The Effects of Electricity Sources on the Cost of Electricity in Panama

The purpose of this study is to identify which electricity production sources have a significant effect on the marginal cost of electricity in Panama. Panama, like most Central American countries, relies on hydroelectric power, thermoelectric power and electricity purchases. We only collected secondary data provided by the official entities in order to ensure that the numbers used for the study were consistent. Linear regression analysis was used to find the model that describes the behavior of the marginal cost more accurately. The findings of the study confirm the need for less oil-dependent electricity production alternatives in order to achieve lower electricity prices in Panama.

Published by SCHOLINK INC. marginal cost has a standard deviation of 43%.
Central America is now going through a stage of integration. Electricity demand in Panama is increasing and if the marginal cost continues to increase, the economic growth of the country may be hindered. A growing economy requires to have an efficient electric network to support its operations.
In order to develop an efficient electric network, we need to understand the effect of the different production alternatives in the marginal cost of producing electricity. The aim of this present study is twofold: 1. Test the significance of the relationship between each production alternative and the marginal cost of electricity through linear regression analysis.
2. Make suggestions about future electric network development based on the results of the linear regression analysis.
One of the objectives of this work is to determine the significance of the relationship between each production factor and the marginal cost of electricity. Each production factor should influence the final value of the marginal cost of producing electricity. Figure 1 shows the research framework in a graphic mode.

Figure 1. Research Framework
The reminder of the paper is organized as follows. The next chapter examines the literature review. The third chapter analyses the research methodology. The fourth chapter discusses the data analysis. Finally, the last chapter concludes our study. Implications for practice are highlighted and limitations and future directions for inquiry are presented.

The Deregulation of Electric Markets and its Impact on the Cost of Electricity
According to Takriti, Krasenbrink, and Wu (2000) electric markets have been following a deregulation pattern in many countries. Initiatives undertaken by the United States since 1978 encouraged separation between generators and distributors. These policies have shaped the market into what it is today, and this model has also influenced many countries in Latin America, including Panama. The purpose of deregulation is to achieve cheaper prices for consumers through the increase in competition. Though, deregulation also has an impact on the electricity demand, which is said to be unpredictable due to trading transactions. Also, because electricity is a perishable, non-storable good, prices are extremely volatile and will depend on the supply and demand of the market at any given moment in time.
However, contrary to Takriti's considerations and thoughts on the purpose of deregulation, it has proven to increase the prices of electricity. Moreover, studies like "Electricity in Latin America" (Hall, 2004), show how deregulation and integration are not really working towards the benefit of the end user, but to the benefit of big corporations who try to squeeze the most profit they can from the markets they are in, in order to get the fastest possible investment recovery. In fact, the prices to the consumer in Panama have increased for almost all customer segments. international petroleum prices on countries which electricity generation is not heavily based on thermal power is comparatively low. Central America is very vulnerable to movements in the international price of petroleum, mainly because during the 1990's Central America modified their production grid to be more dependent on thermal power. The reason for this decision according to the authors is that thermal plants require a lower initial investment than hydroelectric plants and that at the time the investment decisions made petroleum prices lower. Table 2 exhibits the price of producing energy increased as oil prices increased. As illustrated in Figure 2, most of the countries in the study, excluding Costa Rica, followed this trend.
Referring to the case of Costa Rica, the authors assert that switching to renewable types of energy can be a feasible solution to reduce the impact of the high level of petroleum prices in the electricity cost, because electricity production is no longer subject to the variations of petroleum prices. This implies that for any country to be less vulnerable to oil price shocks, investment in non-thermal plants is necessary.

Figure 2. Thermal Generation as a Percentage of Total Electricity Generation
Source: Elaborated by Artana et al. (2007), p. 30 with information from CEPAL. Furthermore, the OPEC basket price in Figure 3 shows an increase of around 80% from March 2005 to December 2010. Contrasting it with the 46% increase of the marginal cost, we can preliminarily infer that thermal power might be a very important driver of the marginal cost.

