A macroeconometric forecasting model for the Jordanian economy is proposed. The model was developed, estimated, and then used for ex-post and ex-ante forecast. Absolute Percentage Forecast Inaccuracy method has been used to measure the precision of the model forecast. The major findings of the paper are that the model (1) provides very useful information on the dynamics of the Jordanian economy; (2) appears to be particularly successful in forecasting GNP, GDP, private consumption, investment, government consumption, imports, and CPI; (3) is a moderate success in forecasting exports, money demand, and employment level; and (4) a non satisfactory forecast of other government expenditures and exchange rate.

Introduction

This paper is devoted to the building of a simultaneous linear macroeconometric model to forecast the key economic variables of the Jordanian economy. Economic variables are interdependent not only on each other, but also on the institutional and technological developments. In addition, the economic system is a set of dynamic simultaneous relationships. There are feedbacks from the dependent variables to the independent variables within this system of equations.

In addition to identifying the structural model and estimating the parameters from the reduced form, it will be necessary to consider the completeness of the model (i.e. identification of issues) by testing the hypotheses, by using the model for forecasts, and by producing a residual analysis. The ex-post and ex-ante forecast simulations of the model will provide information on the dynamics of the model and assist us in further understanding the Jordanian economy.

This model will follow Zellner's (1984) and Diebold (1998) advice to build as simple a model as possible. Even when the model intends to capture many details, the model may be large but should stay simple. The more sophisticated models may have a unique solution, or may have many solutions. Fortunately, the Jordanian econometric model is relatively simple, comprehensible, and formalized in a fairly straightforward manner.

The model developed here is concerned with the demand variables in the economy. The model resembles, to a certain degree, some aspects of the macro models built by Evans (1973) and some aspects of the DRI model (Data Resources Quarterly Model of the U.S. Economy), but it takes into consideration Jordan's economic structural characteristics. Jordan's model consists of four identities and ten behavioral equations. The model can be disaggregated into the goods markets, which includes the exchange rate function, a money market, and a labor market (employment), with the price relationships developed separately. Most structural models are over identified according to Theil (1971), and unfortunately this model is no exception. Over identification is not a problem for our purposes, but if the purpose was to estimate the structural model's coefficients from the reduced form coefficients, there would be a problem. That problem can be overcome by using 2SLS or three stage iterative estimation processes simultaneously for the reduced and structural models.

There are explicit and implicit assumptions that have been made through the construction of the structural model. The explicit assumptions are in the linear functional form, in the variables included, and in the expected sign and size of the coefficients. The implicit assumptions are that human economic behavior can be explained in mathematical form, and that all variables not in the equation are held constant. The latter implies that the mean effects of the omitted variables are captured by the constant terms, and their variabilities are captured by the variance of the error term.

The value of the interest rate on different kinds of saving and loans in Jordan was constant over the entire period and lower than the average growth rate of inflation. Thus, Jordan's interest rate cannot explain any variation in the behavior of any equation. Nevertheless, there is a difference in the prices of durable and luxury items between cash and installment values. The difference, it could be argued, suggests the existence of an implicit interest rate, but no direct data are available to test that argument. A proxy variable is developed by taking the relevant interest rate on Eurodollars in London and treating it as the Jordanian interest rate. This rate will be used to see what influence the interest rate had on the behavior of the Jordanian economy. Therefore, the interest rate in this context is the interest rate on the Eurodollar in London. Also, the interest rate on the Eurodollar is an appropriate cost of borrowing since Jordan has a relatively large foreign debt relative to the size of the economy.

The blockade of Iraq will be represented by a dummy variable, which has a value of 1.00 from 1990 until now and zero elsewhere. The coincidence of the blockade of Iraq, the influx of more than 300,000 Jordanians from Kuwait, and the depreciation of the currency at the same time may produce a challenging result to explain.

The goods market consists of five subsectors: private consumption, investment, foreign trade, government consumption expenditure, and net factor income from abroad. The Jordanian goods market data are constructed in a manner that produce four identities. These identities are defined as gross domestic product, net factor income from abroad, gross national product, and total government expenditure. The GDP identity consists of private consumption, government consumption, investment, and net exports. The net factor income from abroad’s identity consists of the following: Jordanian workers’ remittances from abroad, plus income to Jordanians from their investments abroad, minus foreign labor remittances to countries outside Jordan, minus foreign investment income from foreigners' investments in Jordan. The total government expenditure’s identity consists of the following: government consumption expenditure, government capital, and unspecified government expenditure. Also, The sum of government capital expenditures (GK_t) and unspecified government expenditures (USG_t) will be called other government expenditures. The GNP identity consists of GDP plus net factor income from abroad. The goods market's four identities can be expressed mathematically in the following way:

GDP_t = C_t + I_t + G_t + X_t -L_t

NFI_t = JLR_t + IRJ_t - FLR_t - IRF_t

GNP_t = GDP_t + NFI_t

TG_t = G_t + GK_t + USG_t

where

C_t private consumption for period t.

I_t investment for period t (gross fixed capital formation).

G_t Government consumption expenditure for period t.

X_t Export for period t.

L_t Import for period t.

NFI_t Net factor income from abroad

JLR_t Jordanian labor remittance from abroad to Jordan

IRJ_t Interest received by Jordanians from their investment abroad.

FLR_t Foreign labor remittances from Jordan to other countries.

