Analysis of central bank accounts

The intervention of the central bank in the foreign exchange market, together with the government’s financing of expenditures with oil-related receipts implies that the simple approach to determine fiscal dominance using central bank accounts would not work. The link between changes in the monetary base and fiscal policy is not clearly identified: all changes in the monetary base would be seen as originating from changes in the central bank’s net international reserves. Moreover, it is possible to observe money creation of fiscal origin even in the presence of an overall fiscal surplus.

A simple example may illustrate the problem of using this approach in the case of OEC, using the basic monetary base identity, and the simplifying assumption that all oil exports are carried out by state-owned oil companies.

where,

When an oil export occurs, there is a simultaneous increase in NIR (as the oil company sales the foreign exchange proceeds to the central bank) and in the deposits of the nonfinancial public sector at the central bank, that is, a decline in NDA for the same amount, so the monetary base does not change. As the nonfinancial public sector finances expenditures with those deposits, NDA increase and the monetary base expands. However, if those changes occur within the same period, the central bank accounts would show an expansion in the monetary base due to an increase in NIR. Moreover, if the public sector spends less than what it received from oil exports, NDA would reflect a decline, partially offsetting the impact of the increase in NIR. Without further analysis, one may reach to the conclusion that fiscal policy had a contractionary effect on the monetary base.

Fiscal sustainability approach

To use the fiscal sustainability framework as an instrument to determine fiscal dominance in OEC, it is necessary to change the focus from the stock of debt and the overall primary balance of the government to the net worth and its corresponding flow, the net operating primary balance. The reason for that is that the extraction of oil implies a depletion of a non-renewable resource, that is, a reduction of a government asset.⁷ Financing government consumption or unproductive investments with oil-related fiscal receipts reduces the government’s net worth without immediately affecting the stock of debt. Moreover, oil-related fiscal receipts present simply the change of one less liquid asset for another more liquid and should not be considered as fiscal revenue. The standard definition of primary fiscal balance overlooks this fact by treating oil-related receipts as any other revenue.

One way to properly assess the presence of fiscal dominance in OEC is to use a fiscal balance excluding oil-related net receipts and a more comprehensive indicator of the government’s financial position—net worth (NW)—in addition to the stock of debt. The accounting framework explained below, based on the net worth approach to fiscal analysis, can be used to derive the appropriate indicators.⁸ To visualize the separation between fiscal policy and monetary policy, the government and the central bank will first be considered individually and then consolidated.

The government’s budget constraint and net worth

NOBg:	Net operating balance of the government, defined as revenue minus expense (transactions that increase net worth minus transactions that reduce net worth).
$I_{\mathit{\text{NF}}}^{\mathit{\text{no}}}$, $I_{\mathit{\text{NF}}}^o$:	Net investment in non-oil and oil nonfinancial assets, respectively.
$A_{\mathit{\text{NFg}}}^{\mathit{\text{no}}}$, $A_{\mathit{\text{NFg}}}^o$:	Non-oil nonfinancial assets of the government and oil wealth, respectively.
$A_{\mathit{\text{Fg}}}^f$, $L_g^f$:	Government’s foreign assets and liabilities.
$A_{\mathit{\text{Fg}}}^d$, $L_g^d$:	Government’s domestic financial assets and liabilities, respectively, excluding those with the central bank.
${\mathit{\text{DP}}}_g^{\mathit{\text{cb}}}$, $L_g^{\mathit{\text{cb}}}$:	Government’s deposits at the central bank and liabilities to the central bank.

