Official forecasts and management of oil windfalls

Abstract

Official forecasts for oil revenues and the burden of pensioners are used to estimate forward-looking fiscal policy rules for Norway and compared with permanent-income and bird-in-hand rules. The results suggest that fiscal reactions have been partial forward-looking with respect to the rising pension bill, but backward-looking with respect to oil and gas revenues. Solvency of the government finances might be an issue with the fiscal rules estimated from historical data. Simulation suggests that declining oil and gas revenue and the costs of a rapidly graying population will substantially deteriorate the net government asset position by 2060 unless fiscal policy becomes more prudent or current pension and fiscal reforms are successful.

Appendices

Appendix 1: Official forecasts, permanent values and description of other data

The stock of oil at time t, S _t , must equal the sum of current and future extractions, R _s , s≥t, that is, $S_{t}=\sum^{\infty}_{i=0}R_{t+i}$. The discounted value of oil revenue is $V_{t}=\sum^{\infty}_{i=0}\frac{P_{t+i}R_{t+i}}{(1+r)^{i}}$. We calculate permanent income using only available information. With only projections for income in year t+1, t+2 and t+3, we have

$$N_t^p= \frac{N_t+\frac{N_{t+1}}{1+r}+\frac{N_{t+2}}{(1+r)^2}+\frac{N_{t+3}}{(1+r)^3}}{1+\frac{1}{1+r}+\frac{1}{(1+r)^2}+\frac{1}{(1+r)^3}}. $$

An alternative is to suppose that oil revenue stays constant from t+3, which yields

Another alternative is to suppose iso-elastic demand and zero extraction costs:

$X^{e}_{t+T}$ is the Ministry of Finance’s forecast of X _t+T at time t. X _t equals oil income, oil price, oil production, dependency ratio or GDP. T is the year farthest into the future for which a forecast was given. We base our estimates on the first approach, because this seems to be most realistic. The qualitative nature of our estimates does not vary much if either of the two alternatives is used.

Variables measuring permanent values: Permanent values are calculated by the information approach, i.e., we use only published expectations from the authorities and a 2 % discount rate. Projections of oil production and reserves are comparable over time, since they are given in volumes and we convert all to standard cubic meters oil equivalents (Sm3 oil equivalents). Oil and gas production value is oil and gas production in oil equivalents multiplied with the oil price. Oil price and production value projections are recalculated to 2007-NOK for consistent comparison over time, and measured as shares of GDP in 2007-NOK. The dependency ratio of interest to the Ministry of Finance has changed over time. The lower age of the labor force has increased over time, while the pension age has varied between 65 and 70. We calculate the growth rate of the predicted dependency ratio at the time and apply this growth rate to the current dependency ratio convention of population aged 67 or more relative to population aged 20–66. We use the actual 67+/20–66 ratio for the year the projection was published as the start of each projection. For GDP, the Ministry’s projections focus on growth in real GDP and we apply its projected growth rates to a starting point set by GDP in 2007 prices.

Current oil and gas revenue received by the state is as reported by the Ministry of Oil and Energy, and measured as described above. The current dependency ratio is the population aged 67 and higher divided by the population aged 20–66. Debt (i.e., net debt excluding the petroleum fund, denoted d ^g) and the Fund (the state’s pension fund abroad—previously called the petroleum fund, denoted f) are measured in current NOK as shares of GDP. The output gap is the logarithmic deviation from GDP trend, which is calculated from the Hodrik–Prescott filter of GDP in 2000 NOK with the smoothing parameter set to 1600 (the standard choice of the Norwegian Ministry of Finance for annual data). The Norwegian CPI serves as deflator to measure variables in 2007-NOK.

Up to 2005 we use long-term budgets (Langtidsprogram) as our source for the Ministry’s expectations. The first long-term budget was published in 1953 with budgeting for the succeeding four years (1954–57). The practice of a new long-term budget every fourth year was maintained up to 2001, covering 2002–05. Since then long-term budgeting has been replaced by long-run perspectives (Perspektivmelding 2004 and 2009). We have supplemented the long-term budgets and perspectives with detailed information from three parliamentary documents that explicitly address oil and gas issues (Stortingsmelding 25 1973–74, Stortingsmeldig 30 1973–74 and Tempomeldingen NOU 1988:27). The budget documents up to 1998–2001 were from the Library of Statistics Norway. We focus on the fiscal reaction functions for the central government, which is the receiver of public oil rents.