IRF_t Interest received by foreigners from their investment in Jordan.

TG_t Total government expenditures

GK_t Government capital expenditures.

USG_t Unspecified government expenditures.

The Jordanian consumption function in the model represents a dynamic view of income, consumption patterns, and the cost of credit. It could be derived from the permanent income hypothesis and the habit persistence theoretical hypotheses of consumption. The empirical application of Friedman's (1957) permanent income hypothesis which states that current consumption is a function of current income and lagged consumption, has proven to be a very plausible hypothesis (for more detail see Timbrell (1976), and Gapiniski 1982). Friedman's consumption theory tells us that there is a positive relationship between current consumption and lagged consumption due to habit persistence, and the statistical property reveals the coefficient of the lagged consumption should be less than one if the economy is stable. There is also a positive relationship between current consumption and current income, but as income increases, consumption increases by less than the increase in income. This relationship is true unless the whole economy practices deficit spending which can not be maintained in the long run.

One expects to find a negative relationship between current consumption (C_t) and the current consumer price index (CPI_t), assuming other variables remain constant. When the CPI_t increases, it leads to a decrease in consumption due to the budget constraint, or in other words due to a decline in real income.

It is expected that the higher the interest rate is, the lower the demand will be for durable and luxury goods bought on installment, unless the growth rate of inflation and the administrative cost of lending are higher than the interest rate as is the case for Jordan. Therefore, the relationship could be described mathematically in the following manner:

C_t = f( C_t-1 , GDP_t , CPI_t , R_t )

Where:

C_t-1 private consumption for previous year.

CPI_t consumer price index for period t.

R_t nominal interest rate.

The investment function includes both the "marginal rate of return principle" and the "marginal cost of funds principle." This approach has been described by Anderson (1967), and represents a compromise between the acceleration principle and the residual fund theory. (For more details on each approach see Gapiniski 1982).

As income increases it leads to increases in the  demand for goods and services and a reduction in inventory. If the increase in the demand persists for many months, it will lead to increases in investment.

Since the 1973 increase in oil prices, the demand for Jordanian labor in the Arab oil countries produced an influx  of money to Jordan. These remittances played a major role in stimulating the construction boom during the 1970's and early 1980's. The remittances also  contributed  to  the  capital formation used for investment in different industrial projects as well as the stabilization of Jordanian economy in the 1990's.

Jordan's government is taking a leading role in investing in major agricultural and industrial projects, especially since 1973. The economic and social development plans contributed hundreds of million of Jordanian Dinar (JD) to the implementation of these projects. Thus, investment in Jordan is partially determined by the level of government expenditure.

It is necessary in building econometric models to include the lagged values of capital stock in the investment function (Gapiniski, Skegro, and Zuehlke 1989). The value of capital stock at period t-1 (K_t-1) has influential effects on the investment in period t. There is a negative relationship between K_t-1 and investment (I_t) because if the capital stock is relatively large, then its marginal productivity is low or there is a low rate of capital utilization either of which reduces the incentive to invest. If the capital stock is relatively small, however, and fully utilized, then it is expected that the existing firms in the market are making a profit and want to expand their operation to take advantage of the profit opportunity.

The interest rate theoretically is expected to have a negative impact on investment, not only because it makes the opportunity cost higher for any given firm, but also because it makes the cost of borrowing for expanding their existing size more costly and consequently reduces or eliminates profit margin. It also forces some of the firms operating at the profit margin to shut down.

Jordan’s economy depends on foreign countries to supply it with machinery and equipments to carry forward its development. Therefore the higher the unit price index of import (UPIM) the lower the investment would be materialized .

The blockade of Iraq and the influx of more than 300,000 Jordanians from Kuwait brought a substantial increase in investment to Jordan.

Thus, the relationship could be described mathematically as:

I_t = f( REM_t , K_t-1 , R_t , TG_t , GDP_t , UPIM_t-1, IRBLk_t)

where:

K_t-1 capital stock at the end of period t-1.

REM_t remittance from Jordanian working abroad for period t.

UPIM_t-1 unit price index of imports for period t-1.

IRBLK_t blockade of Iraq for period t.

Government Expenditure (G_t)

All forms of government expenditures are expected to be influenced by the level of GDP, foreign aid, imports, exports, blockade of Iraq, and the interest rate on the debt. During the 1967-1996 period, the government's expenditures have increased every year except 1970 and 1994, reflecting higher wages, an increase in the number of government employees, a higher level of capital expenditures, and the spending for building the infrastructure. 

As GDP increases one expects government expenditure to increase because government income from direct and indirect taxes would increase. As the industrial and mining sector expanded, the government's income would also increase from the return on their shares in these industries.

The higher the foreign aid to Jordan, the higher the total government spending would be, where the relationship between them is positive but less than 1.00 because not all foreign aid goes to the government.

A positive relationship is expected between the level of imports and the level of indirect taxes, while exports in Jordan would have a positive effect on government spending since the Jordanian government has a large share in major export industries such as phosphate, fertilizer, and potash.

The higher the level of the interest rate, the higher the government expenditure on servicing the debt accompanied by less government spending on capital and unspecified expenditures.

The blockade of Iraq and the influx of more than 300,000 Jordanians to Jordan from Kuwait has a positive effect on government spending to accommodate their needs.