View Table

NOBg:	Net operating balance of the government, defined as revenue minus expense (transactions that increase net worth minus transactions that reduce net worth).
$I_{\mathit{\text{NF}}}^{\mathit{\text{no}}}$, $I_{\mathit{\text{NF}}}^o$:	Net investment in non-oil and oil nonfinancial assets, respectively.
$A_{\mathit{\text{NFg}}}^{\mathit{\text{no}}}$, $A_{\mathit{\text{NFg}}}^o$:	Non-oil nonfinancial assets of the government and oil wealth, respectively.
$A_{\mathit{\text{Fg}}}^f$, $L_g^f$:	Government’s foreign assets and liabilities.
$A_{\mathit{\text{Fg}}}^d$, $L_g^d$:	Government’s domestic financial assets and liabilities, respectively, excluding those with the central bank.
${\mathit{\text{DP}}}_g^{\mathit{\text{cb}}}$, $L_g^{\mathit{\text{cb}}}$:	Government’s deposits at the central bank and liabilities to the central bank.

Equation (1) represents the government’s budget constraint: the left hand side corresponds to the traditional definition of overall fiscal balance, while the right hand side represents the financing. Equation (2) defines the government’s net worth as the sum of nonfinancial assets, non-oil and oil, financial assets, and debt. As indicated by equation (3), assuming no valuation adjustments, the change in net worth will be determined solely by the net operating balance, defined as revenue minus expense; since in the net worth framework any receipt related to the sale of assets is not considered revenue but a change of one asset for another, oil-related receipts are excluded. From equation (2) it is clear that even if the government’s overall balance is zero (which may also imply no change in debt), its net worth position might be deteriorating if oil wealth is being used to finance a negative NOB (for example, to finance government consumption).

The equations above can be used to analyze fiscal sustainability and thus, fiscal dominance in cases where the government is financing consumption or unproductive investments with assets instead of debt. An unsustainable fiscal situation could be identified with a path for the net operating balance that implies a continuous deterioration of the government’s net worth.⁹ This framework is suitable for oil exporting countries for which the stock of debt might not fully detect the extent of the problem, particularly during the oil extraction period. After all oil reserves have been depleted, the unsustainable path of the net operating balance would start affecting the stock of debt. The stronger the degree of oil dominance in the fiscal accounts, the more difficult it will be to introduce the fiscal adjustment required to compensate for the loss of oil financing. At that point, the problem becomes the standard case analyzed using the debt sustainability framework.

The central bank’s budget constraint and net worth

where,

NOBcb:	Net operating balance of the central bank.
$I_{\mathit{\text{NF}}}^{\mathit{\text{no}}}$:	Net investment in non-oil nonfinancial assets.
AfFcb, AdFcb:	Financial assets of the central bank measured in local currency, foreign and domestic, respectively, excluding credit to the government.
$A_{\mathit{\text{Fcb}}}^g$:	Central bank credit to the government (equal to $L_g^{\mathit{\text{cb}}}$ in equation (2)).
AnoNFcb:	Non-oil nonfinancial assets of the central bank.
Lfcb, Ldcb:	Central bank’s liabilities measured in local currency, foreign and domestic, respectively. Includes debt and non-debt foreign liabilities and any other liability not included in the concept of monetary base. Excludes deposits of the government at the central bank.
${\mathit{\text{DP}}}_g^{\mathit{\text{cb}}}$:	Deposits of the government at the central bank.
MB:	Monetary base. By definition it includes: currency in circulation and deposits of banks and the nonfinancial private sector at the central bank.

View Table

NOBcb:	Net operating balance of the central bank.
$I_{\mathit{\text{NF}}}^{\mathit{\text{no}}}$:	Net investment in non-oil nonfinancial assets.
AfFcb, AdFcb:	Financial assets of the central bank measured in local currency, foreign and domestic, respectively, excluding credit to the government.
$A_{\mathit{\text{Fcb}}}^g$:	Central bank credit to the government (equal to $L_g^{\mathit{\text{cb}}}$ in equation (2)).
AnoNFcb:	Non-oil nonfinancial assets of the central bank.
Lfcb, Ldcb:	Central bank’s liabilities measured in local currency, foreign and domestic, respectively. Includes debt and non-debt foreign liabilities and any other liability not included in the concept of monetary base. Excludes deposits of the government at the central bank.
${\mathit{\text{DP}}}_g^{\mathit{\text{cb}}}$:	Deposits of the government at the central bank.
MB:	Monetary base. By definition it includes: currency in circulation and deposits of banks and the nonfinancial private sector at the central bank.

