of total factor productivity At−At−1At−1 is interpreted as the contribution of total factor productivity to output growth, the term α(Kt−Kt−1Kt−1) as the contribution of capital to output growth, and the term (1−α)(Lt−Lt−1Lt−1) as the contribution of labor to output growth.

27. The above-described growth accounting framework has obvious shortcomings. First, “productivity” is a residual, and not a direct estimation of an improvement in the quality or performance of production factors. Second, this residual is affected by uncertainties related to the measurement of production factors and output. An error in the estimation of capital stock or labor can significantly change the magnitude of productivity growth. Third, it does not isolate the main factors that have contributed to productivity growth. There is no way to identify whether the improvement of productivity came from the quality of the capital stock, the quality of the human capital stock, or from the managerial or organizational skills of firms. Fourth, the share of each production factor is assumed to be constant throughout the period under consideration. This may not be the case for developing countries that typically tend to move from labor- intensive to capital-intensive activities over time.

28. In spite of the aforementioned shortcomings, the relevance of the growth accounting methodology cannot be ignored. As a first step, it helps to orient the analysis of an economy’s growth by trying to isolate, though in a relatively imprecise way, the contribution of factor accumulation and that of an improvement in the quality of these factors. Such an analysis must be complemented with other information on the historical evolution of such factors during the period under consideration.

Application to Morocco

29. Estimates of the production function for Morocco suggest that the growth rate of total factor productivity has been very modest on average. The long-run growth rate of productivity g and the elasticity of output to capital α are estimated for the Moroccan economy using equation (2).14 The results of the estimation are presented in Table 2 and are in line with a priori expectations.15 The estimated elasticity of output to capital (α) is about 0.4. The point estimate of the annual growth rate or productivity g is not statistically significant. This suggests that over the estimation period, productivity growth in Morocco has been nil on average.

Table 2.

Estimation of the Production Function

article image
1/t-statistics are in brackets. The sample covers the period 1960–2002. All equations include an unreported constant term

30. The results of the growth accounting analysis indicate that GDP growth was essentially driven by factor accumulation over the past thirty-five years (see Table 3). 16 The results of the growth accounting are broadly in line with the stylized trends presented in Section B. Growth in Morocco is mostly driven by factor accumulation and the contribution of total factor productivity is on average negligible. Capital accumulation was the main factor contributing to growth in the 1970s. Since the 1980s the contributions of capital and labor to growth were more or less the same. The contribution of productivity is negative for all the periods considered except during 1982–91. The negative contribution of total factor productivity to growth during some of the periods considered is partly due to the performance of the agricultural sector, which depends on weather conditions.

Table 3.

Growth Decomposition for the Whole Economy

(In percent)

article image
Source: IMF staff estimates.

31. However, the growth accounting exercise for the nonagricultural sector reveals that while the growth of the residual has also been close to zero on average over the whole period 1971–2004, its contribution has gradually improved over time in parallel with a declining contribution of labor (Table 4). Nonagricultural growth in the 1970s was impressive but relied both on unsustainable rates of capital accumulation and a strong expansion of employment. The residual was negative, maybe because the returns to public investment in infrastructure were rather low or did not materialize in the short-run. Nonagricultural growth slowed down sharply in the 1980s but the negative contribution of the residual was smaller. One possible explanation is that, following the investment boom of the previous decade, capacity utilization improved and employers became more selective in their hiring decisions. Consistent with this last hypothesis, employment growth slowed down by more than one percentage point per year during that period and urban unemployment shot up from 12.7 percent in 1982 to 17.3 percent in 1991. The 1991–98 cycle was characterized by several droughts and negative contribution of the residual in aggregate growth as described above, but the residual in the nonagricultural sector actually stabilized. Capital accumulation remained weak, while employment growth continued to decline in a context of increasing urban unemployment and a declining participation rate. More recently during the 1998–2004 cycle, investment picked up but the contribution of employment dropped further. The contribution of the residual in nonagricultural growth reached a significantly positive level for the first time in thirty-five years, indicating that structural reforms may have started to enhance total factor productivity. These productivity improvements, however, seem to have taken place at the expense of employment growth in the short run.

Table 4.

Morocco: Growth Decomposition for the Nonagricultural Sector

(In percent)

article image
Source: Moroccan authorities; Annuaires Statistiques; and IMF staff estimates.

The role of total factor productivity in increasing employment

32. This subsection contrasts scenarios relying on high investment levels with others relying on high total factor productivity growth to reduce unemployment to some targeted level and reach a GDP growth rate of 5 percent per year. All scenarios embed the objective of reducing the unemployment rate by half in ten years, from 10.8 percent in 2004 to 5.4 percent in 2014. Assuming an annual growth rate of the labor force of 2.5 percent over the period, consistent with the trend since the 1990s, employment will have to grow at an annual rate of 3.1 percent in order for Morocco to reach the targeted unemployment level in 2014. The resulting growth rate of real wages would be 1.9 percent per annum. The Moroccan authorities envisage a minimum growth rate of output of 5 percent a year. Three scenarios for achieving this output growth rate are examined (Table 5).

Table 5.

Employment Scenarios, 2004–14

article image
Source: IMF Staff estimates.

33. The first scenario assumes that capital stock would grow at 4 percent per annum, in line with the average growth rate since 1990. Under this scenario total factor productivity will have to grow at a minimum rate of 1.5 percent a year in order to reach the output growth target, compared with a total factor productivity contribution of 0.8 percent to the nonagricultural GDP growth during 1998–2004.

34. The second scenario assumes that capital stock grows in line with GDP at the rate of 5 percent, thus maintaining the investment rate stable at the current level of 25 percent of GDP throughout the ten-year period. In order to reach the output growth target under this scenario, total factor productivity will have to grow at a minimum rate of 1.1 percent a year.

35. The third scenario assumes that productivity grows at 0.8 percent per year, the average growth rate of productivity in the nonagricultural sector for the period 1998–2004. In order to achieve the desired output growth rate, the capital stock will have to grow at the rate of 5.8 percent per year, raising the investment rate from 25 percent in 2004 to 27 percent in 2014.

36. While these scenarios are indicative, the findings highlight that reforms in the future would need to aim at fostering both productivity gains and higher investment as the basis for higher growth and employment.

D. The Role of Macroeconomic Stability and Structural Reforms as Determinants of Morocco’s Growth Performance

37. This section applies the cross country regression model presented in Loayza and others (2005) to study the growth performance of Morocco.17 The evolution of the variables included in the model explains a significant portion of the observed increase in per capita GDP growth of 1.49 percent per annum on average between the cycles 1991–98 and 1998–2004.18 In particular, this acceleration is associated with improvements in macroeconomic stability and structural reforms.

