A. The Monetary Approach

The monetary approach assumes a stable demand for real money balances determined by real income and returns on alternative assets. Given aggregate supply constraints in the short run, an expansionary monetary policy will result in higher prices. OECD (1995) finds that, in the long run, wholesale price inflation is exclusively determined by the money supply, while Togan (1987) and De Santis (1993), using inverted money demand functions, produce estimates of inflation which track the trend of actual inflation rather well. Ozatay (1992a) shows that monetary aggregates can help predict movements of nominal income and the GNP deflator.

B. The Public Finance Approach

The public finance approach emphasizes that, given the limits on domestic and foreign borrowing dictated by financial market conditions and solvency requirements, monetization is the residual form of deficit financing.

Rodrik (1991) finds a one-to-one relationship, at the margin, between public sector deficits and inflation, with inflation inertia not significant over the estimation period. He also notes that, during the 1980s, the inflationary consequences of a given deficit were further exacerbated by a decline in the demand for base money due to the relaxation of foreign exchange regulations, which induced a portfolio reallocation toward foreign currency-denominated assets and a significant erosion of the base of the inflation tax. Anand and van Wijnbergen (1988, 1989), van Wijnbergen (1989) and World Bank (1991) formulate their analyses in a framework that can be used to derive inflation for a given deficit and money demand; however, multiple equilibrium inflation rates can result because the inflation rate is also a determinant of money demand. Finally, Batavia and Lash (1983) find evidence of a vicious circle between inflation and public sector deficits in the period 1950–75: they maintain that inflation increases the public sector deficit because it raises expenditure faster than revenues.

It has been argued that fiscal deficits have been affected by the political cycle and that public price policy has been used in this connection, by keeping public prices unchanged before elections and raising them after the elections (Rodrik). Thus, inflation is lower just before an election and increases afterward, both because of the direct effect of public prices in the total price index and because of the indirect effect of public prices through private prices. The link between public prices and private price dynamics has been investigated for the period 1982–1990 by Ozatay (1992b). He finds that the response of private prices to electrical energy prices, prices of refinery products, and mining sector prices—which account for almost half of the public PPI—is sizable and occurs quickly.

Metin (1995) takes into account various possible sources of inflation, including cost push factors and concludes that fiscal variables dominate the inflationary process in Turkey. In addition, she does not find significant effects of disequilibria in the labor and goods markets, while the imbalances in the money market have an effect, but only in the short run.

C. Structural and Cost-Push Explanations

Finally, a number of studies explore the role that structural and cost-push factors play in the inflationary process. These factors include: (i) the link between the exchange rate and prices; (ii) the markup on final product prices due to the oligopolistic industrial structure; and (iii) wage pressures stemming from indexation rules and entrenched inflationary expectations.

Rising inflation and continued depreciation of the Turkish lira throughout the 1980s give rise to the hypothesis of a devaluation-inflation spiral. In Turkey, this process is facilitated by the relatively high dependence of the economy on imports of capital and intermediate goods, and the predominantly oligopolistic industrial structure that allows a markup over costs by manufacturing firms. Under these circumstances, increases in the price of foreign currency or the dollar price of imported inputs are translated into higher prices of domestic products. However, the empirical evidence on the validity of this argument is mixed.

Onis and Ozmucur (1990) find evidence of a two-way causal link between exchange rates and prices. Rittenberg (1993) and Metin (1995), however, provide counter evidence. Using Granger causality tests, Rittenberg shows that causality runs from price level changes to exchange rate changes, but not vice versa. In a more recent study, Metin finds no evidence that purchasing power parity and uncovered interest parity relationships hold for the period 1948 to 1987, although she cites several structural shifts during the estimation period which may have obscured the transmission effects. Although no consensus has been reached on the causality between exchange rates and inflation, it is worthwhile noting that a key assumption of the devaluation-inflation argument was not fulfilled during the post-1980 period; that is, monetary policy was not uniformly accommodative (da Cunha, Webb, and Isaac, 1990).²

Few studies have tested the relative significance of oligopolistic pricing behavior in the Turkish manufacturing sector on inflation. In general, these studies indicate that markup pricing alone cannot explain the causes of persistent inflation in Turkey. Uygur (1990) finds that firms’ markup, expressed as a function of changes in excess demand, statistically determines private manufacturing wholesale price inflation, but with only a relatively small impact. Instead, inflationary expectations are found to be the more relevant factor, accounting for about 75 percent of the magnitude of the price changes. In turn, these expectations are shown to be determined by: (i) inertia—higher current inflation leads to higher expected inflation; (ii) public sector policy (e.g., energy prices); and (iii) uncertainty.³ The study by da Cunha, Webb, and Isaac resembles Uygur in that a markup rule expressed as a function of changes in excess demand is estimated; they stress, however, the difficulty of determining the degree to which price markup responds positively to changes in excess demand. Estimation of the price equation indicates that the markup is positively associated to changes in excess demand; however, sectoral differences could not be identified.