Table 5 shows data definitions and sources.

Table 5 Variable definitions and data sources

1.1 A.1 Background data and calculations for Fig. 6

The PIH simulations use forecasts of total transfers and suppose that the other components of government spending are a constant share of GDP. In our regressions, we use the dependency ratio. Our estimated coefficient includes in effect a “price”, which links the dependency ratio to government expenses, in addition to the behavioural effect of government spending on the fiscal stance. Given the predictions of transfers and the dependency ratio (Perspektivmeldingen 2009) and our calculations of their permanent values, we estimate g _t =0.069+0.488p _t and $g_{t}^{p}=0.080+0.458p^{p}_{t}$. Our estimates of the effect of the current and permanent dependency ratio on the non-oil/gas primary deficit suggest a coefficient of about 1 for the current dependency ratio and a coefficient of about −0.9 for the permanent dependency ratio (Harding and Van der Ploeg 2009). Translated into government transfers, our estimates imply that a 1 %-point increase in current government transfers increases the non-oil/gas primary deficit with about 2 %-points. A 1 %-point increase in permanent government transfers decreases the non-oil/gas primary deficit with about 2 %-points. To get values for the predicted value of current and predicted oil production as share of GDP (predicted n and n ^p), we use the series for the predicted current government oil revenue as share of GDP (predicted n) and the coefficients as estimated in the first stage; column (a) upper panel of Table 2 (i.e., in the simulations we set v=n/0.456 and v ^p=n ^p/0.456). We assume that government spending g equals total transfers, the discount factor is 2 %, the return on the fund is 4 %, and that total transfers and oil and gas revenue to the state follow the paths presented by Ministry of Finance in January 2009 (Perspektivmeldingen 2009). The estimates are based on percentages of GDP, whereas the projections of Ministry of Finance on mainland GDP. The oil price used in the projections is about 65 USD per barrel. The projected size of the Fund is based on the series for the Fund-to-Mainland GDP and Mainland GDP-to-GDP presented by the Ministry of Finance in Perspektivmeldingen (2009).

Appendix 2: Emergence of the oil and gas windfall in Norway

The history of Norway as a oil and gas nation started in 1969. Ten years after the Netherlands found gas in Slochteren, the first oil field within the territory of Norway—Ekofisk—was discovered. This was one of the world’s largest offshore oil basins and started production in the summer of 1971. Today there are 57 oil and gas fields in production and Norway is ranked as the fifth largest exporter and the 11th largest producer of oil in the world. Norway was in 2006 the third largest exporter and sixth largest producer of gas. In 2007 the oil- and gas industry constituted 24 and 48 percent of GDP and exports, respectively.¹⁷ The Ministry of Oil and Energy (OED 2008) suggests that about 36 percent of expected total production is currently produced. The peak of oil production was probably passed around the turn of the millennium and the composition of production is tilting away from oil and other liquids towards gas. In 2007 gas contributed 40 percent of production.

Both the time path of oil prices and that of oil and gas production and government revenue have a positive trend over the past four decades. Part of volatility of government income is caused by oil price fluctuations, but the tax regime and government’s direct engagements have also contributed substantially to volatility. The implicit tax rate on oil and gas revenues also has an upward trend.