Cross-country regressions

39. Loayza and others’ (2005) econometric specification is fairly parsimonious and can be represented by the following equation:

38. Loayza and others (2005)use cross-country regressions in the spirit of Barro and Lee (1994), which link the per capita GDP growth performance of a country to a group of economic, social, and political variables, in order to construct growth explanations. They identify a robust statistical relationship between a group of variables and per capita GDP growth in a sample of 78 countries19 during the period 1961–99. They classify the growth determinants in five subgroups of variables associated with transitional convergence, cyclical reversion, structural policies and institutions, stabilization policies, and external conditions. Using the estimated coefficients of these growth determinants, it is possible to decompose the growth performance of a particular country.

yit−yit−55=λyit−5+αCyit−5C+β′XitI++γ′XitS+δTit+μt+ηi+εit(4)

where y is log per capita GDP, yC is output gap (the logarithmic difference between actual and trend output), XI is a set of structural policies and institutions, XS is a set of stabilization policies, T is terms-of-trade growth, μt is a time dummy, μi is an unobserved country fixed effect, εi is the regression residual, and λ, αC, β, γ, δ are parameters to be estimated. The subscript i refers to a country and t to time period. Details on the definition, construction, and sources of these variables and the econometric technique used are presented in Appendix II.20

40. The set of explanatory variables in the regression equation (4) can be classified as:

  • Transitional convergence: One of the main implications of the neoclassical growth model is that an economy’s growth rate depends on its original position and that richer countries tend to grow more slowly than poorer countries, after controlling for other determinants of growth (Barro and Sala-i-Martin (1995) and Turnovsky (2002)). This effect is captured by the parameter λ, which should be negative.

  • Cyclical reversion: Countries growing faster than their trend tend to slow down in the future. This effect is captured in equation (4) by the parameter αC, which should be negative.

  • Structural policies and institutions: A vast literature supports the claim that structural policies should have an impact on per capita growth. For example, Lucas (1988) emphasizes the importance of human capital, Beck and others (2000) that of financial development, and Pritchett (1996) that of international trade. Regarding the role of government expenditures, Corden (1991) analyzes the impact of their size, while Barro and Sala-i-Martin (1992) examine the effect of public services and infrastructure. Finally, Kauffman, Kraay and Lobaton (1999) explore the relationship between several governance indicators and growth. Therefore, the set of policies included in the regression contains: education (which should have a positive effect on growth), financial depth (positive effect), international trade openness (positive effect), governance (positive effect), government burden (negative effect), and public services and infrastructure (positive effect). These effects are captured in equation (4) by the β vector.

  • Stabilization policies: Stabilization policies are also expected to have a positive impact on growth as argued by Fisher (1993). The following variables are therefore included in the regression: lack of price stability (negative effect), cyclical volatility (negative effect), real effective exchange rate level21 (negative effect), and financial crises (negative effect). These effects are captured in equation (4) by the γ vector.

  • External conditions: The two included variables are related to global conditions and the evolution of terms of trade. Their effects are captured in equation (4) by the coefficients μt and δ.

41. The results of the estimation of the cross country regression are presented in Table 1422 in Appendix II. They are obtained using econometric techniques that deal with the problems of both a country-specific fixed-effect in a dynamic structure and the potential endogeneity of the explanatory variables.23 All the estimated parameters have the correct sign and are statistically significant, except for the case of governance, which is negative and insignificant.24 Therefore, the coefficient for this variable should be interpreted with caution, as these results mean that the effect of governance on economic growth might rather work through the policies that the Government implements.25 Results also reflect the deterioration of world growth conditions which has taken place since the end of the 1970s, as reflected by the increasingly negative coefficients of the time dummies.

42. When interpreting the results of the regression, one should bear in mind that cross-country regressions have one main shortcoming. By lumping countries at different stages of development, they ignore the possibility that the impact of policies could differ in various countries depending on their level of development. For example, because returns to production factors are potentially higher in poor countries than in richer ones, the effects on growth of a policy that increases the size of a country’s educated labor force could be much higher in a poor country than in a richer one. The same argument applies, for example, to a policy that favors the building of infrastructure.

Growth determinants for Morocco

43. Overall, Morocco has made progress in most policy dimensions. A set of charts in Figure 13 summarizes the evolution of the right-hand-side variables of equation (4) for Morocco and the world median during 1971–2004.26 These graphs present an interesting picture of changes in the Moroccan economy over the last thirty-five years:

Figure 13.
Figure 13.
Figure 13.

Growth Determinants, 1971–2004

Citation: IMF Staff Country Reports 2005, 419; 10.5089/9781451824773.002.A001

  • Transitional convergence: The graph shows a slow increase in GDP per capita, which reflects a lackluster growth performance.

  • Cyclical reversion: The graph shows that the two most recent quinquennial periods have benefited from the cyclical recovery associated with large and negative output gaps in 1995 and 2000.27

  • Structural policies and institutions: Both the absolute and relative evolution of these variables are of interest:

    • ➢ The evolution of secondary enrollment and public infrastructure clearly shows an absolute improvement since 1971, but a worsening in relative terms. Putting it differently, Morocco is improving both indicators, but lagging behind the World in both areas.28

    • ➢ Other variables such as trade openness and financial depth show a significant improvement since 1991 and put Morocco above the median value for these variables. The case of trade openness is worth noting because, despite the country’s improvements in this area, the structure-adjusted trade volume29 indicates that in relative terms, Morocco’s position has worsened in the period analyzed (trade openness was around 15 percent above the world median in the 1970s and it is roughly on the world median during 2001–04), but the 2000s show a takeoff of openness that may further benefit from recent trade reforms and trade agreements.

    • ➢ The evolution of the government burden shows no large differences with the world median. However, it worsens with the deterioration of the fiscal situation in the last period.

  • Stabilization policies: Price stability shows a small positive difference with the world median and a consolidation of the inflation rate around an average of two percent since the mid-1990s. In Morocco, cyclical volatility is associated with weather conditions and drops below the world median in the last period (due to improved weather conditions). In addition, the real exchange rate has been on a declining path. Finally, Morocco suffered from a financial crisis in the early 1980s.30

  • External conditions: The data show a slight improvement of the terms of trade since the mid-1980s followed by a significant deterioration in the last period.

44. The empirical analysis suggests that the above-described variables could shed light on the Moroccan growth performance. The estimated coefficients and the actual values of the variables are used to account for the changes in the growth performance across different periods. From the estimated regression equation (4), it is possible to compute the expected change in Morocco’s annual growth rate between two periods (see Appendix II for details on the estimation procedure).31 Table 6 summarizes the contributions of each category of the explanatory variables to the changes in per capita growth rates during the four periods analyzed in this chapter.32

Table 6.