Yeldan (1993) argues that the underlying sources of inertial inflation in Turkey originate from the prevailing income inequality and conflicting social claims on national output, which act to propagate cost inflation in Turkey in the 1980s. Attempts by each sector to set its price so as to maximize its share of a given output seldom lead to a change in income distribution, but result instead in inflation. However, to reach the order of magnitude of recent inflation rates, the markup implied by Yeldan’s model would not only be unusually high, but would also suggest continually rising markups on the part of firms, irrespective of changes in demand conditions.

Wage pressures are not found to have a substantial impact on the inflationary process in Turkey during the 1980s (Onis and Ozmucur, 1990; da Cunha, Webb, and Isaac, 1990; Ozatay, 1992a; Yeldan, 1993; and Metin, 1995). In a model of wage and price determination, da Cunha, Webb, and Isaac, show that between 1978 and early 1982, wages were positively and significantly associated with changes in excess demand and past inflation, the latter suggesting backward-looking indexation or expectations. However, between 1982 and late 1988, when a more constrained wage policy regime was in effect, wages served to ease rather than trigger inflationary pressures in the economy.

A. A Long-Run Model

Standard aggregate demand and supply functions are assumed to determine the long-run equilibrium level for domestic prices.⁴ Aggregate demand increases if real money balances rise and/or competitiveness improves (i.e., if domestic prices in foreign currency terms decline relative to foreign prices of competing exports). Aggregate supply declines if real wages and/or imported input prices increase. Hence, the balance of aggregate demand, y^d, and supply, y⁸, can be written as:

where P denotes the domestic price level, W nominal wages, E the exchange rate (defined as the price of domestic currency in foreign currency),⁵ M money, P*_x the exogenous price of exports, P*_r exogenous imported input prices, and ∈_d and ∈_s are random demand and supply shocks, respectively. Solving for the price level yields the following long-run price equation:

The expected signs of money, wages, export and import prices are positive, while the expected sign for the exchange rate is negative.

To identify pressures on prices deriving from external sector disequilibrium, the observed level of the real exchange rate is compared with a measure of the “equilibrium” real exchange rate (RER), defined to be the real exchange rate that, given estimated export and import relationships, is compatible with a financeable deficit on the goods and services account. The financeable deficit (K) is estimated by average levels of medium and long-term capital flows, current transfers, and net factor income. Note that the RER, as defined here, is not an equilibrium exchange rate in the broader sense of being sustainable over the medium term, but more narrowly the exchange rate that yields a financeable deficit on goods and services, given existing patters of capital inflows. Given standard export and import functions, with export (import) volumes depending on the real exchange rate and foreign (domestic) income, the RER can be derived as follows:

where P* is a weighted average of foreign export and import prices, δ is the estimated coefficient of domestic income (y) in the import function, β is the estimated coefficient of foreign income (y_f) in the export function, ξ and μ are the estimated constants in the export and import function, respectively, and α and γ are the estimated coefficients of the real exchange rate in the export and import function, respectively. The difference between the actual and equilibrium real exchange rates provides a measure of disequilibrium in the external sector that impacts on price developments.

Equilibrium in the money market is obtained when the public sector deficit, as a ratio to GNP, is fully financed by long-run seigniorage—money creation—and a sustainable level of borrowing. The long-run seigniorage level is assumed to change in proportion to expected inflation and real income growth; for simplicity, a unit elasticity of real money balances demanded with respect to income is assumed. Sustainable borrowing, both domestically and externally, is defined by reference to a debt-to-output ratio that is constant or not rising. The financeable deficit, G*, can then be written as the maximum deficit consistent with the projected demand for real money balances and a constant debt to output ratio:

where ${\dot{y}}^{*}$ is the projected growth in real output, Y is nominal output, π^* is the targeted inflation rate, H base money and ${\dot{D}}^{*}$ are changes in total debt such that the debt-output ratio is constant. The difference between the actual public sector deficit and the financeable deficit, as measured in equation (3), yields the disequilibrium in the money market. If a large positive disequilibrium is obtained then upward pressures on inflation will emerge. This is referred to as the required deficit reduction consistent with the stipulated macroeconomic targets.⁶

The long-run real wage is determined as a weighted average of the real wage offered by firms (i.e., the real wage on the labor demand function) and the real wage demanded by workers (i.e., the real wage on the labor supply function), with weights determined by the relative bargaining power of firms and workers. Hence, the long-run real wage is given by:

where L is employment so that y/L measures productivity, and u is the unemployment rate. The disequilibrium in the labor market is then estimated by the residuals of equation (4).

