Spot of Oil Prices Evaluation - Article Example

Spot of Oil Prices al Affiliation: Qatar This paper presents a review of oil prices between the years 1946 and 2013. Periodic prices listed between these two points in time have been used to construct conclusions about the characteristics of the data, and thereby provide insights into the characteristics of the data. Several methods were used for the review, namely; generation of descriptive statistics for the data, and generation of plots/ graphs. The graphs included histograms for investigating the distribution of the data, box plots for ascertaining the existence of outliers and establishing the points within which these outliers lie, scatter plots for investigating movements of prices over time, and time series plots for elucidating further features of the data, such as variability, seasonality and trend. The histogram and descriptive statistics revealed that the data has a heavy right tail, implying that the prices had risen within the last few years included in the analysis (beyond 2000, where the prices rose tremendously. Accordingly, the data had significant outliers, with prices placed above USD 70 appearing in the outliers’ list. These prices correspond with the latest years that were included in the analysis. This increase in oil prices in the most recent years was clearly represented within the scatter and time series plots. The data also has remarkable trend (an increasing trend), variability, and seasonality. In order to deal with the problems of trend, outliers, variability and seasonality, logarithms of raw oil prices were used to transform the data. Effectively, the outliers were trimmed to the extent the box plot of the transformed data did not have any outliers. However, the method did not address the issues of variability, seasonality and trend. Using the current data, we can conclude that oil prices are set for increments in the future. Keywords: trend, variability, outliers, logarithmic transformations, and detrending. Introduction The prices of oil are an important point of knowledge for all human societies within the globe. Since the turn of the twentieth century when major revolutions established the world as a global village, there has been significant harmonization of the ways in which people handle various issues. For instance, modern modes of transportation mainly depend on fuel obtained from refined crude oil. Equally, people use different forms of refined oil in their homes for different purposes, including heating, lighting, running machines, among other more sophisticated purposes. These simple factors draw our attention to one central factor: oil is one of the most valuable commodities to have in any part of the globe. However, several; factors also determine accessibility of the products of refined oil to people from different parts of the world. For instance, people from across the countries that do not have any oil deposits rely heavily on the production of the products by the producing countries. That is, accessibility depends on the levels of production within the producer markets. While a drop in the overall prices is an unlikely scenario, it is also important to note that the rise in the prices has also been accompanied by major adjustments (mostly positive) in the prices of other commodities. Such a relationship clearly indicates the importance of global oil prices. Often, oil has been used as a global tool for shaping geo-politics, mainly by the producing countries and other big players within the industry. Some of the effects of such politicking with the important product that is oil include massive inflation across countries and continents, with innocent consumers of the oil products always bearing the burden when cartels conspire to trigger reduction in production for the purposes of raising prices. At such time, countries that do not produce the commodity are often left at the mercy of the oil producers, suffering massive inflation and economic losses. As such, the prices of oil are of great importance to the global economy. In this report I have presented the characteristics of data on oil prices between first January 1946 and seventh January 2013. Closely related but slightly different approaches have been used to reveal the patterns within the data. For instance, descriptive statistics generated from the data are presented to show the nature of the data. Complementing these statistics is a set of graphs; all meant to portray the real qualities of the global market oil prices across the span of 68 years. Inevitably, the patterns within the data are reminiscent with the global developments both in terms of major price shifts (which are usually accompanied by changes in the rates of inflation based on the percentage rates of increase or decrease. As such, further research of major global happenings in the light of oil market rise or fall has been provided to complement the observations. The report is divided into major guiding sections; namely, the introduction, the analysis and discussion of descriptive statistics and graphs, and the conclusion section in the last part. The output has been prepared using the software Minitab. Data Analysis Based on the nature of the data, the following descriptive statistics have been obtained to describe the prices of global oil prices between 1946 and January 2013. Table 1. Basic statistics about historic oil prices. Statistic Oil price Log oil price