Therefore, the relationship could be described mathematically in the following manner:

GK_t + USG_t = f( GDP_t , F_t , L_t , X_t , IRBlk_t )

G_t = f( GDP_t , F_t , L_t , X_t ,IRBLK_t )

where:

F_t foreign aids to Jordan for period t.

When the exchange rate of the Jordanian currency in terms of the dollar (E_t-1)  decreases, it indicates a depreciation of the Jordanian currency, making Jordanian goods cheaper to foreigners  which leads to increase in Jordanian exports of goods and services. The lag in the exchange rate reflects the actual conditions in the world trade, where firms sign contracts ahead of delivery.

The higher the CPI_t in Jordan, the more expensive Jordanian products become to Jordanians and to foreigners.  This condition reduces demand for Jordanian products and reduces Jordanian exports to foreign markets. Therefore, the relationship could be described mathematically in the following manner:

X_t = f(E_t-1 , CPI_t )

where:

E_t-1 exchange rate of the dollar in terms of the Jordanian currency at time t-1 (the dollar was used as a proxy for foreign currency).

When there is inflation in Jordan, assuming other  economic factors in the world remain constant, it causes increased demand for the importation of foreign goods as they become cheaper for Jordanian citizens. It is expected that when income increases, people consume more, with a wider variety of goods and services, which leads to an increase in imports. 

As the exchange rate goes down, the price of foreign currencies becomes more expensive than before. This expense is expected to reduce the demand level of imported goods. Jordan's investment is expected to have a positive effect on the level of import, because Jordan depends on foreign countries to supply it with machinery and equipment (all kinds of capital goods) to carry forward its development.

The more remittance to Jordan, the more Jordan is able to finance more imports; therefore, imports have a direct and substantial impact on the standard of living in Jordan.

The relationship could be described mathematically as:

L_t = f( CPI_t , GDP_t , E_t , I_t , REM_t )

A higher value of Jordanian currency in terms of the dollar results when Jordan receives more foreign aid, or a higher remittance, or an increase in net exports. The higher the level of imports for Jordan, however, the more the Jordanian currency will depreciate in terms of dollars. Therefore, it is expected that the remittances (REM_t), foreign aid (F_t), and exports (X_t) have a positive effect on the exchange rate, and imports (L_t) have a negative effect on the exchange rate. But the higher the interest rate, the less demand for Jordanian currency and the lower the exchange rate of the Jordanian currency in terms of the dollar. It is expected that a lagged exchange rate will have a positive relationship with the current exchange rate because the Jordanian economy is small and has few natural resources. Therefore, the relationship could be described mathematically as:

E_t = f( REM_t , F_t , X_t , L_t , R_t , E_t-1 )

Net factor income from abroad is composed of Jordanian worker's remittances from abroad, plus income to Jordanians from their investments abroad, minus foreign labor remittances to countries outside Jordan, minus foreign investment income from foreigners' investments in Jordan. Since these are impossible to quantify, the income will be treated as exogenous.

The value of interest received by Jordanians from other countries is not known by investment categories. These categories include interest received for their cash in the foreign banks and the profit from their investment in real estate or any other kinds of investments. Therefor, NFIF will be treated as an exogenous variable.

The use of the Keynesian approach represented by Keynes (1936) and expanded upon by Tobin (1956) and Baumol (1952) produces the functional relationship of the demand for money as a function of GDP, CPI, and the interest rate. The Keynesians based their model on transactions, speculation, and precautionary demands for money. Tobin used only nonhuman wealth while Friedman emphasized a wider concept of wealth. (For more details on the differences between the two approaches see Friedman (1956), Tobin (1956), Baumol (1952), and Saunders and Taylor (1976) ).

It is reasonable to argue that the higher the nominal GDP, the higher the demand will be for money.  There is an impressive amount of statistical evidence to support this assertion.

Income is not the only factor affecting the demand for money.  The interest rate is a relevant factor affecting the demand for money because money is one form in which individuals can hold their  wealth.  Holders of cash give up the opportunity to  hold other  assets that pay interest.  When the interest rate is  high, people will economize on their holdings of money, and the  demand to hold money declines as the interest rate rises.

When the CPI rises, the money balances of individuals   and businesses become smaller in  real terms, which increases  the demand for money in order to compensate for the loss on their holdings in real terms. This relationship could be described in the following mathematical forms:

M_s = M_d this is an identity which represents an equilibrium in the money market.

M_d = f( GDP_t , CPI_t , R_t )

where:

M_s money supply which is determined by the central bank of Jordan (exogenous).

M_d money demand at period t.

It is generally recognized that there is a positive relationship between the money supply and CPI if the growth rate of the money supply is higher than the real growth rate of the economy (see Friedman 1982). The current money supply, as well as the lagged change in money supply, has a positive effect on the level of the current CPI.

The increase in demand of goods and services was phenomenal after 1973 when Jordanian worker remittances started to increase at an unprecedented rate. Much of the demand was reflected in the construction industry which created a shortage in building materials. Wages increased substantially due to shortages of labor, especially construction skilled labor and professionals such as teachers and technicians. The demand for Jordanian labor from Arab Oil Countries declined with the decline of the price of oil.

Jordan relies heavily on imports to meet the domestic demand for goods and services. Imports are a huge percentage of GDP; the percentage varied between 44.04% and 97.16% during the 1967-1996 period. This high percentage was the result of huge labor remittances from Jordanians abroad sending funds to Jordan. Thus, the unit price index of import (UPIL_t-1) is an important, if not the most important, determinant of the CPI_t level in the country.