Equation (1’) represents the central bank’s budget constraint: the left hand side corresponds to the overall balance and the right hand side represents the financing. Equation (2’) shows the central bank’s net worth and its components: nonfinancial and financial assets and liabilities.

Rearranging the terms of equation (2’), and expressing the variables in terms of period changes, it is possible to obtain the accounting identity for the change in the monetary base:

where L ^ng_cb includes central bank instruments of monetary control held by banks or the nonfinancial private sector, and OL_cb, other liabilities not included elsewhere. The first term of the right hand side of equation (2”) corresponds to the change in net foreign assets of the central bank, which for simplification are assumed to be equivalent to the net international reserves; the second term represents the change in the net domestic credit of the central bank, defined as credit or holdings of government bonds by the central bank, net of government deposits, credit to banks and nonbank private sector, net of holdings of central bank instruments of monetary control, and central bank’s net worth.

Equation (2”) can be used to analyze the impact of oil-related fiscal receipts on the monetary base. Assuming a fixed exchange rate regime (or foreign exchange surrender requirements), if the government accumulates all oil-related receipts, the reduction in net domestic assets (due to the increase in government deposits) will compensate the increase in the central bank’s foreign assets, without any monetary impact. The monetary base will expand only if the government uses the oil-related receipts to finance domestic expenditures: there will be a decline in government deposits (DP^g_cb) and a simultaneous increase in the monetary base (MB).

The consolidated public sector’s budget constraint and net worth

Adding equations (1) and (1’), and (2) and (2”), and rearranging terms, it is possible to obtain the budget constraint and the net worth equations for the consolidated public sector shown below:

The left hand side of equation (4) represents the overall balance of the consolidated public sector, and the right hand side the financing. Assuming for simplification that all oil sales are external, when the government receives the proceeds from oil sales or from collecting oil taxes, there is a reduction in oil wealth (a negative investment in oil, $I_{{\mathit{\text{NF}}}_t}^o$), compensated by an increase in foreign assets (A^f_F). As shown in equation (4), at that point there is no change in public sector wealth, and no monetary impact. However, if subsequently the government finances consumption with oil-related receipts—a negative NOB—there will be a reduction in net worth that can be compensated by an increase in the monetary base and/or a decline in foreign assets.

Equation (4) also highlights the fact that there is little that the central bank can do to avoid the reduction in foreign assets or the increase in the monetary base, say through open market operations. A continuous increase in the monetary base may cause inflationary pressures, while continuous placements of central bank bonds or government bonds generate central bank losses (since usually the return of international reserves is lower than the interest rate on central bank bonds or the forgone return of government bonds). Under those circumstances, the central bank looses the ability to effectively use monetary policy to achieve price stability: fiscal policy dictates monetary policy actions, which implies fiscal dominance as defined in this paper.

The framework presented above seems to be the appropriate one to determine fiscal sustainability in the case of oil exporting countries, which in turn helps to identify the presence of fiscal dominance: a declining net worth would indicate a possible situation of fiscal dominance. Using a fiscal sustainability framework based on net worth however, may be subject to some data problems and uncertainties associated not only to the valuation and volume of the stock of oil wealth, but also to the valuation of nonfinancial assets in general. In addition, while using oil wealth to finance investment might not affect the net worth position of the government in the short-run, it may still have a monetary impact.¹⁰ An alternative and much simpler approach to directly address the problem of the monetary impact is discussed below.

Oil dominance/fiscal dominance: an alternative indicator

The presence of oil dominance/fiscal dominance can be determined using a simple indicator that tries to capture the monetary impact of fiscal policy: the ratio of the domestic fiscal balance to oil-related net receipts.¹¹ The domestic fiscal balance (OB^d) is defined as the difference between revenues from domestic sources minus domestic expenditures.