Changes in Annual Average Per Capita Growth Between Periods

article image
Source: Gallego (2005).

45. A significant share of the acceleration of the annual average per capita GDP growth rate during 1998 – 2004 can be explained by the evolution of the variables included in the model (Table 6 and Figure 14). The model projected an increase of 1.16 percent during 1998–2004.33 This projected increase is explained by improvement in structural policies (0.39 percent), macroeconomic stability (0.79 percent), and cyclical reversion (0.45 percent), in a period with a negative impact of external conditions (-0.44 percent) and with a small adverse transitional convergence effect (-0.03 percent).

Figure 14.
Figure 14.

Explaining Changes in Per-Capita Growth 1998–2004 vs. 1991–98

Citation: IMF Staff Country Reports 2005, 419; 10.5089/9781451824773.002.A001

Source: Gallego (2005).

46. In addition, the contribution of structural and stabilization policies to per capita growth during the 1998 – 2004 period is associated with a significant decrease in cyclical volatility (Figure 15, Table 7). The positive effect of structural policies on growth mainly comes from education, trade openness, and financial depth. The total improvement in structural policies is diminished by the negative contribution of the increase in government burden (a factor related to the increased wage payroll during the latest period). The drop in cyclical volatility seems to be mainly related to improved weather conditions during the 1998–2004 cycle.

Figure 15.
Figure 15.

Changes in Per-Capita Growth Due to Structural Reforms and Stabilization Policies: 1998–2004 vs. 1991–98

Citation: IMF Staff Country Reports 2005, 419; 10.5089/9781451824773.002.A001

Source: Gallego (2005).
Table 7.

Expected Contributions to Annual Average Growth Changes Between Periods

(In percent)

article image
Source: Gallego (2005).

47. Therefore, the analysis suggests that Morocco would benefit from maintaining macroeconomic stability and further accelerating the implementation of structural reforms. Improvements in structural and stabilization policies seem to have contributed positively to the modest growth increase in the last cycle with a magnitude which is consistent with cross-country empirical estimates. According to these estimates, sustained improvements of indicators in the domains of secondary education, trade openness, financial depth and public infrastructure have a large potential to accelerate growth. These policies are essential to achieve the authorities’ growth rate objectives in the medium-term.

E. Conclusions

48. This chapter has analyzed Morocco’s growth performance over the past thirty-five years using three complementary methodologies. Several conclusions emerge from the exercise.

49. The first conclusion is that total factor productivity growth has not contributed to overall growth in Morocco neither over the past thirty-five years nor over the last cycle (1998–2004). Hence, there is room for significant productivity improvements. The finding that total factor productivity growth in the nonagricultural sector has been on an upward trend and has reached a positive level over the past cycle suggests that structural reforms may have started to have an impact.

50. The second conclusion is that reaching the authorities’ GDP growth objective of at least 5 percent in the medium term from its current level of around 3.5 percent will require a further increase in the contribution of capital and total factor productivity to growth. As a consequence, reforms in the future would need to aim at fostering both productivity gains and higher investment, including in the domain of infrastructures, as the basis for GDP and employment growth

51. The third conclusion is that recent improvements in structural and stabilization policies seem to have yielded a modest growth increase in the last cycle of a magnitude consistent with cross-country empirical estimates, and that there is a large potential for further acceleration in growth provided the rhythm of structural reforms is accelerated, in particular in the domains of secondary education, trade openness, and public infrastructure, and provided cyclical volatility continues to decline. Improvements in other policies or indicators not included in the econometric analysis, but emphasized in the World Bank’s 2004 Investment Climate Survey, such as the judiciary system, access to industrial land and the financing of small and medium enterprises would improve the investment climate and also help accelerate growth.

52. The fourth conclusion is that Moroccan exports have not fully benefited from the recent expansion of global trade. Indeed, while the real growth of Moroccan exports was stronger than that of developing countries as a group until the early 1990s, the pattern has reversed since then. The recent signature of several free trade agreements provides an opportunity to catch up in this regard.

53. The fifth conclusion is that Morocco faces the risk of enjoying total factor productivity gains as a result of structural reforms, but of witnessing at the same time a rate of employment growth that is too moderate to sharply reduce urban unemployment. This possibility calls for special attention to the evolution of the labor market following the implementation of the new labor code.

Appendix:I athematical Derivations, Econometric Application, and Data Used in Section C

Derivation of Equation (2)

In order to estimate the growth rate of total factor productivity At−At−1At−1 and the elasticity of output to capital α, the model presented in equation (1) is transformed as it is explained below.

First, it is assumed that total factor productivity grows on average at a constant rate over time, that is it depends on time t in the following way

Ln(At)=a+gt+εt(5)

In equation (5), a is a constant term, and g the average growth rate of total factor productivity At−At−1At−1. εt is an “error term” that is assumed to be stationary around zero.

Second, equation (1) is written as

YtLt=At(KtLt)α(6)

In equation (6), the term YtLt is output per unit of employment, and the term KtLt is capital per unit of employment.

Third, equation (6) is written as

Ln(YtLt)=Ln(At)+αLn(KtLt)(7)

Using equation(5) and equation(7), it follows that

Ln(YtLt)=a+gt+αLn(KtLt)+εt

which is equation (2) that we estimate.

Data Used in Section C

The dataset for the whole economy covers the period 1960–2002.34 Output over employment YtLt is proxied by the ratio of Morocco’s GDP to labor force (“GDP/Labor force”). Capital over employment KtLt is proxied by the ratio of Morocco’s capital stock to labor force (“Capital/Labor force”).35

Data on GDP and labor force is taken from the World Bank’s World Development Indicators database. Data on gross capital formation for the period 1965–2002 is also taken from that database. The data on the stock of capital in constant prices is constructed by applying the perpetual inventory method. First, the data on gross capital formation in constant prices is extended to cover the period 1951–64, using time series on capital stock from Nehru and Dhareshwar (1993) and a depreciation rate of 4 percent, which is the rate they used to construct their capital stock series. Second, we use the Nehru and Dhareshwar’s estimate of capital stock for the year 1950 to compute an estimate of capital stock in 1950. We adjust these series for the difference in the base year that is used in Nehru and Dhareshwar’s investment series and the investment series from the World Development Indicators database. The estimates of capital stock for the period 1951–2002 are constructed using the extended series on gross capital formation and a depreciation rate of 5 percent. Only the estimates of capital stock for the period 1960–2002 are used in the growth analysis.