Figure 3. Marginal Cost of Electricity vs. OPEC Oil Prices
Source: Elaborated with data from the CND (2005-2010) and OPEC basket prices available at www.opec.com

Additional Characteristics of the Electricity Supply
According to Takriti et al. (2000) electricity is traded between regions on a spot price basis. Moreover, market volatility can affect the input of electricity generation and the electricity trading between networks. To handle this situation, he defines his variables as stochastic variables. Takriti's (2000) concept of stochastic variables, which are values that are not known with certainty, can perfectly apply to electricity production by source, as it is a mix between domestically produced electricity and electricity purchased from outside the national electric grid. In other words, electricity produced from outside the system can be considered a stochastic variable, because the electric market has no way to precisely know or control how much electricity will be available for purchase at a certain period.
Moreover, even though the national production of hydropower and thermal power can be controlled, the amount of water which will be available in the reservoirs to produce hydropower is uncertain. Even though generators base their production on forecasted demand values, demand itself is a variable that is very difficult to predict. Additionally, producer decisions of which source of production to use, and whether to produce or to purchase from outside the grid, will depend on several internal and external factors and decision criteria. Therefore, to make our model simple and workable it is assumed that such information is efficiently reflected by the values of the MwH produced or purchased each month. In other words, we assume that the behavior of the entire market is reflected on the production volume, because the electric market is considered efficient. Many studies on the subject targeted by Takriti et al. (2000) use deterministic models to approach electricity production allocation problems. An example of deterministic models used previously, though not mentioned in Takriti's study, is the work of Anderson (1972). Even though this paper dates from the early 1970's, it is still being cited and was found cited as a reference in 363 papers regarding changes in electric network infrastructure, green energy related investments and optimization of electric systems. Its explanations and assumptions on the operating characteristics of electric networks are still valid nowadays, because electricity generation technology has not changed much throughout the years. Interestingly, Anderson (1972) mentioned in his work that hydropower usually has lower operation costs than fossil fuel-based power. Hydroelectric power is considered to be cheaper than fossil fuel alternatives because the water that moves the turbines used to produce electricity taken directly from the water sources available in the region of the plant at no cost other than the operation cost of taking the water from the source. However, hydroelectric power is not as reliable as fuel-based alternatives, and, therefore, needs to be combined with it in order for the system to be able to cover demand peaks and to ensure a stable supply of electricity that complies with the security levels required by the market.
Other studies found on the topic have a more social/political orientation providing good explanation of the ideologies and reasons behind Central American decision criteria for investment decisions pertaining to which kinds of electricity generation facilities to invest in. Nonetheless, these studies do not contribute to the development of our model. If the reader is interested in the topic, a good starting point is the 2004 policy research report of the World Bank (Kessides, 2004).

Data Collection
In order to establish a realistic model, data collected must be reliable and consistent. Accordingly, all information used in the study has been collected from official sources. Thanks to the regulations of the TMME, government entities and market regulatory organizations in Panama provide the required information freely to the public in order to comply with the transparency mandates. In Panama, the CND is the entity that coordinates all transactions and operations of the electric market and therefore, most of the information about the market can be found through their databases. Information such as electricity produced by source, demand and marginal cost, has been obtained from them. Also, most of the descriptions of the electric market have been extracted from information provided on the CND website. Information available on this website dates from March 2005 up to the present. Thus, we can be confident that the data series are broad enough for the analysis. Weather information, such as temperature used in the first section to verify the significance of the effect of temperature in electricity demand was obtained from the Department of Hydrometeorology at ETESA, which manages all national meteorological data in Panama and oil prices were obtained directly from the OPEC official website. No surveys are necessary for this work. Statistical analysis will be the tool used to determine the validity of the relationships between variables. Figure 3 shows four different electricity production alternatives that can be separated in two groups: electricity produced inside the electric grid and electricity produced outside the grid. The average electricity production mix has been explained in the first section. Depending on how much MwH of each production alternative is input into the system, it is expected that the marginal cost will increase or decrease. Based on the literature review and the behavior of the marginal cost schematized in Figure 1, our first assumption is that the electricity cost will decrease as the concentration of hydroelectric power in the mix increases, given that it is significant to the marginal cost. Since hydroelectric plants usually have a lower average operating cost, the average cost of producing electricity should decrease.