From the late 1970s to mid 1980s, effective ordinary and special tax rates on value added saw a volatile but gradual rise followed by a sharp fall in the late 1980s and much lower rates during the 1990s. Recent years have seen a sharp rise in these tax rates. Special taxes on oil and gas have since the early 1990s taken over from ordinary taxes in importance. Together they constitute almost 35 percent of value added. The other big chunk of government revenue is net cash flow from the State’s Direct Financial Interest in the gas/oil industry (SDFI). After initial investment outlays of up to 20 percent of value added in the mid 1980s, net return on state holdings is now more than one fifth of value added. Production fees used to be an important source of public revenue, but nowadays are almost gone. Dividends from Statoil have recently tracked (with a short lag) the development of oil prices and have now reached about 3 percent of value added. Environmental taxes rose from zero in 1990 to 2.5 percent in 2000 but are a bit over half percent of value added. Area fees contribute even less to government revenue. Total public income from oil and gas revenues is now about 60 percent of value added, most of it being special and ordinary takes and returns from stake holdings.

Government expectations of future oil and gas revenue have followed current revenue closely. Current values are lower than permanent values in the beginning of the hydrocarbon area, which is consistent with increasing production soon after oil and gas was discovered. The time path of current oil and gas revenues now lies above the permanent path as oil and gas revenues are expected to fall in the future. This should signal a shift from borrowing to saving oil and gas revenues; something the Norwegian government has started doing for some time.

The 1960s and 1970s saw a gradual rise in both primary spending and non-oil/gas taxes. After the onset of oil and gas revenue in the early 1980s, taxes and spending first fell and then increased relentlessly, roughly in line with each other. The non-oil/gas primary deficit (b) has fluctuated around two plateaus, with the level shifting in the late 1970s. In the post war, pre-hydrocarbon period the government ran a surplus of about 3 percent of GDP. In the later period the average deficit has been around 4 percent of GDP. Oil revenue has allowed for running a higher non-oil/gas primary deficit. The key question is whether the higher non-hydrocarbon primary deficits in the hydrocarbon era are sustainable in view of the long-term budgetary commitments of the Norwegian government.

The fiscal rule backed by the Norwegian parliament in Spring 2001 states that the government on average should keep the structurally adjusted, non-oil/gas deficit in year t to 4 percent of the Fund at the end of the previous year. The deficit relative to the fund was close to 4 percent in 2001, so that the rule may be seen as a formalization of going policy at the time. There has been a gradual increase in net government assets excluding the Fund (negative d) and a switch from a small surplus to a somewhat larger deficit for the non-oil/gas primary budget deficit. The two episodes in the 1980s and 1990s with negative output gaps (high unemployment) are associated with large increases in the non-oil/gas primary budget deficit. The Norwegian Fund started in 1990 and has since then rapidly increased to about 90 percent of GDP. The global financial crisis wiped out a big chunk of the Fund.

We focus on the primary non-oil/gas deficit, cleaned for both net capital income and oil revenue and take all lending and borrowing and their associated revenues and costs should be taken into account. In contrast, the non-oil/gas budget deficit used by the Ministry of Finance includes net capital income (excluding those from hydrocarbon-related assets). The structural deficit used by the Ministry of Finance corrects the deficit for business-cycle adjustments.¹⁸ The biggest deviation in our definition of the deficit and that of the Ministry of Finance occurs in 2000, which is most likely due to different treatment of the state’s direct oil engagement (SDFI).¹⁹

The gradual rise in the 67+ dependency ratio in Norway has led to a gradual rise in the need for funding public pension obligations. In the far future, graying of the economy will increase spending needs even further, so that it is sensible to provide for these future needs by having a smaller budget deficit or a surplus. Given the relatively small size of Norway, its sovereign wealth fund is very large.

In response to the oil and gas windfall, the Norwegian government has produced various policy documents and initiatives. In chronological order, they can be summarized as follows.

1973–75: Analytical work by the Ministry of Finance and the Ministry of Industry led to three important documents covering Dutch disease issues, size of reserves, likely lifecycles of fields and environmental concerns. There was not much discussion of long-run spending needs.

1983: The Tempo Committee headed by Hermod Skånland, Central Bank Governor 1985–94, produced its report on “The Future of the Petroleum Activity” (NOU 1983: 27). It recommended the bird-in-hand approach, which says that the government should put its oil and gas revenues in a fund and spend only the real return on the assets accumulated in this fund.²⁰ The Tempo Committee also discussed in detail how such a petroleum fund and spending rule should work in practice. It pointed out the importance of converting oil and gas assets in the ground into financial assets in a fund and of decoupling oil and gas income from spending. Due to political pressures to spend, the Tempo Committee discounted the likelihood of such a Stabilization Fund being implemented and therefore recommended slow extraction of oil and gas as a way to distribute oil and gas wealth to future generations.