For the growth accounting exercise in the nonagricultural sector, several data sources are used. Real nonagricultural output and investment in current prices are taken from the national accounts. Labor force data come from Annuaire Statistique du Maroc (1973) for the year 1971, a dataset provided by the Département des Etudes Politiques et Financières (DEPF) for the period 1982–2002, and a dataset provided by the Direction de la Statistique for the period 1999–2004. This last dataset takes into account revisions of the series after the 2004 population census. Data prior to 1994 (the previous census year) are assumed to be compatible with the results of the 2004 census, and data for the period 1995–98 are adjusted in light of the new series available for 1999–2004. Finally, investment deflators are taken from the World Development Indicators dataset for the period 1970–80 and from national accounts for the period 1980–2003. This deflator is assumed to be constant between 2003 and 2004.

The elasticity of nonagricultural output to capital is assumed to be the same as for the whole economy.36 The stock of capital in the nonagricultural sector is constructed by assuming that it is equal to 90 percent ofthat of the stock of the whole economy in 1969 and that investment in the nonagricultural sector during 1970–2004 is equal to the sum of the construction, public works and part of the machinery and equipment components of gross fixed capital formation.37The depreciation rate is also assumed to be the same as for the whole economy. In the absence of yearly data on employment in the nonagricultural sector, data on the employed urban population is used instead.

Econometric Application in Section C

The degree of variable (in logarithms) integration was examined using the Augmented Dickey-Fuller tests. (Table 8). “GDP/Labor force” is integrated of order one. “Capital/Labor force” is integrated of order two.

Table 8.

Stationarity Tests

article image
Source: IMF staff calculations.

The number of lags is selected using the Schwartz information criterion ( see Schwarz (1978)). The maximum number of lags used is 9.

For the regression equations without a linear trend, the critical values are from MacKinnon (1996).

The Vector Error Correction Model (VECM) that was initially estimated used 7 lags. Johansen cointegration tests suggest that there is one cointegrating vector at both 1 and 5 percent level of significance (Tables 9 and 10). Only the first lag in this estimated VECM turned out to be significant.38 So we estimated another VECM using only one lag (Table 11).

Table 9.

Cointegration Rank Test (Trace): Production Function

article image
Note: Trace test indicates 1 cointegrating equation at the 0.05 level.

denotes rejection of the hypothesis at the 0.05 level.

Table 10.

Cointegration Rank Test (Maximum Eigenvalue): Production Function

article image
Note: Max-eigenvalue test indicates 1 cointegrating equation at the 0.05 level.

denotes rejection of the hypothesis at the 0.05 level.

Table 11.

Vector Error Correction, Model with 1 Lag 1/

article image

t-statistics in brackets.

Misspecification tests for this VECM are presented in Table 12. The multivariate LM statistics shows that the residuals are not autocorrelated. The multivariate JB test rejects the hypothesis of normality of residuals. However, this rejection is due to excess kurtosis, which has less impact on properties of cointegration estimators, than if the skewness was considered a reason for the rejection.39

Table 12.

Misspecification Tests, Model with 1 Lag 1/

article image

Numbers in brackets are p-values.

A Fully-Modified OLS (FMOLS) procedure proposed by Phillips-Hansen (1990) was also used. This procedure takes into account the endogeneity of variables and the autocorrelation of residuals. Monte Carlo simulations by Hargreaves (1994) suggest that the FMOLS dominates other univariates estimators in the case of small sample simulations. Estimates from FMOLS are also presented in Table 2. Note that for all these estimators the signs and magnitudes of the elasticity of capital to output is in line with a priori expectations. The value of the Lc statistics is 0.39, which corresponds to a probability value of 0.17 (see Hansen 1992). This indicates that the coefficients of the estimated production function in equation (2) are stable. Moreover, since the Lc statistics is also a test of the null of cointegration, the value of the statistics suggest that the estimated relation in equation (2) is a long-term one.

Appendix:2 Econometric Procedure and Data Used in Section D

The estimation of equation (4) poses a basic econometric problem that make traditional techniques inaccurate. This problem relates to the presence of both a country-specific fixed effect and a dynamic structure (captured by the inclusion of the lagged income level) and the potential endogeneity of the right-hand side variables. To deal with these issues, Loayza and others (2005) use a two step Generalized Method of Moments (GMM) estimator, derived from the Arellano and Bond (1991) estimation procedure. This estimator takes differences of regression variables or instruments to control for unobserved effects and uses previous observations of explanatory and lagged dependent variables as instruments (which are called internal instruments). The GMM estimator combines equation (4) in levels and the first difference of equation (4) into one system. In order to obtain consistent estimates, Loayza and others (2005) use lagged levels and differences of the right-hand side variables, as far as these are valid instruments (e.g. using a Sargan test). The GMM estimation uses a panel structure consisting of non-overlapping quinquennial data for 1960–99. In order to take into account important data points that have a role in the correction of the endogeneity problem, time dummies (capturing period shifts) are also included in the estimation, with respect to a benchmark period (1966–70).

The description and source of the data used in Section D are presented in Table 13, but the construction of the measure of trade openness requires further explanation. This measure is the ratio of the volume of trade (real exports plus imports) over GDP, adjusted for the size (area and population) of the country, per capita GDP, whether it is landlocked, and whether it is an oil exporter. This structure-adjusted trade volume is preferred to the unadjusted measure because some of Loayza and others’ econometric estimates and projections are based on cross-country comparisons. Without the adjustment, they argue that the model would be unfairly attributing to trade policy what is merely the result of structural country characteristics. Indeed small countries are more dependent on international trade than large ones, oil exporters can have quite large volumes of trade and at the same time impose high import tariffs, and landlocked countries face large transport and trading costs and thus trade less than countries with port access. Thus, in the particular case of trade openness, the econometric technique requires the use of an auxiliary variable. It is the residual from the regression of the trade intensity ratio to population, area, level of per capita GDP, transport costs and an oil dummy as in Pritchett (1996).

Table 13.

Definitions and Sources of Variables Used in Section D

article image
Source: Gallego (2005).
Table 14.

Determinants of Economic Growth, Basic Regression in Loayza and Others (2005): GMM-IV System Estimator

article image
Numbers in parentheses represent the absolute value oft-statistics.Source: Loayza and others (2005).