B. A Short-Run Model for Inflation

The short-run inflation equation is an error correction representation of equation (1) expanded to include the disequilibrium term from the other three sectors:

where π is the inflation rate, ECM, the error correction term, D_i, i=e,m,w, represent the respective deviation of the actual real exchange rate, public sector deficit and real wage from their long run equilibrium, Z_t is a vector of other exogenous variables such as changes in real government expenditure, and υ_t is the residual.

The dynamic adjustment of the rate of inflation depends not only on the speed at which the disequilibrium within the goods market corrects itself, but also on the dynamic adjustment of the foreign exchange market, the money market and the labor market to their respective long-run equilibria. The expected sign of the exchange rate disequilibrium is negative: when the real exchange rate is more appreciated than its long-run level downward pressure on inflation is exerted. The disequilibrium effects originating from the money and labor markets are expected to be positively signed, as an excess supply of money and a higher-than-equilibrium real wage level raise inflation.

A. Data

The model is estimated with quarterly time series data from 1970 to 1995; see Appendix I for a description of data sources. The choice of the sample period is dictated by the desire to take a long-term view; the rationale for quarterly as opposed to annual frequency, is to capture short-term inflation dynamics. For some series, quarterly data had to be interpolated from annual data.

Prices are measured by the wholesale price index for the private manufacturing sector, thereby excluding administratively influenced state enterprise prices and agricultural prices. However, because the private manufacturing index was not available prior to 1981, the total wholesale price index is used in the earlier period. Wages are estimated from the annual payments made to employees in the private manufacturing sector as well as from survey data of the Turkish Confederation of Employers’ Association. The monetary aggregate used is broad money inclusive of foreign exchange deposits, M2X, and the exchange rate variable is the nominal effective exchange rate as provided by the IMF Information Notice System.

B. Estimates for the Long-Run Model

Prices

The long-run price equation is estimated over the period 1970 to 1995 (Table 1). Both likelihood ratio tests of the Johansen procedure point to multiple cointegrating relationships. The choice of the cointegrating vector to be included in the short-run inflation equation is determined by the vector that has the right signs as expected from economic theory and is shown below:

$P=-0.57E+0.22M+0.08W+0.52P_x+0.01P_r$

Table 1

Long-Run Price Equation ^1/

Johansen Maximum Likelihood Procedure: lag in VAR = 6

94 Observations from 1971-Q3 to 1994-Q4

^*Denotes the vector chosen to represent the long run relationship.

^1/All variables are in logarithm.

^2/r indicates the number of cointegrating vectors.

Table 1

Long-Run Price Equation ^1/

Johansen Maximum Likelihood Procedure: lag in VAR = 6

94 Observations from 1971-Q3 to 1994-Q4

Null 2/	Alternative 2/	Statistic	95% Critical Value	90% Critical Value
Cointegration LR Test Based on Maximal Eigenvalue of the Stochastic Matrix
r = 0	r = 1	61.8627	39.3720	36.7620
r<= 1	r = 2	38.6308	33.4610	30.9000
r< = 2	r = 3	23.8397	27.0670	24.7340
r< = 3	r = 4	14.3923	20.9670	18.5980
Cointegration LR Test Based on Trace of the Stochastic Matrix
r =0	r>= 1	148.6835	94.1550	89.4830
r<= 1	r> = 2	86.8208	68.5240	64.8430
r< = 2	r> = 3	48.1900	47.2100	43.9490
r< = 3	r> = 4	24.3503	29.6800	26.7850
Estimated Cointegrated Vectors
	Vector 1	Vector 2*	Vector 3	OLS (t-ratio)
p	-1.0000	-1.0000	-1.0000
E	-0.4515	-0.5722	-0.4840	-0.612 (16.2)
M	0.3832	0.2230	0.2640	0.158 (4.0)
W	0.1286	0.0750	0.1155	0.150 (7.2)
Px	-1.8061	0.5193	1.3952	0.287 (2.7)
Pr	1.5766	0.0064	-0.4768	0.226 (3.2)
Intercept				-0.307 (0.5)

^*Denotes the vector chosen to represent the long run relationship.

^1/All variables are in logarithm.

^2/r indicates the number of cointegrating vectors.