1988: The policy Committee headed by Professor Erling Steigum, then NHH Bergen and now at BI Oslo, presented its report “The Norwegian Economy in Change—Prospects for National Wealth and Economic Policy in the 1990s” (NOU 1988: 21). This report suggested that government spending should depend on the permanent income of total oil and gas wealth consisting of the financial fund plus the value of oil and gas reserves in the ground. The calculation of total oil and gas wealth requires the prediction of an optimal depletion path given expected oil/gas prices, technology and interest rates. In contrast to the Tempo Committee, the Steigum Committee did argue for the establishment of a financial hydrocarbon fund. It stressed the importance of regarding such a fund and the value of oil and gas reserves in the ground as part of the same portfolio. It also offered arguments for selling oil before extracting the oil as well as for going short in oil stocks.

2001: The Norwegian government implemented its 4 % BIH rule, which allows for a business-cycle corrected deficit equal to 4 percent of the value of the Fund measured in Norwegian kroner at the end of year t−1. Hence, 4 percent of the value of the Fund at the end of the previous year is allowed to be extracted from the Fund and to be used to fund the general government deficit. The Fiscal Policy Guideline (Handlingsregelen) interpreted the 4 % as the expected future real rate of return of the Government Pension Fund, so non-renewable petroleum wealth had to be invested abroad and transformed into financial wealth (5 % of which in property). The basis sustainability rule was inspired by Hartwick (1977) and starts only the expected future rate on return could be used for domestic consumption purposes.

2006: The Government Pension Fund of Norway comprises two separately managed funds. The main one is the Government Pension Fund Global renamed 1 January 2006 and part of the Norwegian Central Bank (formerly The Government Petroleum Fund established in 1990 and receiving money since 1996). It manages the surplus wealth produced by Norwegian petroleum income (taxes and licenses) and had a value of NOK 2.385 trillion in August 2009. This made it the largest pension fund in Europe and the second largest in the world. Its objective is to counter the decline of expected petroleum income and to smooth the disrupting effects of highly fluctuating oil prices. Since 1998, the fund was allowed to invest up to 40 percent of its assets in the international stock market, but this was increased to 60 percent in 2007. Much of the debate surrounding the fund is on how to contain the risks in investing in the international stock market, ensure ethically sound investments (away from arms and tobacco) and avoid inflation when its return is spent. The other fund is the Government Pension Fund Norway renamed 1 January 2006 (formerly The National Insurance Scheme Fund established in 1967) has value of NOK 106.9 billion at end of 2006.

2006: The Norwegian government undertook reforms to trim pension rights. Pensions are no longer indexed to wage growth but indexed to the average of wage growth and inflation (typically, less). Furthermore, the lifetime value of the pension is a fixed amount calculated around age 60 and is based on expected average life expectancy for the cohort of 60-year olds. The focus is on keeping people in work longer and to retire later whilst allowing for some freedom of choice. The individual can decide whether to work long and enjoy a higher pension pay per retirement year or enjoy more retirement years with a lower pension. The average de facto retirement age (including early retirement and partly disabled) is currently (“pre-reform”) around 59 years, but is expected to rise a little. Compared with most OECD countries, Norway’s pension system is still very generous.

Appendix 3: Econometric robustness checks

3.1 C.1 Instrumenting current and permanent hydrocarbon income with production values

Given that the variables in (9a)–(9b) are I(1) and our fiscal reaction functions will be cointegration relationships between the variables b (or g or t), n, n ^p, p−p ^p, f−d ^g and y, instruments can be used to achieve consistent estimates (Phillips and Hansen 1990).²¹ We use current and permanent oil and gas production value as instruments for current and permanent hydrocarbon income (see the main text for a discussion of exogeneity).