Finally, the decomposition of growth accelerations presented in Table 6 requires to solve the problem of fitting regressors corresponding to periods of varying length using the econometric model described by equation (4) that has parameters estimated using five-year lagged variables. The technique to solve this problem consists in the application of a general recursive equation, as in equation (8):

E[gt,t+c−gt−d,t]=λ^A(yit−yit−d)5d+α^A(yitC−yit−dC)5d+β^′(Xit+cI−XitI)+γ^′(Xit+cS−XitS)+δ^(tit+c−tit)+(μ^t+c−μ^t)(8)

where gt,t+c is the average growth rate between t and t+c; and t+c and t-d are the upper and lower bounds of the periods under estimation. This form makes it possible to apply the coefficients from Loayza and others (2005) to the periods of Table 6. Two adjustments are made. First it is necessary to correct the lagged term related to transitional convergence and cyclical recovery that are originally defined for five-year periods. The lambda and alpha coefficients are defined for variables under five-year lags, but in the periods of Table 6 there are no five-year lagged terms anymore, but terms with d lags. Therefore, a transformation of the lagged terms on a d-year period change into a five-year equivalent change is needed. To do this, the lagged term is divided by d—to obtain a sort of average change per year—and multiplied by 5 so the annual average is transformed into a five-year change. The second problem to be solved is that the lambda and alpha coefficients are constant terms that distribute the lagged differences into a five-year period, as the original coefficients refer to five-year periods. However, in this general case, the next period is not a five-year, but a d period. As a result, the lambda and alpha coefficients require a simple transformation of five-year annual ‘rates’ into a c-year annual rate (using an exponential form), such that they also are consistent with the five-year lagged terms. In algebraic terms, this transformation can be written as follows:

λ^A=[(1+λ^)]c5−1andα^A=[(1+α^C)]c5−1

Reference

  • Arellano, Manuel and Stephen Bond, 1991, “Some Tests of Specification for Panel data: Monte Carlo Evidence and an Application to Employment Equations,” Review of Economic Studies 58 (2) 277–97.

    • Search Google Scholar
    • Export Citation

  • Barro, Robert J. and J.W. Lee, 1994, “Sources of Economic Growth,” Carnegie-Rochester Series on Public Policy, 40: 1–57.

  • Barro, Robert J. and Xavier Sala-i-Martin, 1992, “Public Finance in Models of Economic Growth,” Review of Economic Studies, Vol.59 (4): 641–61.

    • Search Google Scholar
    • Export Citation

  • Barro, Robert J. and Xavier Sala-i-Martin, 1995, Economic Growth (New York: McGraw-Hill).

  • Baxter, Marianne and Robert. G. King, 1999, “Measuring Business Cycles: Approximate Band-Pass Filters for Economic Time Series,” Review of Economics and Statistics 81 (4): 575–93.

    • Search Google Scholar
    • Export Citation

  • Beck, Thorsten, Asli Demirgüç-Kunt, and Ross Levine, 2000. “A New Database on Financial Development and Structure,” World Bank Economic Review 14 (3): 597–605.

    • Search Google Scholar
    • Export Citation

  • Calderón, César and Luis Serven, 2004, “The Effects of infrastructure Development and Growth and income Distribution,” Policy Research Working Paper, World Bank, No. WPS 3400–04.

    • Search Google Scholar
    • Export Citation

  • Canning, David. 1998. “A Database of World Stocks of Infrastructure, 1950–95,” World Bank Economic Review 12 (3) 529–47.

  • Caprio, Gerard, and Daniela Klingebiel, 2004, “Episodes of Systemic and Borderline Financial Crises,” Manuscript, World Bank.

  • Corden, Max, 1991, “Macroeconomic Policy and Growth: Some Lessons of Experience,” Proceedings of the Annual Conference on Development Economics 1990, 59–84, (Washington, DC: The World Bank).

    • Search Google Scholar
    • Export Citation

  • Dollar, David, 1992, “Outward oriented Developing Economies Really Do Grow More Rapidly: Evidence from 95 LDCs, 1976–1985,” Economic Development and Cultural Change 40 (3) 523–44.

    • Search Google Scholar
    • Export Citation

  • Dollar David and Art Kraay (2003). “Institutions, Trade and Growth,” Journal of Monetary Economics 50 (1) 133–62.

  • Easterly, William, and Mirvat Sewadeh, 2002, “Global Development Network Growth Database” (Washington, DC: World Bank).

  • Fisher, Robert, 1993, “The Role of Macroeconomic Factors in Growth,” Journal of Monetary Economics 32 (3) 485–511.

  • Gallego, Francisco, 2005, “Economic Growth in Morocco: Explanations and Forecast Using Cross-Country Regressions,” Background Paper for the World Bank Morocco: Country Economic Memorandum.

    • Search Google Scholar
    • Export Citation

  • ITU (International Telecommunications Union), 2004, World Telecommunication Indicators Database, 8th ed. Geneva.

  • Hargreaves, C., 1994, “A Review of Methods of Estimating Cointegrating Relationships” in Nonstationary Time Series Analysis and Cointegration, ed. Hargreaves, 99–131 (Oxford: Oxford University Press).

    • Search Google Scholar
    • Export Citation

  • Kaminsky, Graciela L., and Carmen M. Reinhart, 1998, “Financial Crisis in Asia and Latin America: Then and Now,” American Economic Review 88 (2): 444–48.

    • Search Google Scholar
    • Export Citation

  • Kauffman, Daniel, Art Kraay and Pablo Zoido Lobaton, 1999, “Governance Matters,” PRWP 2196 (Washington, DC: The World Bank).

  • Loayza, Norman, Pablo Fajnzylber and Cesar Calderon, 2005, “Economic Growth in Latin America and the Caribbean: Stylized Facts, Explanations and Forecasts” (Washington, DC: The World Bank).

    • Search Google Scholar
    • Export Citation

  • Lucas, Robert, 1988, “On the Mechanics of Economic Development,” Journal of Monetary Economics 22 (1) 3–42.

  • Miller, Stephen and Mukti Upadhyay, 2000, “The Effects of Openness, Trade Orientation and Human Capital in Total Factor Productivity,” Journal of Development Economics 63(2): 399–423.

    • Search Google Scholar
    • Export Citation

  • Nehru Vikram, and Ashok Dhareshwar, 1993, “A New Database on Physical Capital Stock: Sources, Methodology, and Results,” Revista de Análisis Económico, Vol. 8, No 1 (June), pp. 37–59.

    • Search Google Scholar
    • Export Citation

  • Nsouli, Saleh M., Sena Eken, Klaus Enders, Van-Can Thai, Jorg Decressin and Filippo Cartiglia, 1995, “Resilience and Growth Through Sustained Adjustment,” IMF Occasional Paper No 117.

    • Search Google Scholar
    • Export Citation

  • Paruolo, P., 1997, “Asymptotic Inference on the Moving Average Impact Matrix in Cointegrated I(1) VAR Systems,” Econometric Theory, Vol. 13, pp.79–118.

    • Search Google Scholar
    • Export Citation

  • PRS Group, 2004. International Country Risk Guide. East Syracuse, NY. Available at www.prsgroup.com/icrg/icrg.html.