View Table

Table 1

Long-Run Price Equation ^1/

Johansen Maximum Likelihood Procedure: lag in VAR = 6

94 Observations from 1971-Q3 to 1994-Q4

Null 2/	Alternative 2/	Statistic	95% Critical Value	90% Critical Value
Cointegration LR Test Based on Maximal Eigenvalue of the Stochastic Matrix
r = 0	r = 1	61.8627	39.3720	36.7620
r<= 1	r = 2	38.6308	33.4610	30.9000
r< = 2	r = 3	23.8397	27.0670	24.7340
r< = 3	r = 4	14.3923	20.9670	18.5980
Cointegration LR Test Based on Trace of the Stochastic Matrix
r =0	r>= 1	148.6835	94.1550	89.4830
r<= 1	r> = 2	86.8208	68.5240	64.8430
r< = 2	r> = 3	48.1900	47.2100	43.9490
r< = 3	r> = 4	24.3503	29.6800	26.7850
Estimated Cointegrated Vectors
	Vector 1	Vector 2*	Vector 3	OLS (t-ratio)
p	-1.0000	-1.0000	-1.0000
E	-0.4515	-0.5722	-0.4840	-0.612 (16.2)
M	0.3832	0.2230	0.2640	0.158 (4.0)
W	0.1286	0.0750	0.1155	0.150 (7.2)
Px	-1.8061	0.5193	1.3952	0.287 (2.7)
Pr	1.5766	0.0064	-0.4768	0.226 (3.2)
Intercept				-0.307 (0.5)

^*Denotes the vector chosen to represent the long run relationship.

^1/All variables are in logarithm.

^2/r indicates the number of cointegrating vectors.

The residuals from this estimation are included as the error correction term ECM in equation (5). The expected sign is negative because when prices are above their equilibrium level, due for instance to a shock, downward pressure on prices is exerted, thus “correcting” the error and driving prices back toward equilibrium.

The real exchange rate

All variables in the export and import equations—with the exception of the sum of capital inflows, transfers and income—were transformed by an eight-quarter moving average, in an effort to capture long-term trends. The financeable deficit for the goods and services balance was calculated as a four-quarter moving average of the sum of medium- and long-term capital inflows, transfers, and net factor income. No cointegrating relationships were identified for both the export and import functions when estimated over the period 1970 to 1995, which is not surprising given the significant structural break that occurred in the trade regime with the shift from an inward-looking import substitution strategy in the 1970s to an export-oriented growth strategy in the 1980s. To account for this shift in the trade regime, the estimation period was divided into two subsample periods: (i) from 1970 to 1983 and (ii) from 1984 to 1995.⁷ A cointegrating relationship was obtained for both the export and import functions in each of the subsample periods (Tables 2a and 2b). The cointegrating vector for exports includes exports of goods and services (deflated by the export price index), the real effective exchange rate, and an index of world imports, which proxies the level of demand in partner countries. On imports, the cointegrating vector includes imports of goods and services (deflated by the import price index), the real effective exchange rate and real domestic income. All the variables have the expected sign in both periods.

Table 2a

Exports of Goods and Services ^1/

Johansen Maximum Likelihood Procedure: lag in Var = 4

^*Denotes vector chosen.

^1/All variables are transformed by an eight-quarter moving average

^2/See footnote 2 in Table 1.

Table 2a

Exports of Goods and Services ^1/

Johansen Maximum Likelihood Procedure: lag in Var = 4

Null 2/	Alternative 2/	Statistic	95% Critical Value	90% Critical Value
Cointegration LR Test Based on Maximal Eigenvalue of the Stochastic Matrix
1972-Q4 to 1983-Q4
r =0	r = 1	39.6016	20.9670	18.5980
r<= 1	r = 2	9.0659	14.0690	12.0710
1984-Q1 to 1994-Q4
r =0	r= 1	43.2854	20.9670	18.5980
r<= 1	r = 2	13.4363	14.0690	12.0710
Cointegration LR Test Based on Trace of the Stochastic Matrix
1972-Q4 to 1983-Q4
r =0	r>= 1	51.7204	29.6800	26.7850
r<= 1	r> = 2	12.1188	15.4100	13.3250
1984-Q1 to 1994-Q4
r =0	r> = 1	60.1443	29.6800	26.7850
r<= 1	r> = 2	16.8589	15.4100	13.3250
Estimated Cointegrated Vectors
		1972-Q4 -1983-Q4		1984-Q1 - 1994-Q4
	Vector 1	OLS (t-ratio)	Vector 1*	OLS (t-ratio)
X	-1.000		-1.000
RER	-0.274	-0.196 (9.5)	-0.133	-0.195 (9.9)
yf	0.312	0.080 (2.2)	0.820	0.834 (75.2)
Intercept		37.528 (8.4)		-9.493 (4.2)

^*Denotes vector chosen.

^1/All variables are transformed by an eight-quarter moving average

^2/See footnote 2 in Table 1.