Mean 21.665 1.0349 Standard error of the mean 0.987 0.0189 Standard deviation 25.537 0.5369 Variance 652.13 0.2883 Minimum 1.17 0.0682 Maximum 133.93 2.1269 Range 132.76 2.0587 Q1 2.97 0.4728 Median 14.85 1.1717 Q3 28.765 1.4589 IQR 25.795 0.9861 Skewness 1.86 0.10 Kurtosis 3.12 -1.33 Number of cases 811 811 Table 1 above shows the various basic statistics pertaining to the historical oil prices. From the onset, the values of kurtosis (3.12) and skewness (1.86) are greater than 1, implying that the data may not be suitable for parametric tests. However, we could overcome this hurdle by applying a suitable method of removing or minimizing both the skewness and kurtosis. For this analysis, the method of logarithms was applied based on its ease of understanding and lack of distortion of the important features of the data. The average price of oil for the study period was USD 21.665 (std. dev. = 25.537). Again, the standard deviation indicates a relatively very large standard deviation with relation to the mean. Apparently, with a mean price of 21.665, and without a possibility of acquiring negative values for the variable, it is clear that some years have presented a massive rise in the prices. This, indeed, is an important point about the data – comparing the mean and the standard deviation shows that for the better part of the 68 years under research, and considering the minimum (1.17) and maximum (133.93) prices listed within the period, it is apparent that the prices shot high instantly over a very small period of time. Figure 1. Histograms comparing the distributions of oil prices and logarithms of the prices. The figure above shows the distributions of the data with comparison to a normal distribution. Apparently, it can be observed that the histogram for the raw oil prices has both a more skewed and more peaked distribution. The normal line indicates the ideal “normal” path for the data to follow. As can be observed from the figure, the data is positively skewed, which means we cannot assume it is normally distributed. Equally, the peak, which in actual sense represents the highest number of equal values (also the mode) is more exaggerated. For a normal distribution, we would hope to find a histogram that is bell shaped, with the data fitting closely perfectly within the bell shape. However, this is not the case with the prices of oil. The table on descriptive statistics provides a good basis from which we can evaluate the properties of the data. The measures of skewness (1.86) and kurtosis (3.12) are both beyond the conventional threshold of 1, implying that the data is not normally distributed with respect to the two measures. The concurrent examination of the two sets of evidence obtained regarding the raw data indeed shows a perfect level of corroboration of each set of results. Examining the transformed data, we realize that the data has acquired a fairly normal distribution. With the exception of the kurtosis (-1.33), the measure of skewness (0.10) suggests a full transformation of the data since the value falls within the acceptable threshold of ±1 (Casineau & Chartier, 2010). Therefore, applying logarithms to the data has led to a fairly normal distribution. Now, the above discussion highlights a situation in which some oil prices, probably the most recent according to the review of the data, are way higher than the average prices witnessed in the oil market. That means that prices have grown significantly, to the point it is possible to observe outliers in the data. In order to investigate the presence of outliers, box plots have been selected to provide the necessary visual effects. As in the above histogram, the box plot presents a comparative view of oil prices and logarithms of the prices. Figure 2. Box plot of oil prices for identifying outliers. As shown in figure 2 above, all prices above USD 70 were enumerated as being outliers. The prices, therefore, do not lie within the acceptable limits based on the positions of measures of central tendency, specifically the mean, median and the mode. Such observations are treated differently in statistics, mainly in response to the type of method used in generating the data. For instance, in experimental studies, outliers may be treated as experimental errors, or merely errors in data recording. They may also be attributed to variability within a dataset. The net effect of having outliers is sometimes undesirable, as they may be representative of errors in measurement or recording of data. As such, they are often removed from the dataset. For the oil prices dataset, the outliers are indicative of the existence of a heavy tail on one side of the distribution. Referring back to the histogram of raw data, we can accurately attribute the existence of the outliers to the positive tail, denoted as positive skewness. In the current analysis, the outliers were not removed from the dataset; they were instead treated using logarithmic transformation. This method ensures that the data do not lose their initial characteristics, but rather coalesce closer together in accordance with their logarithmic values. Figure 3 below shows the effects of such transformation on the prices of oil. The effect is clearly notable in the sense that the new transformed data does not have any outliers, again based on a criterion of distances of individual values from the measures of central tendency listed before. The upper and lower tails of the data are