1

Principal authors are Jacques Bouhga-Hagbe, Jérôme Vandenbussche (both IMF), José R. López-Cálix (World Bank) and Francisco Gallego (Massachusetts Institute of Technology). Research assistance was provided by Fernanda Sayavedra. The Moroccan authorities kindly provided data for and helpful comments on earlier versions of this chapter.

2

By convention, annual GDP growth over a given period of time represents the geometric average growth rate between the GDP level at the beginning of the period and that at the end of the period.

3

The set of developing economies is taken from the World Economic Outlook database and comprises 152 countries for the whole period 1971–2004. Averages are computed using PPP weights.

4

These four periods are chosen to ensure consistency across all sections in the chapter and to provide a basis for a meaningful analysis of Morocco’s growth performance over time. The first period (1971–82) runs between two population censuses. The other periods run from peak to peak, where the noncensus peak years (1991, 1998 and 2004) are obtained after plotting the path of the log of real GDP between 1970 and 2004 and fitting a polynomial trend line of order three.

5

The growth rate of the population has steadily declined from 2.6 percent in the early 1970s to 1.1 percent in 2004.

6

The volatility of real GDP growth in a decade is defined as the standard deviation of annual growth rates observed during that decade.

7

As a result of significant public wage increases the share of government in nominal GDP has expanded over the last period, as shown in Figure 4.

8

In this paragraph and the next, investment is defined as gross fixed capital formation and therefore excludes stockbuilding.

9

The nongovernment sector includes public enterprises.

10

Data available for 2001–04 show that investment by public and semi-public companies have supported the improvement over the past cycle.

11

These subperiods are the same as those analyzed in Section B.

12

The nonagricultural sector includes all sectors of the economy except agriculture, forestry and fishing which together represent 17 percent of real GDP on average over the period.

13

The details of the derivation are in Appendix I.

14

It is assumed that the error term εt is stationary. The estimation is done in levels using cointegration techniques on a dataset covering the period 1960–2002. The Vector Error Correction Model (VECM) that was initially estimated used 7 lags. Only the first lag in this estimated VECM turned out to be significant. So we estimated another VECM using only one lag.

15

Further details on the econometric techniques and tables are presented in Appendix I.

16

The growth accounting exercise uses the estimated elasticity of output to capital of 0.4. One should note that some authors do not estimate this elasticity, but rather assume a value usually between 0.3 and 0.4 to perform their growth accounting exercise. For example, Bosworth and Collins (2003) assume that the value of this elasticity is 0.35.

17

Since the analysis is based on cross country regressions, GDP is in 1985 PPP-adjusted US dollars. Hence growth rates in Section D do not exactly match those in Sections II and III for identical periods. Deviations between the two growth series, however, are minimal.

18

Growth of GDP per capita—in 1985 PPP-adjusted dollars—accelerated from an annual average of 0.26 percent during 1991–98 to an annual average of 1.75 percent during 1998–2004.

19

The sample consists of 78 countries, covers all continents and includes Morocco (see Appendix C in Loayza and others (2005)).

20

For the determinants of growth, actual values are used until 2002, 2003 or 2004, depending on the availability of the series.

21

The level of the real effective exchange rate is adjusted such that the average for 1976–85 equals the index of real exchange rate distortion reported in Dollar (1992), which is 123 for Morocco during 1976–85. Using Dollar’s methodology, a country without any distortion would have an index of 100 for 1976–85.

22

Only results for the basic equation estimated in Loayza and others (2005), which contains additional estimates and robustness checks, are presented.

23

More specifically, an Arellano and Bond (1991) Generalized Method of Moments (GMM) estimator is used as explained in Appendix II.

24

Loayza and others (2005) find that this result is robust to alternative indicators of governance. It is also consistent with Dollar and Kraay (2003) who find that various measures of governance have a relatively weak effect on growth, particularly over the medium-term horizon.

25

Caution is also required because Kauffman, Kraay and Lobaton (1999) find a positive association between governance indicators and growth performance.

26

Countries included in the empirical analyses in Loayza and others (2005) are used to compute the world median. The median, and not the mean, is used to take account of the outliers problem. The figures present the evolution of the variables at a quinquennial frequency, as in their paper.

27

This observation deserves a caveat since the Moroccan economy is typically subject to supply-side rather than demand-side shocks. Therefore the ‘output gap’ should be interpreted with that characteristic in mind.

28

Note that main telephone lines are a proxy for public infrastructure, as discussed in Loayza and others (2005). Alternative proxies for public infrastructure are energy generation capacity (for example, megawatts of electricity produced per capita), transport facilities (for example, kilometers of paved roads per capita), and main telephone lines plus mobile phone subscribers. This last point is important because including mobile phone subscribers, a significant increase is observed in the 2000s. However, most results are basically unaffected if a composite index of public infrastructure is used, as suggested by Calderón and Servén (2004).

29

As explained in Appendix II, the measure of trade openness is a structure-adjusted trade intensity index. It is estimated as the ratio of real exports and imports to GDP adjusted for some country-specific structural characteristics such as size (both area and population) of the country, per capita GDP, transportation costs (whether the country is landlocked), and resources endowment (whether it is an oil exporter). For further details, see Pritchett (1996).

30

Morocco suffered from a balance of payment crisis between 1980 and 1983, with a current account deficit that reached more than 12 percent of GDP and foreign exchange reserves covering less than a week of imports at the height of the crisis.

31

See Appendix II for details.

32

The model’s ability to predict the changes in per capita growth is mixed in some cycles, especially in the period 1991–98 when the actual growth declined rather than increased, and in the period 1971–82 where the expected decline is larger than the actual decline. In the other two cycles, the actual changes are close to the model’s predictions. Results take into account, but do not show, the effect of the time dummies.

33

For the determinants of growth, actual values are used until 2002, 2003 or 2004, depending on the available series.

34

Estimates of the elasticity of output to capital α for the period 1970–2002 do not seem plausible. Some estimates were either negative, close to one or greater than one The growth accounting exercise includes also 2003 and 2004. Given that real investment figures for these years are not yet available, estimates of capital stock for these years are constructed using the 2003 and 2004 growth rates of nominal investment. Similarly, the labor force series is extended to 2003 and 2004 using Moroccan authorities’ population projections.

35

Note that by using labor force instead of employment, it is implicitly assumed that the unemployment rate is stationary. Estimation for the shorter period for which unemployment data is available does yield estimates of the elasticity of output to capital α that are not plausible. A growth accounting exercise was done for the period 1971–2004, using the authorities’ employment figures that are available. A value of 0.4 was used for the parameter α. It was found that the contribution of the residual to growth was -0.23 percent for that period, compared to a growth rate of 0.03 percent presented in Table 3.