The decline of the value of Jordanian currency in terms of the dollar has a significant impact on the CPI due to the Jordanian reliance on imports.

The relationship could be described mathematically in the following manner:

CPI_t = f(M_t , ChngM_t-1 , UPIL_t-1 , E_t)

where:

ChngM_t-1 represents change in M1 money supply for previous year.

The employment function is constructed from the Cobb-Douglas production function. Where

GDP_t = AL^aK^bU

where

L represents labor employment.

K represents capital stock in use.

U represents an error term.

This production function can be expressed in an implicit form such that

L_d = f( GDP_t , K_t-1 )

where

L_d represents labor employment (demand).

The above formalization of labor demand is the same as that of Evans (1973) without the time factor that represents technological progress. The production function that Evans used was Ae^ctL^aK^bU. The time factor was ignored because it has no significant impact on the Jordanian economy. It is expected that these variables will have a positive influence on labor demand. One would expect that a higher level of available capital stock for use would have a positive impact on the level of employment.

Jordan, like other countries in the world, keeps no public data about the value of the capital stock. For the Jordanian data, the following approach has been used to produce the value of the capital stock. The following approach is more theoretically sound than the typical approach (see Evans 1973) for calculating the value of initial capital stock.

Let

d_t represent total depreciation at period t where there is a published data available.

K_t-1 represent capital stock at the end of period t-1.

r_t represent the depreciation rate at period t.

r_t+1 represent the depreciation rate at period t+1.

Assuming that the depreciation rate in 1964 and 1964 is the same, and varying after that. Such that

r_t =r_t+1 = r

then

r K_t-1 = d_t (1)

r K_t = d_t+1 (2)

but

K_t-1 ( 1-r ) + I_t = K_t is an identity.

r ( K_t-1 - rK_t-1 + I_t ) = rK_t (3)

substitute equation (1), and (2) into equation (3)

d_t - rd_t + rI_t = d_t+1 (4)

r = ( d_t+1 - d_t ) / ( I_t - d_t ) (5)

From equation (5) it is possible to calculate the depreciation rate of each year for the 1964-1995 period because the values of the variables on the right hand side of equation (5) are available. The value of (r) is expected to vary over time due to different accounting procedures and the capacity utilization of the existing stock. After calculating r_t for each year for the entire period of 1964-1995, the average rate of depreciation for the whole period is 4.4%. This average depreciation rate applied to equation (1) to calculate the capital stock at the end of 1964, is equal to JD 163.36 million. Then use the following identity:

K_t = K_t-1 + I_t - d_t

to calculate K_t for the 1964-1995 period.

The data for depreciation increased as capital stock increased when the economy worked at full capacity. When there was a decline in economic activity as in 1985, or a short period of shut down during the civil disturbances in 1970, the depreciation value decreased.

The total number of workers employed in Jordan during 1975 and during the 1979-1996 period has been taken from various publications of the Jordan Ministry of Labor and Social Development (1978 - 1987) and Central Bank of Jordan (1997). During 1967-1974 the unemployment rate was 10% of the usual 20% of the total participation rate of the population. Total employment was calculated from the population figures. For 1976, there was a 20% participation rate of the population without unemployment. The 1977 and 1978 participation rates were respectively 20.9% and 20.8% of total population. The rest of the employment data has been taken from Central Bank of Jordan (1997).

The structural model describes the economy and the behavior of the economic agents. This structural model will be integrated as a whole system, which produces the reduced form model. The reduced form is used for forecasting purposes. The reduced form expresses each endogenous variable in terms of exogenous and predetermined variables and provides a measure of the effect of change of each independent variable on the endogenous variables.

There are twelve reduced form equations which are derived from fourteen structural equations and identities. There are twelve of these equations because the fourteenth, money supply, is exogenous and GNP is an identity. The reduced form equations are designed to estimate the main variables in the Jordanian economic model. These variables are gross domestic product (GDP_t), private consumption (C_t), total investment (I_t), government consumption (G_t), other government expenditure (government capital and unspecified expenditure (GK_t + USG_t)), total government expenditure (TG_t), exports (X_t), imports (L_t), exchange rate (E_t), consumer price index (CPI_t), money demand (M_d), and employment level (L_d). All these variables are a function of the same exogenous and predetermined variables which are a function of lagged private consumption (C_t-1), Jordanian remittance (REM_t), foreign aid (F_t), lagged capital stock (K_t-1), lagged changed in money supply (ChngM_t-1), lagged unit price index of imports (UPIL_t-1), a dummy variable for the blockade of Iraq (IRBLK_t), a lagged exchange rate (E_t-1), and the interest on the Eurodollar (R_t).

The OLS method of estimation was applied to each equation of the reduced form of the econometric model. This method of estimation has been used rather than the simultaneous SUR method because one method is not more advantageous than the other when the right hand side of all equations in the system are the same according to Judge et al. (1985) and Greene (1988). All equations have plausible signs.

