Targeting debt in Lebanon: a structural macro-econometric model

ABSTRACT

This paper estimates a structural macro-econometric model of the Lebanese economy to simulate the implications of accumulated debt changes on GDP and other economic indicators, and to project the growth–fiscal nexus for the six years following the last year for which national statistics are systematically available, 2015–2020. To these ends, historical and up-to-date national accounts data for the years 1992–2014 were painstakingly collected from individual government agencies, and an economic framework with five macroeconomic blocks (macroeconomic; government; price; monetary and financial sector; and external accounts) was constructed. Our simulations predict that additional deficits and more debt accumulation would deter growth, while fiscal consolidation such as paying off government debt is growth promoting. We use the prospect of future natural gas revenues as a potential external source to pay off Lebanese national debt – rather than as a collateral for additional borrowing – in order to promote sustainable economic growth. These results have important implications for the fiscal position and macroeconomic policy in Lebanon as well as other transitional, potentially resource-rich, open economies, where debt servicing is crowding out other growth promoting government spending activities.

KEYWORDS:

• Macroeconomic framework
• public debt
• Lebanon
• national accounts
• crowding out

JEL CODES:

• E27
• E42
• N15
• O13
• Q32

Acknowledgements

This document has been reproduced in the form in which it was received, without formal editing The opinions expressed are those of the authors and do not necessarily reflect the views of ESCWA or any other UN office or agency. Research assistance by Kristine Najjar is gratefully acknowledged. The Authors relied in their descriptive analysis for the first two sections of this paper on an unpublished research paper commissioned by ESCWA in 2015. The current version was supported by the Economic Research Forum (ERF). The contents and recommendations do not necessarily reflect ERF's views. We are grateful to Lyn Squire, Magda ElSayed Kandil and an anonymous referee for constructive comments.

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Vladimir Hlasny http://orcid.org/0000-0002-6029-1634

Notes

1 This review of economic trends and challenges was highly selective and tailored to the estimable macro-econometric framework in the following section. More discussion of trends in various sectors of the Lebanese economy is provided in the working paper version of our study (Araji, Hlasny, Mansour Ichrakieh, & Intini, 2017).

2 Multiple robustness tests have been conducted along the entire modeling process. Model selection criteria were used to choose the appropriate form of dependent variables and sets of independent variables. All time-series have been checked for non-stationarity and a unit root, and corrections were performed to convert them to stationary processes.

3 One limitation with regards to unemployment data is that they come from ILO estimates, which are different from national estimates. The IMF uses different unemployment estimates. Regressions in this study rely on ILO estimates available for all years 1992–2014.

4 We have estimated this equation also without first-differencing, and while controlling for lagged consumption in agreement with the habit persistence utility hypothesis. Indeed, a strong, highly persistent relationship was found. Because of this persistence, the equation was ultimately estimated in the first-differenced form. Remittance inflows were considered as a factor affecting consumption, but the time series exhibited poor statistical properties on account of inconsistent data (refer to the annex) and limited significance in explaining fluctuations in private consumption.

5 In a previous version of our model, government capital expenditures were treated as behaviorally dependent on budget deficit, but the equation was omitted from the current model for conceptual reasons (indeterminate direction of causality, and inappropriate inference). The results of that equations were as follows: $\widehat{\mathrm{\Delta }GCE}=\begin{array}{c}-{0.493}^{\ast \ast }\\ \left(0.183\right)\end{array}\times \mathrm{\Delta }B{D}_t+\begin{array}{c}0.336\\ \left(0.440\right)\end{array}\times Y{1996}_t-\begin{array}{c}0.200\\ \left(0.417\right)\end{array}\times Y{2006}_t+\begin{array}{c}0.034\\ \left(0.091\right)\end{array}$. N = 22, R²  = 0.29, F = 2.44*, DW = 1.99. Change in government capital expenditure is thus found to be related negatively and significantly to change in budget deficit, in agreement with expectations.

6 To evaluate the quality of our real national income data and correlations among variables, we have estimated the following regression, even though this specification is not entirely appropriate conceptually as it violates exogeneity: $\widehat{\mathrm{\Delta }{\mathrm{Y}}^r}=0.004+0.92\times \mathrm{\Delta }\mathrm{C}+0.26\times \mathrm{\Delta }\mathrm{I}-0.56\times \mathrm{\Delta }\mathrm{Im}+0.13\times \mathrm{\Delta }\mathrm{Ex}+0.11\times \mathrm{\Delta }\mathrm{G}$. Reassuringly, all coefficients have the expected signs, and all are significant at a 1 percent significance level. R²  = 0.97, Pr(F) = 0.00, DW = 1.76.