Equation (4) can be reformulated to capture the relationship between the domestic fiscal balance and the change in the monetary base as follows:

Where G includes total expenditures (total expense and net investment in non-oil nonfinancial assets), R^no refers to non oil-related revenue, and R^o represents oil-related receipts (equivalent to −$I_{\mathit{\text{NF}}}^o$ as noted above, assuming that gross investment in oil assets is equal to zero). The left hand side of equation (4’) represents the deficit related to domestic transactions: if it is zero (or negative), oil wealth is being transformed into a financial asset, for example, net foreign assets (international reserves); if it is positive, oil-related receipts are financing part or all of the domestic deficit, causing an increase in the monetary base, unless the central bank intervenes either by selling foreign assets, or domestic debt. Thus, it is possible to establish the following relation:

Where γ is equal to the ratio domestic deficit to oil-related receipts. Since R^o is always positive, a positive γ would indicate that a monetary expansion of oil origin initially took place. As noted above, that monetary expansion could have been offset by a fall in net international reserves or through central bank open market operations (to mop up the excess liquidity), an unsustainable monetary policy when continuously implemented for long periods of time. Therefore, γ could be used as an indicator of oil dominance/fiscal dominance: the larger γ, the higher the possibility of oil dominance/fiscal dominance. The approach followed in section IV to analyze oil dominance/fiscal dominance in the case of Venezuela is based on the same framework, but considers a more comprehensive definition of the domestic fiscal balance.

1. Testing the oil dominance/fiscal dominance hypothesis.

The general VAR representation is as follows:

where y_t and x_t are the vectors of endogenous and exogenous variables, respectively, A₁,…., A_p and B are the matrices of the coefficients to be estimated, and ε_t is the vector of innovations.

If all variables included in y_t result non-stationary I(1) variables, and a cointegration relationship among them is found, it is possible to rewrite equation (7) as:

Equation (8) represents the Vector Error Correction (VECM), where ${\stackrel{\wedge }{e}}_{t-1}$ embodies the deviation from long-run equilibrium in period (t-1)¹⁹.

2. Testing the hypothesis of a relationship between the monetary base and prices.

If the transmission mechanism described above holds, in addition to finding a relationship between oil, fiscal and monetary variables, a close relationship between prices and money also should be found, both in the long-run and short-run. Again the VAR-VECM methodology is used to test for this relationship.

The details of the two sets of empirical analysis are described below.

B.1 Testing the oil dominance—fiscal dominance hypothesis

Model 1: variables in nominal terms

To test for the link between the domestic primary deficit and the monetary base, a first model using the variables in nominal terms is estimated. All variables are measured in log-levels and are identified as follow:

LPON:	Nominal oil prices (average of the Venezuela’s oil basket)- exogenous;
LDDPN:	Nominal domestic primary deficit;
LNEXEP:	Nominal exchange rate (official rate Bs./US$; end of period);
LR:	Stock of nominal net international reserves (in U.S. dollars);
LMB:	Nominal monetary base.

From a long-run perspective, the finding of a cointegrating relationship between the monetary base (LMB) and the domestic primary deficit (LDDPN) is supportive of the OD/FD hypothesis.

Since all nominal variables in levels contain a unit root (as indicated by the augmented Dickey-Fuller (ADF) and the Phillips-Perron (PP) tests) the Johansen approach to test for cointegration was applied. The most satisfactory result obtained was from the application of the Johansen test to LMB and LDDPN which indicates the presence of a cointegrating vector (normalized) in which the coefficient of LDDPN exhibits a statistically significant and negative sign as expected, and with a value close to one:²¹

$\text{COINT}1{A}=\text{LMB}-{1.075578}^{*}{LDDPN}$

As in the case of cointegration for the long-run, I-R functions, variance decompositions and Granger-causality tests lend support to the OD/FD hypothesis in the short-run.