36

The obtained residuals are not very sensitive to that assumption, and the analysis would yield similar results with an elasticity of .3 instead of .4.

37

The nonagricultural share of the machinery and equipment component is assumed equal to the nonagricultural share of capital good imports in the balance of payments.

38

The cointegrating equation for this VECM is presented in Table 1. Other tables are available upon request.

Appendix:I A Summary of Exchange Arrangements (Position as of End-2004)

Exchange rate structure and market. The exchange rate is pegged to an undisclosed basket of currencies dominated by the Euro. BAM intervenes in the market by continuously setting buying and selling rates based on the basket of currencies of the peg. Banks may provide forward contracts for commercial and financial operations for up to 12 months.

Foreign currency accounts. Residents are allowed to open: (a) foreign currency accounts which may be credited with transfers from abroad or from a domestic bank in accordance with foreign exchange regulations. Residents exporters are allowed to open these accounts and keep up to 20 percent of their foreign exchange receipts. Others must obtain authorization. Authorized debits include purchases of foreign currency denominated government debt, purchases of dirhams, or transfers to convertible dirham accounts, and transfers abroad. (b) dirham accounts convertible into foreign currency may be open by resident foreign nationals (see below). Nonresidents (foreign nationals and Moroccans residing abroad) are allowed to open: (a) foreign currency accounts. These accounts are regulated by provisions applicable to resident foreign nationals. They can be credited with transfers from abroad or from a domestic bank with authorization. Authorized debits include transfers abroad, subscriptions in foreign currency to government paper, or transfers to convertible dirham accounts.(b) dirham accounts convertible into foreign currency. These accounts are regulated by the provisions applicable to convertible dirham accounts of residents foreign nationals. They can be credited with the proceeds of the sale of foreign currency. Authorized debits include purchase of foreign exchange, dirham payments, and transfers to term deposits in convertible dirham. Nonresident foreign nationals not entitled to maintain convertible dirham accounts may open (c) forward convertible accounts. Funds on these accounts may be transferred within maximum periods of four years in annual installments of 25 percent, and may be used without authorization to settle dirham transactions.

Capital transactions. Nonresidents may purchase shares of a participating nature and foreign direct investment inflows are free. The purchase or issuance of securities (shares, bonds, bills etc.) abroad by residents is subject to authorization. Outward direct investments are subject to authorization. Corporations may obtain financial credits from nonresidents, provided the proceeds are used to finance investment or foreign trade.

Provisions specific to financial institutions: Commercial banks may place foreign currency with correspondents abroad, purchase sovereign bonds and securities issued by international financial institutions, and borrow abroad only to finance foreign trade or investment operations. Bank loans or purchase of securities domestically in foreign currency and loans to nonresidents are subject to authorization. Convertible and dirham deposits are excluded from reserve requirements. All deposits are taken into account in the calculation of liquid asset requirements. There are no special provisions applicable to institutional investors which are not allowed to purchase assets abroad.

Appendix: II

Countries Included in the Husain (2004) Paper

article image

Appendix: III A Model of Balance of Payments Equilibrium Exchange Rate (bpeer)52

Description of the Model

In the BPEER model, the equilibrium exchange rate is defined as the rate that is consistent with BOP equilibrium. The model therefore relies on the BOP identity. The sum of the current account balance (ca), the capital account balance (ka), and a change in the foreign reserves (rez) equal to zero, that is:

cat+kat−Δrezt=0(5)

In equation ), the variables are in percent of GDP. The current account balance is defined as the sum of net trade (nt), including services, and interest payments on foreign assets (nfa):53

cat=nt1+it⊕nfat(6)

where it⊕ stands for nominal interest on net foreign assets. We have assumed that transfers are constant in percent of GDP. Net trade depends on the level of real effective exchange rate (q), internal demand (y) and external demand (y*) :

nt1=−α1(qt−α0)−α2yt+α3yt*+ε1,t(7)

Where α1, α2, α3 > 0 are respectively the elasticities of net trade with respect to the exchange rate, domestic demand and foreign demand. An appreciation (increase) of the exchange rate of 1 percent reduces net trade by α1 percent of GDP. A real appreciation is expected to make a country loose in competitiveness, and therefore reduce its exports. Similarly, an increase in internal demand by one percent reduces it by α2 percent of GDP. An increase in internal demand boost demand for all goods and services, including foreign ones. Therefore, it tends to raise imports. An increase in external demand by one percent raises net trade by α3 percent of GDP. Foreign demand is expected to boost exports. Using equations (5)–(7) it follows that

−α1(qt−α0)−α2yt+α3yt*+it⊕nfat+(kat−Δrezt)=−ε1,t(8)

Which then implies

qt=α0−α2α1yt+α3α1yt*+1α1it⊕nfat+1α1(kat−Δrezt)+εt(9)

Where

εt=ε1,tα1

is assumed to be normally distributed.

Therefore, in this BPEER model, the equilibrium level of real exchange rate is permanently influenced by internal demand, foreign demand, interest on net foreign assets and capital account flows, excluding reserve assets accumulation.

Various factors can influence capital mobility.54 In Morocco, the capital and financial account being relatively closed, flows are mostly driven by government transactions, and not interest rates differential. Holding the current account balance constant, capital and financial account flows could by proxied by net foreign assets.

Econometric Application to Morocco

The data and estimation procedure

In this section, we discuss the data used for estimating the BPEER parameters. The data are for the period from the first quarter of 1980 to the fourth quarter of 2003. The real exchange rate used is the real effective exchange rate deflated by consumer prices. The internal (y) and external (y*) demands were approximated by seasonally adjusted gross domestic product at constant prices in Morocco and in advanced economies weighted by trade exports.

The proxy for net foreign assets (nfa) were computed by taking Morocco’s net foreign asset position at end 1979, and then adding the cumulative quarterly changes in net foreign assets from Morocco’s monetary survey to get the value of the proxy at the end of subsequent quarters. As noted above, net foreign assets could also be seen as a proxy for capital and financial flows, holding the current account balance, that is foreign and domestic output constant.

The estimation procedure is conducted in four steps. First, the order of integration of the series is determined. Second, the existence of a cointegrating relation among the variables used in the BPEER model is verified. Third, the coefficients of the BPEER model are estimated.

Stationarity tests

The degree of variable integration was examined using the modified version of the Dickey-Fuller t test proposed by Elliott, Rothenberg and Stock (1996)55 REER denotes the real effective exchange rate, GDPMAR real GDP in Morocco, GDPADV real GDP in advanced economies,56 NFA net foreign assets. The symbol Δdenotes the first difference of the variables. The lag structure was chosen using the Schwartz information criteria. The maximum number of lags was set at 11. The results are presented in Table 6. All variables considered in the BPEER model seem to be integrated of order one.57

Table 6.