GDP_t C_t I_t G_t

Constant -571.894 -121.807 -409.949 -76.464

(-2.158) (-.325) (-2.066) (-4.513)

C_t-1 .33 .465 .08667 .07932

(3.355) (3.337) (1.177) (2.561)

REM_t 1.355 .644 1.547 .111

(8.038) (2.696) (12.258) (2.095)

F_t .321 -.246 .05103 .119

(.791) (-.428) (.168) (.935)

K_t-1 .291 .09109 -.03748 .08982

(7.449) (1.648) (-1.283) (7.313)

ChngM_t-1 .461 .741 .735 -.04694

(1.59) (1.805) (3.385) (-.514)

UPIL_t-1 .867 4.694 -2.282 1.173

(.697) (2.662) (-2.447) (2.993)

IRBLK_t 510.33 -49.236 699.29 104.715

(4.087) (-.278) (7.478) (2.664)

E_t-1 203.543 -8.529 179.176 115.228

(2.189) (-.065) (2.573) (3.936)

R_t .281 7.584 -2.883 4.951

(.043) (.812) (-.584) (2.384)

DW 1.893 2.836 1.445 2.091

Adjusted R² .998 .992 .991 .996

F(9, 18) 1656.111 376.422 350.087 860.984

SSR 76404.43 153283.432 42844.436 7573.449

X_t L_t E_t TG GK_t+USG_t

Constant 7.33 -328.995 -.773 -522.83 -146.366

(.032) (-1.230) (-1.376) (-2.235) (-.641)

C_t-1 -.205 .0964 .0003957 .336 .256

(-2.428) (.971) (1.896) (3.864) (3.022)

REM_t .627 1.575 -.000413 .104 -.007137

(4.344) (9.253) (-1.155) (.699) (-.049)

F_t 1.123 .727 -.001448 .686 .567

(3.236) (1.778) (-1.684) (1.917) (1.622)

K_t-1 .256 .108 -.0001509 .09058 .0007554

(7.65) (2.754) (-1.823) (2.629) (.022)

ChngM_t-1 1.243 2.211 -.003096 .401 .448

(5.006) (7.556) (-5.036) (1.567) (1.792)

UPIL_t-1 -1.07 1.648 .003072 -1.182 -2.356

(-1.004) (1.311) (1.163) (-1.075) (-2.194)

IRBLK_t 518.363 762.802 .529 388.359 283.644

(4.846) (6.051) (1.997) (3.522) (2.634)

E_t-1 -20.741 61.591 1.283 223.626 108.398

(-.26) (.656) (6.503) (2.723) (1.352)

R_t 2.564 11.934 .00004739 -9.862 -14.813

(.454) (1.793) (-.003) (-1.694) (-2.605)

DW 2.222 2.872 1.993 2.437 2.516

Adjusted R² .995 .997 .962 .99 .947

F(9, 18) 669.842 982.304 79.323 312.132 56.383

SSR 56073.334 77861.265 .344 59583.964 56806.927

Estimates of Money Demand (M_t), Consumer Price Index (CPI_t),

M_d CPI_t L_d

Constant -235.273 -2.209 -63.512

(-.952) (-.266) (-.452)

C_t-1 .493 .007037 .06062

(5.372) (2.285) (.116)

REM_t -.674 -.0008154 .34

(-4.289) (-.155) (3.8)

F_t .102 .04573 .303

(.27) (3.605) (1.409)

K_t-1 .06897 .006036 .05093

(1.89) (4.942) (2.461)

ChngM_t-1 1.921 .05136 -.29

(7.104) (5.661) (-1.889)

UPIL_t-1 1.553 .115 -.634

(1.337) (2.952) (-.961)

IRBLK_t 242.759 19.232 278.187

(2.084) (4.920) (4.2)

E_t-1 70.211 4.117 143.499

(.809) (1.414) (2.909)

R_t -1.475 -.154 -8.187

(-.24) (-.748) (-2.34)

DW 2.294 1.539 2.001

Adjusted R² .99 .996 .979

F(9, 18) 314.965 852.869 146.247

SSR 66486.268 74.879 21498.998

Implications of the Estimated Econometric Model

The signs of all coefficients were as expected. The impact of lagged consumption level is positive, and less than 1.00. This result implies that the consumption behavior for the Jordanian economy is stable.

The most significant factor positively influencing the formation of capital is the remittances of Jordanians from abroad. The Jordanian remittances from abroad are sensitive to the oil prices and the political stability of the region.

The impact of Jordanian remittances from abroad on employment is positive, because all Jordanian remittances went to the private sector and were invested in the private sector, especially in construction, during the 1967-1996 period. Jordanian remittances and foreign aid have a positive effect on GDP, but the latter has a smaller impact than the former. The difference in impact is the result of the larger effect of Jordanian remittances on investment compared with that of foreign aid. Also, there is a positive relationship between Jordan's imports with each of the above two factors.

Foreign aid to Jordan had a negative impact on private consumption. The negative impact may be the result of increased government consumption, which produced an increase in CPI, and adverse consequences on private consumption.

Lagged capital stock had a positive effect on the GDP, as was anticipated. The positive impact of lagged capital stock on government expenditure was due to the income the government receives from its share on industrial and mining projected.

The positive relationship of lagged change in money supply and current imports was the result of the positive impact of lagged change in money supply on current CPI. The inflationary impact causes increased demand for importation of foreign goods as they become cheaper for Jordanian citizens. The positive effect of lagged change in money supply on exports may reflect the negative impact of the lagged money supply on the exchange rate. Also, the impact of lagged change in money supply on government capital spending and total government spending was positive but had no significant impact on government consumption expenditure.