With the previous result, a VECM with one lag is estimated including as endogenous variables the first difference of the nominal variables (DLDDPN, DLMB, DLR, DLNEXEP) and as exogenous variables the residuals from the cointegrating vector lagged one period [COINT1A(-1)], and the first difference of the logarithm of the nominal oil price lagged one period [DLPON(-1)]. The validity of the OD/FD hypothesis in the short-run is evaluated by first studying the accumulated I-R functions and VD derived from the VECM model, using the following order for the Cholesky decomposition: DLDDPN, DLMB, DLR, and DLNEXEP. For the complete sample period (1960–2005) the I-R functions and VD indicate the following:

• A one standard deviation shock to DLDDPN has a positive effect on DLMB that explains around 17 percent of its forecast error variance in a twenty-period horizon.

• The response of DLDDPN to shocks to DLMB is positive, but very close to zero according to the variance decomposition analysis.

The block Granger-causality test indicates one way Granger-causality running from the nominal domestic primary deficit to the monetary base. Although within the VECM there is not causality in either way between DLDDPN and DLMB, the coefficient of the lagged value of the residuals of the cointegrating vector [COINT1A(-1)] is significantly different from zero and has the expected (negative) sign in the DLMB equation, while it has the expected (positive) sign, but it is not statistically significant (at a 10 percent level of significance) in the DLDDPN equation.

Another interesting result from the VECM that reinforces the validity of the OD/FD hypothesis, is that there seems to be a connection between oil prices, the primary domestic fiscal deficit and the monetary base. In fact, the coefficient of DLPON (-1) is positive and statistically different from zero in the DLDDPN equation (at a 1 percent level of significance) and in the DLR equation (at a 10 percent level of significance).

Accumulated Response to Cholesky One S.D. Innovations ± 2 S. E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Accumulated Response to Cholesky One S.D. Innovations ± 2 S. E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

Accumulated Response to Cholesky One S.D. Innovations ± 2 S. E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

Variance Decomposition ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Variance Decomposition ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

Variance Decomposition ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

Analysis for two sub periods

Although the results from the divided sample should be interpreted with care due to the reduction in the degrees of freedom available for estimation, the separation produces I-R functions and VD that suggest that the OD/FD hypothesis is valid in the two periods.

A comparison of the I-R functions and VD for the two sub-periods gives the following main results:

• In both periods shocks to DLDDPN have a positive effect on DLMB, but the response is substantially different. In the soft-peg period shocks to DLDDPN explain approximately 60 percent of the forecast error variance of DLMB, while in the variable period they explain around 20 percent.

• In the soft-peg period DLDDPN responds positively but mildly to shocks to DLMB: shocks to DLMB explain close to 5 percent of DLDDPN forecast error variance. In the variable-peg period the response of DLDDPN to shocks to DLMB is negative, but close to zero according to the variance decomposition analysis.

The examination of other results from the VECM, however, shows some puzzling results for the soft-peg period: there seems to be Granger-causality running from the monetary base to the domestic primary deficit, but not from the domestic primary deficit to the monetary base. In fact, the coefficient of COINT1A(-1) is positive (as expected) and statistically different from zero in the DLDDPN equation. However, in the DLMB equation this coefficient is not statistically significant. Another result for this period is that the coefficient of DLPON(-1) is not longer statistically significant in the DLDDPN and DLR equations.

In contrast, the results of the variable-peg period are more in line with those for the complete sample (although shocks to DLDDPN have a smaller impact on DLMB in this period compare to the soft-peg): there seems to be Granger-causality from the domestic primary deficit to the monetary base and not the opposite. In the variable-peg period the coefficient of COINT1A (-1) is negative and significantly different from zero in the DLMB equation. This coefficient is positive, but not statistically significant in the DLDDPN equation. Additionally, the coefficient of DLPON (-1) continues to be positive and statistically different from zero in the DLDDPN equation, but no in the DLR equation.

In both periods the block Granger-causality test does not allow to reject the null hypothesis of no causality in either direction between DLDDPN and DLMB (at standards levels of significance).