Hypotheses Testing
Therefore, we state our first hypothesis as follows: H1: There is a negative relationship between hydroelectric power and marginal cost.
Additionally, as mentioned in the literature review, thermal power is normally used to cover demand peaks, because hydro power is said not to be as reliable as thermal power. This assertion is logical in that hydroelectric plants take their input (water) from the adjacent sources and the amount of water on those reservoirs depends on the weather. Weather, being an unpredictable and uncontrollable variable, gives water levels the same characteristics. Even though techniques such as constructing dams have been developed to minimize the variability of water levels and to provide the plants with more accessible water, it is impossible to control how much water is going to be available at any moment in time. Following the fact that thermoelectric plants, in theory, are more reliable than hydroelectric plants because we can control the input and because the utilized input (fossil fuel) is a limited resource with high price volatility, and following the findings summarized in the literature review and in the first section, we would expect that thermoelectric plants contribute to increase the marginal cost, given that the relationship is significant. Thus, we can state our second hypothesis as follows: H2: There is a positive relationship between thermoelectric power and marginal cost.
Even though hydroelectric and thermoelectric power accounts for 89% of the electricity generated in the country, the remaining requirement is supplied by energy purchased from the Panama Canal and from outside the borders of the country. Such sources are "out of the grid" and therefore there is no control over their production costs. Therefore, stochastic variables are now introduced into the model, as the government and the distributors have no control over how that energy is produced out of the grid.
Therefore, even if the quantity is small, compared to the sum of the other two alternatives, we would expect some extra variability in the marginal cost to be caused by those factors. Electricity purchases could present either a buffering effect on the marginal cost or contribute to its increase. The amount of available power for purchase will depend upon the capacity of the seller and the current demand of the electric systems where it is generated. However, despite such considerations, we can still track the effect of electricity purchases on marginal cost by studying the historical data provided by the time series obtained from the CND. It will be assumed that electricity purchased from outside the grid is more expensive than in-house produced electricity. Therefore, the third and fourth hypotheses are stated as follows: H3: There is a positive relationship between electricity purchased from the Panama Canal and marginal cost.
H4: There is a positive relationship between electricity purchased from outside the country and marginal cost.

Independent Variables
Our model has four independent variables. Each one of them accounts for a different source of producing or purchasing energy. We consider two groups: national produced electricity and out-of-system purchases, which also includes the electricity purchased from the Panama Canal. All the independent variables are measured in volume of production, measured in MwH, because of the lack of information about actual average production cost for each electricity generation alternative.
then the resultant equation can be used to explain the relations between independent and dependent variables. A graphic representation of the complete data analysis procedure, including data collection is presented in Figure 4.

Descriptive Statistics
The average, standard deviation, maximum and minimum of the data were calculated and summarized in Table 3, where H is hydropower, TH is thermal power, ACP stands for Panama Canal purchases, EI stands for electricity imports and MC represents the marginal cost. The factor with the greatest variability is electricity imports, with a standard deviation that represents 110% of its average value.
The rest of the production factors show a lower standard deviation with values ranging from 18% to 29% of their average value. However, the marginal cost itself displays substantial variability as well, with a standard deviation of 43% of its average value.  Table   5 and Table 6 and the correspondent equations are the following: 1-month season model: 2-month season model: M Comparing Tables 5 and 6，it can be inferred that the 1-month seasons model have a better fit to the data than the 2-month seasons model. The regression analysis, after removing the non-significant terms for the 1-month season model is shown in Table 6.  Table 6 shows the 1-month seasons model. The other production alternatives were found not to have a significant relation to the marginal cost. The model above also reveals that the coefficient of the thermal power is in the order of 10 -4 . Please note that while marginal cost (dependent variable) is measured in USD/MwH, which is in the order of tenths and hundreds, the volume of production by each alternative (independent variables), given in MwH, is usually in the order of thousands and hundreds of thousands. Rescaling couldn't be applied to the data series because there were some zero values in the time series of electricity imports, which didn't allow us to apply LOG function to the data series. However, the model is significant according to the F-Statistic value and the P-value of the variable shows that it is also significant. After conducting the multiple regression analysis, we can conclude that H1, H3, and H4 are rejected and that H2 is accepted. In other words, all production alternatives except for thermoelectric power are not significantly related to the marginal cost. Therefore, we can conclude that thermal power generation is an important driver of the marginal cost. Our findings are in line with Artana's (2007) research, which indicates that fossil fuel-based electricity production has a significant impact on the cost of producing electricity. The equation that describes the model is obtained from the regression analysis and can be written as: Where TH represents the thermal power and MC is the marginal cost. The equation shows that the marginal cost has a lower bound of 39.81USD/MwH if the value of TH is set to 0 and increases as thermal power utilization scales up. Our model further confirms the assumptions about thermal power based production of electricity having an adverse effect on the marginal cost, as exposed by Artana (2007). Moreover, the fact that the other production alternatives were found not to have a significant effect also means that by reducing the dependence of fossil fuel-based alternatives the marginal cost can be decreased. However, we must remember that thermal power is one of the most reliable electricity production alternatives.