View Table

Table 2a

Exports of Goods and Services ^1/

Johansen Maximum Likelihood Procedure: lag in Var = 4

Null 2/	Alternative 2/	Statistic	95% Critical Value	90% Critical Value
Cointegration LR Test Based on Maximal Eigenvalue of the Stochastic Matrix
1972-Q4 to 1983-Q4
r =0	r = 1	39.6016	20.9670	18.5980
r<= 1	r = 2	9.0659	14.0690	12.0710
1984-Q1 to 1994-Q4
r =0	r= 1	43.2854	20.9670	18.5980
r<= 1	r = 2	13.4363	14.0690	12.0710
Cointegration LR Test Based on Trace of the Stochastic Matrix
1972-Q4 to 1983-Q4
r =0	r>= 1	51.7204	29.6800	26.7850
r<= 1	r> = 2	12.1188	15.4100	13.3250
1984-Q1 to 1994-Q4
r =0	r> = 1	60.1443	29.6800	26.7850
r<= 1	r> = 2	16.8589	15.4100	13.3250
Estimated Cointegrated Vectors
		1972-Q4 -1983-Q4		1984-Q1 - 1994-Q4
	Vector 1	OLS (t-ratio)	Vector 1*	OLS (t-ratio)
X	-1.000		-1.000
RER	-0.274	-0.196 (9.5)	-0.133	-0.195 (9.9)
yf	0.312	0.080 (2.2)	0.820	0.834 (75.2)
Intercept		37.528 (8.4)		-9.493 (4.2)

^*Denotes vector chosen.

^1/All variables are transformed by an eight-quarter moving average

^2/See footnote 2 in Table 1.

Table 2b

Imports of Goods and Services ^1/

Johansen Maximum Likelihood Procedure: lag in Var = 4

^*Denotes vector chosen.

^1/All variables are transformed by an eight-quarter moving average.

^2/See footnote 2 in Table 1.

Table 2b

Imports of Goods and Services ^1/

Johansen Maximum Likelihood Procedure: lag in Var = 4

Null 2/	Alternative 2/	Statistic	95% Critical Value	90% Critical Value
Cointegration LR Test Based on Maximal Eigenvalue of the Stochastic Matrix
1972-04 to 1983-Q4
r =0	r = 1	26.1287	20.9670	18.5980
r<= 1	r = 2	11.0492	14.0690	12.0710
1984-01 to 1994-Q4
r = 0	r = 1	49.0748	20.9670	18.5980
r<= 1	r = 2	12.5883	14.0690	12.0710
Cointegration LR Test Based on Trace of the Stochastic Matrix
1972-Q4 to 1983-Q4
r =0	r> = 1	38.5671	29.6800	26.7850
r<= 1	r> = 2	12.4384	15.4100	13.3250
1984-Q1 to 1994-Q4
r =0	r>= 1	66.8997	29.6800	26.7850
r<= 1	r> = 2	17.8249	15.4100	13.3250
Estimated Cointegrated Vectors
		1972-Q4 -1983-Q4		1984-Q1 -1994-Q4
-	Vector 1	OLS (t-ratio)	Vector 1*	OLS (t-ratio)
X	-1.000		-1.000
RER	0.059	0.069 (2.7)	0.141	0.235 (4.5)
Yf	0.003	0.002 (5.2)	0.005	0.006 (36.7)
Intercept		-5.061 (8.4)		-80.940 (12.7)

^*Denotes vector chosen.

^1/All variables are transformed by an eight-quarter moving average.

^2/See footnote 2 in Table 1.

View Table

Table 2b

Imports of Goods and Services ^1/

Johansen Maximum Likelihood Procedure: lag in Var = 4

Null 2/	Alternative 2/	Statistic	95% Critical Value	90% Critical Value
Cointegration LR Test Based on Maximal Eigenvalue of the Stochastic Matrix
1972-04 to 1983-Q4
r =0	r = 1	26.1287	20.9670	18.5980
r<= 1	r = 2	11.0492	14.0690	12.0710
1984-01 to 1994-Q4
r = 0	r = 1	49.0748	20.9670	18.5980
r<= 1	r = 2	12.5883	14.0690	12.0710
Cointegration LR Test Based on Trace of the Stochastic Matrix
1972-Q4 to 1983-Q4
r =0	r> = 1	38.5671	29.6800	26.7850
r<= 1	r> = 2	12.4384	15.4100	13.3250
1984-Q1 to 1994-Q4
r =0	r>= 1	66.8997	29.6800	26.7850
r<= 1	r> = 2	17.8249	15.4100	13.3250
Estimated Cointegrated Vectors
		1972-Q4 -1983-Q4		1984-Q1 -1994-Q4
-	Vector 1	OLS (t-ratio)	Vector 1*	OLS (t-ratio)
X	-1.000		-1.000
RER	0.059	0.069 (2.7)	0.141	0.235 (4.5)
Yf	0.003	0.002 (5.2)	0.005	0.006 (36.7)
Intercept		-5.061 (8.4)		-80.940 (12.7)

^*Denotes vector chosen.

^1/All variables are transformed by an eight-quarter moving average.

^2/See footnote 2 in Table 1.

