What does Egypt's Revolution Reveal about its Economy?

ABSTRACT

On the third anniversary of the Egyptian revolution and against the backdrop of lingering political instability and deteriorating economic conditions, we diagnose the constraints to sectoral growth in Egypt using the 2011 Egyptian revolution as a natural experiment. We combine quantile regressions to study sector outliers with a difference in difference methodology to capture sectoral behavior before and after revolution. We find that the revolution's effect has been adverse, on average, but heterogeneous across sectors. We identify and characterize sectors most and least impacted. Results reveal that Egypt's fastest growing sectors before Revolution have been the most vulnerable after Revolution. This evidence is supported by our diagnosis approach that shows that faster growing sectors are constrained by continuous increases in prices that threaten export competitiveness (as they erode the benefits accrued to nominal depreciation of currency). Such sectors also benefited from higher monetary growth and fewer constraints on credit availability that have mitigated somewhat the speed of deterioration in the aftermath of the revolution. Our results, which hold under several robustness checks, inform policy priorities as to how to revive investors’ confidence, boost competitiveness, and design priorities in industrial policy to ease structural impediments and align sectoral growth with macro priorities.

Keywords:

• Egyptian revolution
• quantile treatment effect
• natural experiment
• sectoral growth

Notes

¹The second quarter of 2012 marks the end of the transitional period governed by the Supreme Council of the Armed Forces and the election of the first democratic President, who lasted in power for one year before he was forced out of office in response to growing dissatisfaction with power domination, mismanagement and failure to deliver to expectations on the economic front.

²The problem of heterogeneity is of course not entirely alleviated. While the event is unique across statistical cells, the heterogeneity across sectors still remains.

³The fiscal year data end in June 2012, which precedes the election of the first president after the revolution, who assumed power in July 2012. The growth rate for the following fiscal year, ending with his toppling from office, was around 2% and is likely to remain below 3% for the current fiscal year.

⁴Sectors’ numbers follow the order of the comprehensive list in Table 2.

⁵It has been argued in the literature that one should use the test proposed by Elliott, Rothenberg, and Stock (1996) for maximum power against very persistent alternatives, i.e., series that are very close to being non-stationary. The rejection of the null hypothesis in the case of the ADF test of the nominal effective exchange rate (neer) and the nominal exchange rate (exr), illustrates the relevance of using the ERS test in our context.

⁶Investment refers to both private and public investments, i.e., total investment.

⁷It is worth noting that adding sector dummies to the pooled OLS regressions as in columns 2 and 4 of Table 4 incorporates individual sector heterogeneity and is equivalent to estimating the model as a panel fixed effects model.

⁸Results are not reported for space considerations but are available from the authors upon request. The reported results are based on the above specification, which has several advantages. First, CPI inflation is the most tracked measure of inflation in Egypt, which has a direct impact on the cost of business. Secondly, monetary growth, based on M1, is more easily controlled by the policymakers, as it accounts for currency in circulation and demand deposits. Finally, we use the real effective exchange rate to measure relative competitiveness with respect to major trading partners for Egypt, mainly in Europe and the United States.

⁹In other columns in the table, we use different specifications as a means to check the robustness of our findings. As can be seen from columns 2 and 3 in Table 5, adding dgcomm or dfreserve delivers almost the same result. In columns 2 and 3, the non-affected sectors are 1, 2, 3, 5, 7, 10, 15 and 17. These sectors are: Agriculture, Mining Crude Oil, Mining Natural Gas, Manufacturing of Oil Products, Electricity, Transportation, Insurance & Social Security and Real Estate Activities, respectively. The resilience of these sectors stems from the nature of the output, which is less dependent on external demand and less vulnerable to cyclicality in domestic consumption. All other sectors report a negative and significant coefficient on the variable of interest. These are sectors 4 (Other Mining), 6 (Other Manufacturing), 8 (Water & Sewerage), 9 (Construction Building), 11 (Communication & Information), 12 (Suez Canal), 13 (Internal Trade), 14 (Financial Intermediation), 16 (Restaurants & Hotels), 18 (Education), 19 (Health) and 20 (Other Services). Complete results are not reported for space considerations, but are available from the authors upon request.

¹⁰Rosenbaum and Rubin (1983) define propensity scores, p(X), as the conditional probability of receiving the treatment given the covariates.

¹¹According to Frölich and Melly (2010), if quantile effects are conditional and treatment is exogenous (conditional on covariates) then the estimator proposed by Koenker and Bassett (1978) is the most appropriate. If the treatment is endogenous, the IV estimator of Abadie, Angrist, and Imbens (2002) is applied. Third, for estimating unconditional quantile effects, one would adopt the approach proposed by Firpo (2007), when the treatment effect is exogenous, and by Frölich and Melly (2008), when the treatment effect is endogenous. Surveys of the relevant literature include Imbens and Wooldridge (2009), while applications include Martincus and Carballo (2010), among others.

¹²Using different specifications has only added to these three sectors but they were always among those reported to be non-negatively affected.

¹³The Firpo (2007) estimation method is essentially a reweighed version of the procedure proposed by Koenker and Bassett (1978) for the quantile estimation problem.

¹⁴A check function ρ_τ(x)=τ*x if x≥0 and (τ−1)*x if x<0 (see Koenker & Bassett, 1978, for more details).

¹⁵This does not mean that other sectors have never witnessed growth rates above 2% throughout the sample period. It only means that these specific sectors identified above have witnessed growth rates of 2% and above (sixth quantile and above) most of the time throughout the sample period, particularly when the economy experienced robust high growth rates, while others have lagged behind for the most part.

¹⁶Sectors 1 (Agriculture), 4 (Other Mining), 9 (Construction Building), 11 (Communication & Information), 13 (Internal Trade) and 16 (Restaurants & Hotels) report a positive and statistically significant coefficient on the revolution/sector/dreer variable. On the other hand, the same coefficient is negative and significant for sectors 7 (Electricity), 10 (Transportation), 14 (Financial Intermediation), 17 (Real Estate Activities) and 19 (Health). With the exception of 19, these sectors are mostly slower growing sectors in the bottom of the distribution of the outcome variable and are, therefore, expected to report negative coefficients.

¹⁷Sectors 1 (Agriculture), 9 (Construction Building), 16 (Restaurants & Hotels) and 19 (Health) report a positive and statistically significant coefficient, as predicted by the QTT results. On the other hand, sectors 8 (Water & Sewerage), 14 (Financial Intermediation) and 17 (Real Estate Activities) which roughly fall among the bottom two quantiles of the outcome variable, report negative and significant money supply coefficients.