Model 2: variables measured as deviations from their trend values

In this VAR model nominal variables are measured as deviations from their Hodrick-Prescott (H-P) trend values. Both the augmented Dickey-Fuller and Phillips-Perron unit-root tests suggest that all variables are I(0). The variables are defined as follows:

LPONG:	nominal oil prices gap - exogenous;
LDDPNG:	nominal domestic primary deficit gap;
LNEXEPG:	nominal exchange rate gap;
LRG:	nominal net international reserves gap;
LMBG:	nominal monetary base gap.

The results based on the I-R functions and VD analysis with the ordering LDDPNG, LMG, LRG LNEXPG give a rather weak support to the OD/FD hypothesis, both for the whole sample as well as for the two sub-periods, particularly when compared to model 1:

• A shock to LMBG has a negative impact on LDDPNG.

• The VD analysis indicates that the percentage of the forecast error variance of LDDPNG explained by LMBG is close to zero (approximately 1.4 percent in a 20-period horizon).

• LMBG varies positively to shocks to LDDPNG.

• The VD analysis indicates that the percentage of the forecast error variance of LMBG explained by LDDPNG is very low (approximately 8.7 percent in a 20-period horizon).

• There is no Granger causality in either direction between LDDPNG and LMBG.

• Similar to the VECM with nominal variables, we observe that the coefficient of LPONG in the equations for LDDPNG, LRG and LMBG is positive and statistical significant.

Accumulated Response to Cholesky One S.D. Innovations ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Accumulated Response to Cholesky One S.D. Innovations ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

Accumulated Response to Cholesky One S.D. Innovations ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

Variance Decomposition ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Variance Decomposition ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

Variance Decomposition ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

Model 3: variables measured as percentage of GDP or in real terms

In this case, the VAR model is constructed with all variables measured as percentage of GDP or in real terms. The augmented Dickey-Fuller and Phillips-Perron unit-root tests suggest that all variables are I(0). The variables are defined as follow:

DLPOR:	first difference of the logarithm of real oil price―exogenous;
VDDPYN:	change in the nominal domestic primary deficit as a percentage of GDP;
DLREXEP:	first difference of the logarithm of the real exchange rate;
DLRREAL:	first difference of the logarithm of the real net international reserves
SM:	seigniorage measured as the change in the nominal monetary base as a percentage of GDP.

This model also produces a relatively weak support of the OD/FD hypothesis compared to model 1, but the results are more robust than in model 2. The results for the two subperiods are similar to those for the whole sample. Based on the impulse-response functions and Variance Decomposition analysis with the ordering VDDPYN, SM, DLRREAL, DLREXEP, it is possible to observe the following:

• VDDPYN response to shocks to SM is positive but very small.

• The VD analysis indicates that the percentage of the forecast error variance of VDDPYN explained by SM is close to zero.

• SM exhibits a positive reaction to shocks to VDDPYN.

• The VD analysis indicates that the percentage of the forecast error variance of SM explained by VDDPYN is very low (approximately 9.4 percent in a 20-period horizon).

• The block Granger-causality test indicates that SM does non Granger cause VDDPYN, but the latter Granger cause SM (P-value = 0.072).

• As in the case of the VECM with nominal variables and the VAR model with the nominal variables measured as deviations from the trend, a positive and statistical significant coefficient of DLPOR was found in the equations for VDDPYN, DLRREAL and SM.

Accumulated Response to Cholesky One S.D. Innovations ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Accumulated Response to Cholesky One S.D. Innovations ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

Accumulated Response to Cholesky One S.D. Innovations ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

Variance Decomposition ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Variance Decomposition ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

Variance Decomposition ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

B.2. Testing the hypothesis of a relationship between the monetary base and prices

The general conclusion that can be drawn from the results is that there seems to be a close relationship between prices and money, both in the long-run and short-run, for the 1960–2005 period. Overall, the results are in line with those reported in Olivo and Miller (2000) and Olivo (2002).