The effect of a lagged unit price index of imports (UPIM_t-1) on investment was negative, which was also expected. The increased cost of importing capital goods had an adverse effect on the level of investment and government expenditure on capital. The positive impact of UPIM_t-1 on private consumption, government consumption, and imports was the result of the inelasticity of the Jordanian demand for foreign products because of absence of similar Jordanian products.

The blockade of Iraq represented by a dummy variable had a positive impact on all Jordanian economic variables except private consumption. This impact may be the result of the coincidence of the blockade, influx of more than 300,000 Jordanians from Kuwait, and the depreciation of the currency at the same time. Also, the blockade did not impact the size of the imports from Iraq but had a large negative impact on the size of exports to Iraq (see Central Bank of Jordan Monthly Statistical Bulletin of February 1991, April 1992, and August 1997)

The effect of the interest rate on the private consumption, total investment, and GDP was not significant, because the average growth rate of inflation was higher than the nominal interest rate minus the administrative lending cost. However, the effect of interest rate on the government capital expenditure was negative as it was anticipated.

The following are the results of the ex-post and ex-ante forecasts for the econometric model. The evaluation of the ex-post forecast accuracy is based on the Absolute Percentage Forecast Inaccuracy (APFI) for the forecasted period of 1993-1996. The APFI is used to evaluate the accuracy of the performance of the model's forecast for one year after the data have been used.

According to Evans (1973) the best test for a model's performance is the ex-ante forecast. Unfortunately, the ex-ante forecast evaluation cannot be accomplished without future data. The ex-post forecast is, therefore, the only alternative measure that the model builder in these circumstances can rely on. It is necessary to clarify what the meaning of ex-ante and ex-post forecast is before continuing the discussion about the model forecasts. The ex-post forecast is the result of the estimated model over the first 26 observations (1967-1992 period) and the use of the actual values of the independent variables to forecast for 1993. The above process of re-estimating the model repeated over the observation of a 1967-1993, 1967-1994, and 1967-1995 period to forecast for 1994,1995, and 1996 respectively (see appendix A). The ex-post forecasts that are associated with all values of the endogenous and the exogenous variables are known in advance with certainty. The ex-post forecasts are called unconditional forecasts.

The ex-ante forecast estimates the models over the entire available data set of 1967-1996, and then uses the models for forecasting the values of all endogenous and exogenous variables for the period of 1997-2000 (see appendix C). The forecast value of 1997 was obtained by taking the forecast values of the Box-Jenkins model for remittance, Foreign aid to Jordan, net factor income from abroad, unit price index of imports, and the Eurodollar interest rate. The above Box-Jenkins forecast values were combined with the actual values of the lagged independent variables. These figures were applied to the econometric model to forecast the values of all endogenous variables for 1997. The econometric model forecast for 1997-2000 is accomplished by using the forecast values of Box-Jenkins for the remittance, Foreign aid to Jordan, net factor income from abroad, unit price index of imports, and the Eurodollar interest rate (see appendix D), while the values of all other lagged variables were taken from the econometric forecast for 1996, 1997, 1998,and 1999 (see appendix C).

Finally, the total net capital stock for 1996 is obtained by adding the 1996 econometric forecast value for investment to net capital stock of 1995, and then multiplied the above by .956 which is equal to 1.00 minus the average depreciation rate of capital stock for Jordan during 1967-1995, which was 4.4%. Incorporating the net capital stock of 1996 is necessary to obtain the econometric model forecast for 1997. Ex-ante forecasts, or conditional forecasts, are the forecast values for all variables during 1997-2000, where some of the independent variables are not known with certainty. The error of ex-ante forecast is therefore, a combined error of the model and the error of the expected value of the exogenous variables. Fair (1984) argued that the econometric model should not be penalized for the second source of error.

Jordan’s forecasting model is no exception to Pindyck’s (1981) arguments that some of the equations in a given model can produce a good forecast while other equations may not, forcing model builders to accept the equations which do not produce good forecasts. This was the case for the total government expenditure equation, because total government expenditure increased from an annual average growth rate of 12.68% during 1967-1992 to a growth rate of 22.18% in 1993 and declined 4.1% in 1994 for the first time in 25 years. Therefore, the econometric model under-forecasted the actual data for 1993 and over-forecasted for the year of 1994.

The influx of more than 300,000 citizens from Kuwait in 1991 caused the model to under forecast the employment level and exports, but the forecast improved every year over the period of 1993-1995. The exchange rate of Jordanian currency in terms of the dollar was relatively stable during the period of 1967-1989; then a drastic devaluation of the currency in 1989, 1990, and 1991 occurred . The exchange rate has been stabilized since that time. Therefore, the model over-forecasted for the years of 1993 and 1994 and under-forecasted for 1995. The forecast will improve as the economy moves toward its typical condition.

GNP forecast is the result of the sum of GDP and net factor income from abroad (NFIFA). Since 1995, Jordan has fixed its exchange rate of the Jordanian currency (Dinar) to $1.41, and policy makers announced that the exchange rate will be fixed for the foreseeable future. Therefore, the model was used for the exchange rate ex-post forecast and not for ex-ante forecast.

Box and Jenkins (1976) developed a practical procedure for an entire family of models, the autoregressive integrated moving-average (ARIMA) models.  The ARIMA models are applicable only to a stationary data series, where the mean, the variance, and the autocorrelation function remain constant through time.  ARIMA models are appropriate for series with strong trend characteristics, random walk series, and seasonal and nonseasonal time series.