Implications of the Findings
It is in the interest of the market in general, to reduce the marginal cost. It can be discussed that according to the equation obtained from the regression analysis, thermal power dependence needs to be reduced. However, there are some limitations to this approach. The fact that thermal power generation is still present around the globe is an indicator that there are reasons to keep these plants running. Even though hydroelectric power, Panama Canal purchases and electricity imports were found not to be significant, producing electricity through hydroelectric plants does generate an operation cost and electricity originated from outside the electric network has a purchasing cost. The fact that the regression analysis found them not significant doesn't mean they don't add up to the cost, but that the variation of their value does not significantly impact the outcome of the marginal cost. Moreover, the literature review and the further regression analysis of the impact of OPEC oil prices on the marginal cost of electricity in Panama also support our argumentation. The relationship between oil prices and electricity cost is already proven, therefore it makes sense that the thermal power has a significant impact on the marginal cost. The real problem is how to reduce the dependence on thermoelectric power to cover the demand peaks. If we refer to Table 4, we can see that the maximum historical electricity production of hydroelectric power is around 408,434MwH. On the other hand, the historical average demand is around 516,921MwH. This means that the historical maximum production of hydroelectric power can only cover approximately 79% of the average demand.
Electricity purchases are not enough to cover the demand, because their contribution to the total electricity production is low and are not controllable variables, because they are external to the system.
That means that even if we wanted to purchase electricity from the Panama Canal or import it from outside the country to supply what hydro power is not able to supply, it makes no sense to establish goals for the purchases if there is no security that the demanded power will be provided. As explained before by Takriti (2000), hydro plants are used to cover average demand, leaving the peaks to be covered by thermal power. This means that, if we can increase the level of coverage of non-fossil fuel-based production alternatives, we could make the peaks that need to be covered by thermal power smaller in magnitude. In that way, the marginal cost could be reduced. It is not by chance that green technology shows an increasing trend all over the world as mentioned in the Renewables Global Status more efficient and is being more widely supported.

Conclusion
This study demonstrates that the electric system in Panama is vulnerable to petrol price shocks. The resultant regression model clearly supports the studies of Artana (2007). The high dependability of the electric system on thermal power is the main cause of the increase of the marginal cost during the studied period. To test whether our hypotheses were supported, we used linear regression. Hypothesis 1 (H1): The relationship between hydroelectric power and marginal cost is significant, and hypothesis 3 (H3): The relationship between Panama Canal purchases and marginal cost is significant, and hypothesis 4 (H4): The relationship between electricity purchased from outside the country and marginal cost is significant were all not supported in our study. Hypothesis 2 (H2): The relationship between thermoelectric power and marginal cost is significant was supported in our research.
Following the objectives stated in the introduction we can state that after analyzing the different production factors we found out that only one of them was significant. The regression analysis of the independent variables (thermoelectric power, hydroelectric power, purchases from Panama Canal and Electricity imports) shows that the only significant factor is the thermoelectric power.
The main contribution of this paper, if not achieving any major leap on numerical methods or offering any new solution to solve the electricity problem, is to provide the end users in Panama with a realistic point of view, different from the one provided by the multinational companies. Those companies in most Latin American countries (Hall, 2004), and whose main objective is to profit from our markets as much as they can with the least investment. This study also aims at opening the eyes of the end users to the reality of our markets in order to motivate changes towards more green electricity production approach if they truly want the TMME initiative to fulfill its objectives and if they really want to improve the quality of life of their population. Otherwise, the TMME is just a bigger market for multinational companies to profit from in a more cost-effective way. Moreover, renewable energies present a clear benefit for each individual country and therefore we want to encourage Panamanian citizens to promote renewable energies. An electric system that is less vulnerable to petroleum price shocks needs to be developed. Fossil fuel-based alternatives not only bear a higher production cost, but also have high social and environmental costs. The use of thermal plants to produce electricity also fosters the need for oil transportation, which can lead to natural disasters such as the one which Our model, as any study, is bound by some limitations. The most critical one is that the financial cost of each MwH produced by the different alternatives was not available. Therefore, we had to base our regression model on volume of electricity produced. Performing a similar analysis with actual prices for each MwH produced or purchased by each different alternative could probably allow a more accurate model. Time limitations and privacy policies from the involved entities did not allow us to obtain these data in a timely manner and therefore future researchers should focus on developing these models. We think that it is important to research more in-depth how clean production alternatives affect the marginal cost of producing electricity in order to motivate the population to push the government to support clean energy production for sustainable development. Moreover, as stated by Artana et al. (2007), investing in non-thermal electric production has beneficial effects on the cost of electricity in the long run. Integrating more green sources of electricity makes the network less vulnerable to petroleum prices shocks. We hope that this study, being first in its class for Panama and probably Central America, can be referenced for future research on the performance of the Central American Electric Market and of each country in particular, thus contributing to the development of an efficient regional electric market that will really create tremendous benefits for end-users.