Figure 3 plots the actual and estimated equilibrium real exchange rates; the difference between the actual and equilibrium exchange rates is included as D_e in the inflation equation (5) (Figure 4, top panel). It would appear that the real exchange rate was overvalued during most of the 1970s but this was reversed in the 1980s when the authorities actively pursued a real depreciation policy. The appreciation of the equilibrium exchange rate in the early 1990s reflects the improving terms of trade and the increased availability of external financing, a trend reversed in 1994 with the cutoff in access to capital markets.⁸

View Full Size

Turkey: Real Effective Exchange Rate, 1973-95

(1990=100)

Citation: IMF Working Papers 1997, 170; 10.5089/9781451858389.001.A001

Source: IMF, Information Notice System. Equilibrium real effective rates are as calculated from equation (2)

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Turkey: Market Disequilibria, 1972-94

(In percent)

Citation: IMF Working Papers 1997, 170; 10.5089/9781451858389.001.A001

Source: As estimated from the model.

• Download Figure
• Download figure as PowerPoint slide

The real wage

Although a cointegrating vector was obtained for the wage equation, it was not satisfactory in that the coefficient for the unemployment rate was incorrectly signed (Table 3). Failure to obtain a satisfactory result may partly reflect data deficiencies, but may also be due to heavy regulation of labor markets, notably during the 1980s. The residuals from the cointegrating vector, D_w, were included in the inflation equation (5) in an attempt to capture the inflationary impact of the disequilibrium in the labor market, but did not prove to be significant.

Table 3

Long Run Real Wage Equation ^1/

Johansen Maximum Likelihood Procedure: lag in VAR = 8

92 Observations from 1972-Q1 to 1994-Q4

^1/All variables are in logarithm except for the unemployment rate, UNR, and the dummy variable, DUM1 (1970- 1987=0; 1988-994=1).

^2/See footnote 2 in Table 1.

Table 3

Long Run Real Wage Equation ^1/

Johansen Maximum Likelihood Procedure: lag in VAR = 8

92 Observations from 1972-Q1 to 1994-Q4

Null 2/	Alternative 2/	Statistic	95% Critical Value		90% Critical Value
Cointegration LR Test Based on Maximal Eigenvalue of the Stochastic Matrix
r = 0	r = 1	30.2914	27.0670		24.7340
r<= 1	r = 2	13.1140	20.9670		18.5980
Cointegration LR Test Based on Trace of the Stochastic Matrix
r = 0	r > = 1	52.7715	47.2100		43.9490
r<= 1	r > = 2	22.4800	29.6800		26.7850
Estimated Cointegrated Vectors
		Vector 1		OLS (t-ratio)
RW		-1.000
RPP		0.139		-0.029 (0.24)
UNR		0.241		0.098 (2.7)
DUM1		0.745		0.561 (7.4)
Intercept				11.122 (7.6)

^1/All variables are in logarithm except for the unemployment rate, UNR, and the dummy variable, DUM1 (1970- 1987=0; 1988-994=1).

^2/See footnote 2 in Table 1.

View Table

Table 3

Long Run Real Wage Equation ^1/

Johansen Maximum Likelihood Procedure: lag in VAR = 8

92 Observations from 1972-Q1 to 1994-Q4

Null 2/	Alternative 2/	Statistic	95% Critical Value		90% Critical Value
Cointegration LR Test Based on Maximal Eigenvalue of the Stochastic Matrix
r = 0	r = 1	30.2914	27.0670		24.7340
r<= 1	r = 2	13.1140	20.9670		18.5980
Cointegration LR Test Based on Trace of the Stochastic Matrix
r = 0	r > = 1	52.7715	47.2100		43.9490
r<= 1	r > = 2	22.4800	29.6800		26.7850
Estimated Cointegrated Vectors
		Vector 1		OLS (t-ratio)
RW		-1.000
RPP		0.139		-0.029 (0.24)
UNR		0.241		0.098 (2.7)
DUM1		0.745		0.561 (7.4)
Intercept				11.122 (7.6)

^1/All variables are in logarithm except for the unemployment rate, UNR, and the dummy variable, DUM1 (1970- 1987=0; 1988-994=1).

^2/See footnote 2 in Table 1.

C. The Inflation Equation

A general specification of the short-run inflation equation (5) is estimated using ordinary least squares (OLS) and instrumental variables (IV) for the sample period 1972 to 1995.¹⁰ Table 4 reports the estimated equation obtained after eliminating variables with t-ratios less than one. The inflation equation includes (i) the first and third lags of inflation; (ii) the contemporaneous and one-period lagged change in the nominal effective exchange rate; (iii) the second lag of the change in money; (iv) the first lag of wage inflation; (v) the fourth lag of the change in export prices; and (vi) all the market disequilibrium terms, except that for the labor market.¹¹ The estimated equation explains 77 percent of the variation in inflation and tracks the actual inflation developments quite well (Figure 5, top panel). The results obtained using IV estimation do not differ significantly from those obtained using OLS, and all diagnostic tests from both estimation procedures are generally satisfactory. However, Chow tests indicate significant structural breaks in the inflation equation between 1980 and 1983, in line with the substantial structural changes that the Turkish economy underwent in that period.