Cleary and Levenbach (1982), Anderson (1976), and Pankratz (1983) point out that the Box-Jenkins approach is a powerful and flexible method  for short term forecasting because ARIMA models place more  emphasis on the recent past and where structural shifts occur gradually, rather than suddenly. This makes the ARIMA models especially valuable when we are dealing with economic time series data.  This emphasis on the recent past makes long term forecasts less reliable due to accumulation of error terms.

The process of selecting the model is a process of evaluation, adaptation, and trial and error.  When a phenomenon is completely understood it is possible to describe it exactly in a mathematical expression.  Nevertheless, in economics we use incomplete theoretical knowledge to indicate a suitable class of mathematical functions which can then be fitted empirically.  In the first stage of selecting a model we need to identify a rough class of models, followed by identification of their subclasses. This tentative model is then fitted to the data and  estimated for its parameters.  The rough estimates obtained during this identification stage are used as starting values for estimating the parameters.  Finally, we use diagnostic checks to discover any lack of fit.  If no inadequacy is indicated the estimated model will be used for forecasting. However, if any inadequacy is found, the iterative cycle of identification, estimation, and diagnostic checking is repeated until a suitable model is found.

The ARIMA (p,d,q) models for remittance (REM_t), foreign aid (F_t), interest rate on the Eurodollar (R_t) , unit price index of imports (UPIM_t), and net factor income from abroad (NFIFA) have been scrutinized through all diagnostic checks needed for viable models. The process of the remittances(REM_t), foreign aid (F_t), and NFIFA conforms to ARIMA (1, 1, 0), while the process of unit price index of imports (UPIM_t) and interest rate on the Eurodollar (R_t) conforms to ARIMA (0, 1, 1) and (2,1,0) respectively.

REM_t F_t NFIFA R_t

Constant 26.09 0 0 0

t-ratio 1.89 0 0 0

AR(1) .3499 -.5474 .5119 .3559

t-ratio 1.69 -3.46 3.15 2.14

AR(2) -.4749

t-ratio -2.86

SS residual 146110 52995.8 18215.6 105.17

DF 27 28 28 28

Lag 12 24 12 24 12 24 12 24

chi square 11.1 15.9 8.6 12.2 3.3 10.5 12.4 15

UPIM t-ratio

Constant 5.234 1.51

MA(1) -.5353 -3.3

SS residual 4002.61

DF 27

lag chi square

12 11.8

24 17.4

The measure of the absolute percentage forecast inaccuracy (APFI) gives the percentage of deviation of the forecasted value from the actual value for each year. APFI had been used to measure the inaccuracy of ex-post forecast (see appendix B). The APFI can be expressed mathematically in the following manner:

APFI = absolute value of (_Forecast - _Actual)/_Actual

The model was successful in ex-post forecasting for most of the key Jordanian economic variables and provides very useful information on the dynamics of the Jordanian economy. The model shows the significant impact of the Jordanian remittances on investment relative to the weak influence of foreign aid on investment. The difference in impact shows that the future growth of Jordan relies on the private sector and not on the public sector.

1993 1994 1995 1996

Gross National Product 3812.686 3939.449 4554.028 5272.379

Gross Domestic Product 3971.486 4090.849 4670.828 5383.879

Private Consumption 3218.658 3161.024 3076.473 3413.474

Investment 1351.748 1331.923 1599.674 1881.843

Government Consumption 901.531 950.8469 1097.313 1235.857

Exports 1733.434 1884.396 2380.661 2776.492

Imports 3234.016 3241.57 3475.284 3925.829

Exchange Rate (Jordanian

Currency in Terms of the $) 1.8106 1.7564 1.14051.2384

Total Government Expenditures 1471.77 1767.827 1753.806 2062.369

Other Government Expenditures 580.5853 813.7158 658.0225 827.388

Money Demand 1745.187 1722.772 1899.636 1904.559

Consumer Price Index 102.3763 105.1294 115.1552 119.823

Employment Level in 000 727.6092 815.4913 996.2099 1191.041

1993 1994 1995 1996

Gross National Product .0438 -(.0273) .00358 .0471

Gross Domestic Product .042 -.0263 .0035 .0461

Private Consumption .1832 .1394 .0176 .0059

Investment -.0499 -(.0821) .0338 .0447

Government Consumption .0509 -(.0397) .0149 .0351

Exports -(.1165) -(.0998) -(.0236) .0691

Imports .0261 .0431 .0116 .0226

Exchange Rate (Jordanian

Currency in Terms of the $) .2487 .2283 -(.2024) -(.1217)

Total Government Expenditures -(.1068) .1187 -(.0514) .1169

Other Government Expenditures -(.2650) .3789 -(.1428) .2679

Money Demand .0087 -(.0134) .0882 .2374

Consumer Price Index -(.0089) -(.0175) .0516 .0276

Employment Level -(.1533) -(.1457) -(.0512) .0895

1997 1998 1999 2000

Gross National Product 5631.616 6305.279 7255.632 7958.88

Gross Domestic Product 5740.403 6412.678 7362.32 8065.212

Private Consumption 3522.72 4190.473 5044.761 5373.024

Investment 1785.099 2071.367 2480.747 2254.763

Government Consumption 1379.509 1510.712 1689.631 1951.72

Exports 2660.957 3422.817 4284.365 4120.556

Imports 3759.684 4781.09 6135.221 5632.511

Exchange Rate (Jordanian

Currency in Terms of the $) 1.41 1.41 1.41 1.41

Total Government Expenditures 2049.78 2341.495 2890.446 3189.185

Other Government Expenditures 667.2179 827.6134 1197.19 1233.207

Money Demand(1997 is actual) 1662.4 2265.91 2480.75 3362.53

Consumer Price Index 118.2179 145.4611 184.0511 181.8712

Employment Level in 000 1266.293 1248.948 1206.297 1430.972

1997 1198.95 186.647 -108.787 187.002 5.66(actual)