Stationarity Test (Elliott-Rothenberg-Stock DF-GLS Test) 1/

article image
Source: IMF staff calculations.

The dataset covers the period from the first quarter of 1980 to the fourth quarter of 2003. The test used is the one proposed by Elliot, Rothenberg and Stock (1996). The null hypothesis is that the variable has a unit root.

The number of lags is selected using the Schwartz information criterion (see Schwarz (1978)). The maximum number of lags used is 11.

For the regression equations without a linear trend, the critical values are from MacKinnon (1996). For the regression equations with a linear trend, critrical values are from Elliott-Rothenberg-Stock (1996, Table 1).

The real effective exchange rate index uses weights based on 1989–91 trade data.

The real effective exchange rate index uses weights based on 1999–2001 trade data.

Table 7.

Cointegration Trace Test: Exports, REER, and GDPADV 1/

article image

The data frequency is annual and covers the period from 1980 to 2002.

The real effective exchange rate index uses weights based on 1989–91 trade data.Trace test indicates 1 cointegrating eqn at the 0.01 level and 2 cointegrating eqn(s) at the 0.05 level* denotes rejection of the hypothesis at the 0.05 level.
Table 8.

Cointegration Eigenvalue Test: Exports, REER, and GDPADV 1/

article image

The data frequency is annual and covers the period from 1980 to 2002.

The real effective exchange rate index is uses weights based on 1989–91 trade data.Max-eigenvalue test indicates 2 cointegrating eqn(s) at the 0.01 level.denotes rejection of the hypothesis at the 0.05 level.
Table 9a.

Estimating Long-Run Exports, Based on 1989–91 Trade Weights 1/

article image

t-statistics are in brackets.

All regression equations include an unreported constant.

The data frequency is annual and covers the period from 1980 to 2002.

The real effective exchange rate index is uses weights based on 1989–91 trade data.

Table 9b.

Estimating Long-Run Exports, Based on 1999–2001 Trade Weights 1/

article image

t-statistics are in brackets.

All regression equations include an unreported constant.

The data frequency is annual and covers the period from 1980 to 2002.

The real effective exchange rate index uses weights based on 1999–2001 trade data.

Testing for cointegration

The number of cointegrating vectors was tested (see Tables 10a, 10b, 11a and 11b) using Johansen’s (1991) trace and max-eigenvalue tests. The tests indicate that there is one cointegrating vector at 1 percent level of significance. In order to calculate the test statistics, we estimated a vector error correction (VEC) model of the form:

Δxt=Γ+Πxt−1+∑i=1pAiΔxt−i+Dt(10)

where xt is the vector of endogenous variables:

xt=[qtytyt*nfat]
Table 10a.

Cointegration Trace Test: Balance of Payments Equilibrium Exchange Rate, Based on 1989–91 Trade Weights 1/

article image

The dataset covers the period from the first quarter of 1980 to the fourth quarter of 2003.

The real effective exchange rate index is uses weights based on 1989–91 trade data.

The vector autoregression includes 7 lags. A dummy is added to account for the surge in remittances and therefore net foreign assets that occurred in 2001.

Trace test indicates 1 cointegrating eqn at the 0.01 level and 2 cointegrating eqn(s) at the 0.05 level.

Table 10b.

Cointegration Trace Test: Balance of Payments Equilibrium Exchange Rate, Based on 1999–2001 Trade Weights 1/

article image

The dataset covers the period from the first quarter of 1980 to the fourth quarter of 2003.

The real effective exchange rate index is uses weights based on 1999–2001 trade data.

The vector autoregression includes 7 lags. A dummy is added to account for the surge in remittances and therefore net foreign assets that occurred in 2001.

Trace test indicates 1 cointegrating eqn at the 0.01 level and 2 cointegrating eqn(s) at the 0.05 level.

Table 11a.

Cointegration Eigenvalue Test: Balance of Payments Equilibrium Exchange Rate, Based on 1989–91 Trade Weights 1/

article image

The dataset covers the period from the first quarter of 1980 to the fourth quarter of 2003.

The real effective exchange rate index is uses weights based on 1989–91 trade data.

The vector autoregression includes 7 lags. A dummy is added to account for the surge in remittances and therefore net foreign assets that occurred in 2001.

Max-eigenvalue test indicates 2 cointegrating eqn(s) at the 0.05 level.

Table 11b.

Cointegration Eigenvalue Test: Balance of Payments Equilibrium Exchange Rate, Based on 1999–2001 Trade Weights 1/

article image

The dataset covers the period from the first quarter of 1980 to the fourth quarter of 2003.

The real effective exchange rate index is uses weights based on 19992001 trade data.

The vector autoregression includes 7 lags. A dummy is added to account for the surge in remittances and therefore net foreign assets that occurred in 2001.

Max-eigenvalue test indicates 1 cointegrating eqn at the 0.01 level and 2 cointegrating eqn(s) at the 0.05 level

The matrix Π is then decomposed into an adjustment αp×r and a cointegrating βp×r matrices, that is

Π=αβ′

where r is the number of cointegrating relationships. Dt is an exogenous dummy variable that takes the value of 1 for the first and second quarter of 2001, and zero elsewhere. It is supposed to capture the sudden surge in remittances that occurred at the beginning of 2001. Including such a dummy considerably improves the performance of the model, as this phenomenon is unlikely to be explained by this model. Note that critical values used for both the trace and max-eigenvalue in table Johansen assume no exogenous series. However, these test still suggest that there are cointegration relations among the endogenous variables even when the exogenous dummy is removed.

An estimate of the cointegrating vector is presented is presented in Table 12. The VEC was estimated using 9 lags58 (p=9) (Table 13). Lag exclusion tests are presented in Table 15, and misspecification tests in Table 16). The multivariate LM statistics show that the residuals are not autocorrelated. The multivariate JB test rejects the hypothesis of normality of residuals. However, this rejection is due to excess kurtosis, which has less impact on properties of cointegration estimators, than if the skewness was considered a reason for the rejection.59

Table 12.

Estimating the Long-Run or Equilibrium Real Effective Exchange Rate (REER) 1/

article image

t-statistics are in brackets.

All regression equations include an unreported constant.

The real effective exchange rate index uses weights based on 1989–91 trade data.

The real effective exchange rate index uses weights based on 1999–2001 trade data.

Table 13a.

Chi-Squared Test Statistics for Lag Exclusion, Based on 1989–91 Trade Weights 1/

article image

Numbers in brackets are p-values. Column headings are the dependent variable of the equation under consideration. The real effective exchange rate index is uses weights based on 1989–91 trade data.