Given that the logarithm of the monetary base and the price level (measured by the CPI) are I(1) according to the ADF and PP tests, the test for cointegration between them is performed. The Johansen test indicates the existence of one cointegrating equation, but a visual inspection of the residuals from this equation suggests that they are not stationary. To capture the effects of a possible regime shift, the Gregory and Hansen (G&H, 1996) cointegration tests are applied, following Olivo and Miller (2000) and Olivo (2002).

The three tests proposed by G&H allow to reject the null hypothesis of no cointegration at the 1 percent level of significance with a structural brake in 1968. However, the level shift model (model C in G&H notation) seems to produce the most robust results. This model generates the following cointegrating equation (t-statistics in parenthesis):

$\text{LCP}=\underset{(-10.72852)}{-1.199748}+\underset{(47.26916)}{{0.972699}^{*}}\text{LMB}-\underset{(-7.12508)}{{1.058871}^{*}}\text{DU}68$

In the above equation DU68 is the dummy variable that captures the level shift in the cointegrating vector, taking values equal to one from 1969 on. Notice that in the cointegrating equation the coefficient of LMB is close to one. Therefore, taking into account the possibility of a structural break in 1968, there is evidence of a long-run equilibrium relationship between the monetary base and the price level.

Given the presence of cointegration between LCP and LMB, a five lag VECM is estimated, with DLMB and DLCP as endogenous variables and the residuals from the G&H cointegrating vector lagged one period [COINN2(-1)] as an exogenous variable.

The I-R functions and VD derived from that model for the period 1960–2005, show the following results:

• A one standard deviation shock to DLMB produces a positive response in DLCP, explaining approximately 20 percent of its forecast error variance.

• A one standard deviation shock to DLCP affects positively DLMB, explaining around 33 percent of its forecast error variance.

Accumulated Response to Cholesky One S.D. Innovations ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Accumulated Response to Cholesky One S.D. Innovations ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

Accumulated Response to Cholesky One S.D. Innovations ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

Variance Decomposition ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Variance Decomposition ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

Variance Decomposition ± 2 S.E.

Citation: IMF Working Papers 2008, 142; 10.5089/9781451870008.001.A001

• Download Figure
• Download figure as PowerPoint slide

Consistent with these results, the block Granger-causality test indicates a two-way causality between DLCP and DLMB.

The coefficient of COINN2(-1) has the expected sign in both equations of the VECM, but is only statistically significant in the DLMB equation, implying Granger-causality from DLCP to DLMB.

These results are interpreted as evidence of an accommodative monetary policy, which in turn may be the consequence of the OD/FD problem that affects the implementation of an independent monetary policy in Venezuela.

Dividing the sample between the soft-peg and variable-peg periods generates the following results for the I-R functions and VD:

• In both periods shocks to DLCP produce a positive response of DLMB, but this response is stronger in the variable-peg period when the forecast error variance of DLMB explained by shocks to DLCP rises to 40 percent and then declines to 31 percent in a ten-period horizon. In the soft-peg period the comparable number increases gradually up to 21 percent.

• In both periods shocks to DLMB have a positive impact on DLCP, but the effect is stronger in the variable-peg period. The VD show that a shock to DLMB explains up to 15 percent of the forecast error variance of DLCP in the soft-peg period and up to 29 percent in the variable-peg period.

• Turning to the block Granger-causality tests, they indicate the absence of causality in either direction for the soft-peg period and bidirectional causality in the variable-peg period. However, for the soft-peg period the coefficient of COINN2(-1) has the expected positive sign and is statistically different from zero (at a 10 percent level of significance) in the DLMB equation, which implies Granger-causality from DLCP to DLMB. In the DLCP equation the coefficient of COINN2(-1) has the expected negative sign, but is not statistically different from zero, indicating no Granger-causality from DLMB to DLCP. In the variable-peg period it was not possible to estimate a stable VECM model including COINN2(-1), thus a simple 5-lag VAR was estimated for this period.