1998 1261.48 189.194 -107.399 192.236 5.99

1999 1309.44 187.8 -106.688 197.47 6.003

2000 1352.31 188.563 -106.324 202.704 5.85

2001 1393.4 188.145 -106.138 207.938 5.7906

Anderson, W. H. L. (1967), Business Fixed Investment: A Marriage of Fact and Fancy, (eds). by Ferber, R. Columbia University. 413-425.

Anderson, O,D. (1976) Time Series Analysis and Forecasting. London: Butterworths.

Baumol, W. J. (1952) "The Transactions Demand for Cash: An Inventory Theoretical Approach", Quarterly Journal of Economics 66, 545-556.

Box, G.P. and G.M. Jenkins (1976) Time Series Analysis Forecasting and Control. San Francisco: Holden-Day.

Central Bank of Jordan (1997) Monthly Statistical Bulletin. Amman: Department of Research and Studies. Vol. 33, No. 8.

_______ (1992) Monthly Statistical Bulletin. Amman: Department of Research and Studies. Vol. 28, No. 4.

_______ (1991) Monthly Statistical Bulletin. Amman: Department of Research and Studies. Vol. 27, No. 2.

_______ (1996) Yearly Statistical Series (1964-1995), Special Issue. Amman: Department of Research and Studies.

Cleary, J.P. and H. Levenbach (1982). The Professional Forecaster: The Forecasting Process Through Data Analysis. Belmont: Lifetime Learning Publications.

Diebold, F. X. (1998) Elements of Forecasting. South-Western College Publishing. Cincinnati, Ohio.

Evans, M. K. (1973) Econometric Models: Method and Techniques of Business Forecasting, Butler, W., Kavesh, R., and R. Platt (eds) (Englewood Cliffs: Prentice-Hall) 161-189.

Fair, C. R. (1984) Specification, Estimation, and Analysis of Macroeconometric Models Cambridge: Harvard University Press.

Friedman, M. (1956) "The Quantity Theory of Money: A Restatement", In Studies in the Quantity Theory of Money, eds. M. Friedman (eds), Chicago: University of Chicago Press.

________ (1957) A Theory of the Consumption Function. Princeton: Princeton University Press.

________ (1982) "Monetarism in Rhetoric and in Practice", Monetary policy in our Times: Proceedings of the First International Conference held by the Institute of Monetary and Economic Studies of the Bank of Japan, Albert A., H. Eguchi, R. Farmer, and Y. Suzuki (eds), Cambridge : MIT Press. 15-28.

Gapiniski, J. H. (1982) Macroeconomic Theory: Statics, Dynamics, and Policy. New York: McGraw-Hill.

Gapiniski, J. H., Skegro, B., and Thomas W. Zuehlke (1989) "A Model of Yugoslav Performance", Journal of Comparative Economics, 13, 15-46.

Greene, W. H. (1988) limdep: Version 5. New York: William Greene. 133.

Judge, G.G., Griffiths, W.E., Lutkepohl, H. and T. Lee (1985) The Theory and Practice of Econometrics, Second ed. New York: John Wiley and Sons. 599-600.

Jordan Ministry of Labor & Social Development (1978) Annual Report. Amman, Jordan.

______________________ (1980) Annual Report. Amman: Ministry of Labor.

______________________ (1981) Annual Report. Amman: Ministry of Labor.

______________________ (1983) Annual Report. Amman: Ministry of Labor.

______________________ (1984) Annual Report. Amman: Ministry of Labor.

______________________ (1985) Annual Report. Amman: Ministry of Labor.

______________________ (1986) Annual Report. Amman: Ministry of Labor.

______________________ (1979) Quarterly Report. Amman: Ministry of Labor. 7/8.

______________________ (1981) Quarterly Report. Amman: Ministry of Labor. 13/14.

______________________ (1987) Quarterly Report. Amman: Ministry of Labor. 39.

Pankratz, A. (1983) Forecasting with Univariate Box-Jenkins Model. New York: John Wiley and Sons.

Pindyck, R.S. and D.C. Rubinfeld (1981) Econometric Models and Economic Forecasts, Second ed. New York: McGraw-Hill.

Saunders, P. G., and D. J. Taylor (1976) "The Demand for Money", Topics in Applied Macroeconomics, Healthfield,D. (eds) London: Macmillan Press. 186-207.

Theil, Henri (1971) Principles of Econometrics. New York: John Wiley and Sons.

Timbrell, M. C. (1976) "Consumption Functions", Topics in Applied Macroeconomics, Heathfield, D. (eds), London: Macmillan. 164-185.

Tobin, J. (1956) ‘Liquidity preference as Behavior towards Risk’, Review of Economic Studies, 25(4), 81-95.

Zellner, Arnold (1984) Basic Issues in Econometrics. Chicago: The University of Chicago Press.