7 Further, disaggregation of the balance of trade of Lebanon revealed that Lebanon's imports are dominated by imports of refined oil, and other commodities mainly from China. However, accounting for the price of oil and LBP-RMB exchange rate did not yield significant effects, and these controls were dropped from the equation. Oil prices do not appear to significantly affect imports in Lebanon, since this commodity is considered a necessity for electricity production (no substitutes such as water or nuclear power in Lebanon), private consumption (absence of efficient transportation) and production.

8 An indicator for the Syrian civil war was also considered as a factor in equations 1 and 23, but it had insignificant impacts on the change in unemployment, and on the change in exports, after controlling for the selected economic indicators, and given the short span since the outbreak of the civil war. The lack of significance could also be explained by the informal nature of jobs taken by refugees, the pre-war Syrian employment which was proportionally high, and the low economic growth in the region resulting from low oil prices, low investment, decline in remittances, and political instability. Controlling for relevant economic indicators led us to finding insignificance for the remaining effects of the Syrian war.

9 Akaike and Schwarz information criteria were also evaluated in the model-selection process, but are not reported with the final version of the model since equations in Table 1 are not nested, and there are no benchmarks to compare the statistics to.

10 Beside equations estimated here, we have considered adding an equation for aggregate demand $Y_t$ as a function of consumption, investment, net exports and government spending (similar to the identity equation 6). This equation was removed for fear of its misspecification. Linking this equation for demand with equations 2–4, the effects of exports, consumption and investment on total output were positive and mostly significant. These results agreed with macroeconomic theory and suggested that the economic cycle is heavily dependent on consumer spending, as this affects aggregate demand the most. One surprising results involved the effect – negative – of government capital expenditures on aggregate demand. Capital investment typically leads to output growth. However, many developing countries encountered a negative relation between government investments and output growth (Robinson and Torvik, 2005). This is a result of misallocation of investments into ‘white elephant’ projects and effectively underinvestment, which usually leads to negative social surpluses. This type of projects may be common in Lebanon, as our evidence suggested.

11 We are grateful to a referee for making this observation.

12 While the specification for exports reported here does not control for exchange rates, our alternative specifications revealed that a stronger dollar facing the Lebanese pound would result in a lagged increase in exports during 1992–2014. This and additional control variables were omitted here to keep the model parsimonious.

13 Equations 12, 13 and 25 have Durbin-Watson statistics greater than 2, showing some evidence of negative autocorrelation in residuals (insignificant). This may indicate that the regression specifications in first differences have introduced autocorrelation to model residuals. Still, these specifications are believed to be more robust to other issues with model variables and residuals, including strong dependence, and the small negative autocorrelations are weaker than the positive autocorrelations in the specifications in levels.

14 Both private- and public-sector financing approaches have a role to play and can interact as part of an effective strategy to debt reduction. This calls for careful modeling that is beyond the scope of our empirical model, without formalizing private-finance determination and scenarios, and without access to relevant, adequately decomposed data, such as disaggregated data on capital activities. This study limits itself to investigating several scenarios for public finance development.

15 Existing theory and evidence from other countries (Arouri, Boubaker, & Grais, 2018; Elbadawi & Selim, 2016) lead to several predictions. The anticipated economic benefits of gas extraction to national investment, national wealth and economic growth will only accrue if the authorities adhere to the Extractive Industries Transparency Initiative global standards along with satisfactory fiscal and price rules.

16 Clearly, analyzing the impact of debt reduction without reviewing the relationship between the components of public finance and debt reduction is a shortcoming. This is an unsurmountable problem because data on budget activities for the entire period 1992–2014 are missing. Our aim has been to simulate growth by imposing shocks on the expenditure/investment components of the fiscal budget, but data were missing for the fiscal budget component of interest. Also to be noted, Lebanon had no approved budget for many years indicating the ad hock spending nature of the government.

Additional information

Funding

The current version was supported by the Economic Research Forum (ERF).