View Full Size

Turkey: Actual and Fitted Inflation, 1972-94

Citation: IMF Working Papers 1997, 170; 10.5089/9781451858389.001.A001

Source: As estimated from equation (5).

• Download Figure
• Download figure as PowerPoint slide

Table 4

Inflation Equation

Dependent variable is AP

t-ratios in parenthesis.

Table 4

Inflation Equation

Dependent variable is AP

	Ordinary least squares (OLS) 1973-94	Instrumental variables (IV) 1973-94
Regressor	Coefficient	Coefficient
Intercept	-0.09 (2.5)	-0.11 (2.7)
ΔP(-l)	0.32 (3.2)	0.31 (3.1)
ΔP(-3)	0.20 (2.9)	0.19 (2.8)
ΔE	-0.52 (13.2)	-0.55 (10.4)
ΔE(-l)	0.14 (2.2)	0.14 (2.2)
ΔM(-2)	0.12 (2.1)	0.12 (2.1)
ΔW(-1)	0.05 (1.8)	0.05 (1.7)
ECM(-1)	-0.13 (2.9)	-0.15 (3.1)
De(-1)	-0.03 (1.4)	-0.04 (1.6)
Df(-3)	0.20 (1.4)	0.31 (1.5)
ΔPx(-4)	0.08 (1.6)	0.09 (1.7)
${\overline{R}}^2$	0.77	0.77
S.E. regression	0.03	0.03
Sargan’s x2(4)		0.63
	Diagnostic tests
Serial Correlation, χ2(4)	3.05	2.93
Normality, χ2(2)	5.90	8.76
Heteroscedasticity, χ2(1)	0.78	0.02
Chow Test	X2(ll)=34.33	F(ll,67)=3.12

t-ratios in parenthesis.

View Table

Table 4

Inflation Equation

Dependent variable is AP

	Ordinary least squares (OLS) 1973-94	Instrumental variables (IV) 1973-94
Regressor	Coefficient	Coefficient
Intercept	-0.09 (2.5)	-0.11 (2.7)
ΔP(-l)	0.32 (3.2)	0.31 (3.1)
ΔP(-3)	0.20 (2.9)	0.19 (2.8)
ΔE	-0.52 (13.2)	-0.55 (10.4)
ΔE(-l)	0.14 (2.2)	0.14 (2.2)
ΔM(-2)	0.12 (2.1)	0.12 (2.1)
ΔW(-1)	0.05 (1.8)	0.05 (1.7)
ECM(-1)	-0.13 (2.9)	-0.15 (3.1)
De(-1)	-0.03 (1.4)	-0.04 (1.6)
Df(-3)	0.20 (1.4)	0.31 (1.5)
ΔPx(-4)	0.08 (1.6)	0.09 (1.7)
${\overline{R}}^2$	0.77	0.77
S.E. regression	0.03	0.03
Sargan’s x2(4)		0.63
	Diagnostic tests
Serial Correlation, χ2(4)	3.05	2.93
Normality, χ2(2)	5.90	8.76
Heteroscedasticity, χ2(1)	0.78	0.02
Chow Test	X2(ll)=34.33	F(ll,67)=3.12

t-ratios in parenthesis.

To allow for these structural shifts, the inflation equation was re-estimated for the period 1970–1980 and 1981–1995.¹² The final results obtained are reported in Table 5 and described below; the estimated inflation equations account for up to 74 percent of the variation in inflation in the 1970s, and as much as 86 percent during the 1980s and 1990s (Figure 5, middle and bottom panels). All diagnostic tests are satisfactory. Figure 6 plots the actual and dynamic forecasts of inflation for both periods.

View Full Size

Turkey: Actual Inflation and Dynamic Forecasts, 1971-94

Citation: IMF Working Papers 1997, 170; 10.5089/9781451858389.001.A001

Source: As estimated from equation (5).

• Download Figure
• Download figure as PowerPoint slide

Table 5.

Inflation Equation

Dependent variable is ΔP

t-ratios in parenthesis

Table 5.