Panel A of Table 6 shows the first-stage IV regression for our econometric time-series model (9a)–(9b) in panel B. Current oil and gas production value predicts current oil and gas revenue with a coefficient of 0.46. For permanent production value, we find a robust positive coefficient of 0.2. Partial R-squared and F-tests indicate that predictive power of the instruments is good.

Table 6 Fiscal responses with permanent oil and gas revenue—IV-estimates

Test statistics of the Augmented Engle–Granger test for cointegration in the first stage are shown in the bottom rows of panel A. We reject a unit root in the residuals from the current oil and gas revenue equation (see statistics in bold), but the evidence against a unit root is weaker for permanent oil and gas revenue. The non-stationarity in the endogenous variables should be driven by the non-stationarity of the exogenous variables through a cointegration relationship for 2SLS to be valid for non-stationary data, so we must interpret our 2SLS results with caution (Hsiao 1997, 2006).²²

Turning to the second stage, the unit root tests of the residuals from the estimated fiscal reaction functions reported in panel B of Table 6 indicate that a unit root cannot be rejected in the residuals. However, the estimates are similar to what we obtain with our other estimators and spurious correlation does not seem to be a big concern.

The IV-estimate of the reaction for the non-oil/gas primary deficit suggests that a third of each extra Krone of oil and gas revenue is spent on increasing the deficit. For permanent revenue, the effect is 0.7, compared to 0.3 under OLS. The IV-estimate of the effect of the current over permanent dependency ratio is the same as the OLS-estimate. For the previous year’s net assets, the coefficient is now 0.08. Comparing with Table 1 in the main text, the OLS bias seems only severe for permanent oil and gas revenue and net assets. Based on signs, magnitudes and statistical significance, the estimates of the fiscal responses presented in Table 6 are thus consistent with the theory put forward in Sect. 2 and the OLS-estimates presented in Table 1.

3.2 C.2 FMOLS and CCR estimates

If regressors are not strictly exogenous, the OLS estimates of the cointegrating relationship do not have an asymptotically efficient distribution, thus invalidating standard tests. As a robustness test we re-estimate in Table 7 the reaction functions with Fully Modified Ordinary Least Squares (FMOLS, Phillips and Hansen 1990) and Canonical Cointegration Regression (CCR, Park 1992). We now also offer additional tests on cointegration next to the Engle–Granger test. As shown in the lower panel of Table 7, the Engle–Granger and Phillips–Ouliaris tests do not reject a unit root for the formulation with oil and gas revenue, but do reject a unit root in the formulation with oil and gas production value (consistent with Tables 1 and 2). The Hansen Parameter Instability and Park Added Variables tests do not reject cointegration for either specification. Not rejecting a unit root in the residuals of Table 1 with the Augmented Engle–Granger tests might thus be due to low power of the tests rather than lack of cointegration.

Table 7 Estimates of non-oil/gas primary deficit (b) with permanent oil and gas revenue

Regardless of the estimation method, current oil and gas revenue/production is quantitatively and statistically more important than their permanent value. As before, using oil and gas revenue rather than the more exogenous oil and gas production decreases the coefficients on the dependency ratio measure and net assets dramatically. We prefer the latter estimates in view that dealing with identification is essential also under non-stationarity and cointegration (Hsiao 2006).

3.3 C.3 Permanent values based on time-series model

The left-hand panel of Fig. 8 shows the two alternative measures of the permanent value of oil and gas revenue—based on official forecasts and a time-series model—together with the series for the current values. The right-hand side panel does the same for the production value. The series based on the time series models under-predict both the official and the current ones, except for in the end of the sample.

Fig. 8

Permanent values from time-series estimates compared to permanent values based on official forecasts. Note: “ts” indicates permanent value series based on estimates of (11) and calculated by (13) in the main text. T=100. a ₀ and a ₁ in (11) before 1986 are set to the 1986 values, as a ₁>1 before 1986. The value of “Oil and gas revenue ts permanent” for year 2001 was an outlier and was replaced by the average of the 2000 and 2002 value