Inflation Equation

Dependent variable is ΔP

	Ordinary least squares (OLS) 1971 to 1980 1981 to 1994		Instrumental variables (IV) 1971 to 1980 1981 to 1994
Regressor	Coefficient	Coefficient	Coefficient	Coefficient
Intercept	-0.15 (1.8)	-0.09 (2.1)	-0.20 (2-1)	-0.09 (2.2)
ΔP(-l)		0.33 (2.8)		0.35 (2.8)
ΔP(-3)	0.33 (2.8)	0.12 (1.9)	0.34 (2.9)	0.12 (1.9)
ΔE	-0.46 (13.4)	-0.57 (14.9)	-0.54 (5.3)	-0.59 (11.8)
ΔE(-l)		0.20 (2.7)		0.22 (2.8)
ΔM(-2)	0.51 (4.1)		0.52 (4.1)
ΔW(-1)	0.29 (2.4)		0.32 (2.5)
ECM(-l)	-0.16 (1.7)	-0.15 (2.7)	-0.23 (2.0)	-0.15 (2.7)
De(-1)		-0.11 (3.5)		-0.11 (3.4)
Df(-3)		0.32 (2.2)		0.30 (2.0)
${\overline{R}}^2$	0.75	0.86	0.74	0.86
S.E. regression	0.03	0.02	0.03	0.02
Sargan’s statistic			χ2(6)=2.9	χ2(8)=2.8
	Diagnostic tests
Serial Correlation, χ2(4)	1.65	3.92	0.75	4.30
Normality, χ2(2)	3.00	1.21	5.32	1.24
Heteroscedasticity, χ2(1)	0.24	0.00	0.08	0.32

t-ratios in parenthesis

View Table

Table 5.

Inflation Equation

Dependent variable is ΔP

	Ordinary least squares (OLS) 1971 to 1980 1981 to 1994		Instrumental variables (IV) 1971 to 1980 1981 to 1994
Regressor	Coefficient	Coefficient	Coefficient	Coefficient
Intercept	-0.15 (1.8)	-0.09 (2.1)	-0.20 (2-1)	-0.09 (2.2)
ΔP(-l)		0.33 (2.8)		0.35 (2.8)
ΔP(-3)	0.33 (2.8)	0.12 (1.9)	0.34 (2.9)	0.12 (1.9)
ΔE	-0.46 (13.4)	-0.57 (14.9)	-0.54 (5.3)	-0.59 (11.8)
ΔE(-l)		0.20 (2.7)		0.22 (2.8)
ΔM(-2)	0.51 (4.1)		0.52 (4.1)
ΔW(-1)	0.29 (2.4)		0.32 (2.5)
ECM(-l)	-0.16 (1.7)	-0.15 (2.7)	-0.23 (2.0)	-0.15 (2.7)
De(-1)		-0.11 (3.5)		-0.11 (3.4)
Df(-3)		0.32 (2.2)		0.30 (2.0)
${\overline{R}}^2$	0.75	0.86	0.74	0.86
S.E. regression	0.03	0.02	0.03	0.02
Sargan’s statistic			χ2(6)=2.9	χ2(8)=2.8
	Diagnostic tests
Serial Correlation, χ2(4)	1.65	3.92	0.75	4.30
Normality, χ2(2)	3.00	1.21	5.32	1.24
Heteroscedasticity, χ2(1)	0.24	0.00	0.08	0.32

t-ratios in parenthesis

For the 1970s, the results indicate strong influences on prices from wages, money, and the exchange rate. The independent role for wage growth presumably reflects the sizable influence of trade union activity on wage growth during this period; real wages are estimated to have increased by some 100 percent between 1970 and 1979. Both the third lag of inflation and the error correction term are also significant and correctly signed, suggesting that inflation inertia was already of significance during the 1970s; the estimated coefficient on the error correction term implies an adjustment period to equilibrium of four quarters. The terms representing real exchange rate disequilibrium and the public sector deficit are not statistically significant, implying that these variables did not have an influence on inflation separate from variables that were statistically significant (i.e., money, wages, and the nominal exchange rate).

For the post-1980 period, the results indicate strong influences on price growth from the exchange rate and lagged inflation. Interestingly, the independent role for wage and money growth observed during the 1970s disappears. This likely reflects the more restrictive environment for trade unions (at least through the late 1980s) and the liberalization of external trade and financial relationships, which would have expanded both the direct role of the exchange rate in price determination and the importance of the exchange rate as a channel through which monetary policy influences prices. Up to three lags of inflation are positively signed with both the first and third lag statistically significant at the 5 percent level. Moreover, the estimated coefficient of the error correction term implies an even longer return to equilibrium—almost seven quarters—for the price level compared to the 1970s. In other words, the inflationary process exhibited greater inertia during this later period as inflationary expectations became more entrenched.¹³

The terms representing the exchange rate disequilibrium and the pressure of the public sector deficit are both significant for this period. The significance of the former likely captures the effects of the policy of real exchange rate depreciation pursued by the authorities during much of the 1980s as an export promotion tool, and supports the view that at times the exchange rate has been a key factor in generating, as distinct from merely accommodating, inflation.¹⁴ The significance of the deficit measure supports the public finance view of the inflationary process in Turkey. The fact that money growth does not have a significant separate influence on inflation in this latter period reflects a change in the link between fiscal deficits and inflation because of the increased importance of debt financing over monetary financing.