also almost equal in the box plot of the transformed data, unlike the raw prices data that had a significantly smaller lower tail than the upper one. Using logarithmic transformation, we successfully repositioned the outliers within acceptable margins. One more notable change in the data is the revised position of the mean in relation to the number of values on either sides of the tails. Apparently, the transformation moved the mean-line from a visibly more central position to one closer to the upper tail. This is due to the fact that values initially unaccommodated within the actual margins of tolerance were reincorporated within the plot. Figure 3. Box plot of log oil prices showing the transformation of outliers. Figure 4. Scatter plot of oil prices versus logarithms of the prices indicating the increase in prices as we move further away from the base year. Figure 4 above shows the prices of oil plotted against time, and the logarithms of these prices again plotted against time. Based on the first scatter plot, there are different phases in the increment of oil prices to what was recorded in 2013. These phases can be classified in accordance to the existence of sudden changes and extensive lengths of relatively close prices Friendly & Denis, 2005). For instance, the period between 1946 and 1974 appears to have had relatively similar prices. This indicates price stability around this time period. The period that followed after 1974 was marked with a sudden increase in prices. However, the increase did not continue for long, and the prices remained within a certain margin up to the year 2000 when a sudden and continuous increase started. The period between 1975 and 2000 was marked with changing prices, but they always changed within limits established in the market in the early years. However, between 2001 and 2013, the prices rose to new, unprecedented highs, reaching an all-time high of USD 140 in 2009. Figure 5. Time series plot of oil prices showing trends and seasonality. Figure 5 above is the plot of the time series plot of oil prices across the time period 1946 to 2013. The importance of the plot stems from its ability to concurrently shed light on such characteristics as variability of data, seasonality and trends. While these features of a dataset are invaluable in understanding such mechanisms as operations in the oil market, the existence of any of them is often undesirable for further analysis of data (Cleveland, Cleveland, McRae & Terpenning, 1990). For instance, many statisticians and market analysts prefer to detrend the data for convenience of further analysis (Rosales & Krivobokova, 2012). Several methods are used in detrending data. As already done in this analysis, logarithms were used to not only remove the effect of heavy outliers in the data, but to also deal with such observed characteristics as seasonality and variability. Variability is expressed in the sudden rises or drops between data for consecutive time points. Seasonality is observed in the tendency of the prices to fluctuate within a certain limit for a short period of time, while trend is the overall tendency of the prices to gradually increase over time (Dong, Gao & Wang, 2013; Cleveland, 1979). The overall image from the time series analysis plots is that the prices of oil have increased tremendously between 1946 to data. The use of logarithms did not appear to fully address the issues of variability, trend and seasonality. As shown in figure 6 below, the data still shows all three characteristics. For instance, an increasing trend is still imminent despite the use of logarithmic detrending, while seasonality is still evident from the plot. Doubting the efficiency of logarithms as a powerful tool for detrending data at this point, two more methods were attempted – use of moving averages and exponential smoothing. Both methods vindicated the usefulness and correctness of the logarithmic detrending method by producing graphs very similar to the one in figure 6 below. Figure 6. Time series plot of log oil prices showing the effects of logarithmic smoothing. Conclusion The prices of oil have steadily increased between the year 1946 and 2013. However, much of the rise has been observed after the year 2000, when the current high increment in prices began. A look at the various plots and the descriptive statistics shows consistently similar characteristics about the data. Due to the sudden and high increase in prices after 2000, most raw prices within the period up to 2013 appear as outliers. The use of logarithmic transformation was effective in removing the effects of outliers, but did not successfully address other issues such as trend, variability and seasonality. The presence of these features shows that the data is effectively a time series. Using the current trends, we can safely extrapolate future oil prices, which will be predictably higher than they are at current. This analysis is important in shedding light on important features of the oil prices, and they provide a basis upon which further analysis such as market oriented studies can be based. References Casineau, D., Chartier, S. (2010). Outliers detection and treatment: A review. International Journal of Psychological Research, 3(1), 58-67. Cleveland, W. S. (1979). Robust locally weighted regression and smoothing scatterplots. Journal of the American Statistical Association, 74(368), 829-836. Cleveland, R. B., Cleveland, W. S., McRae, J. E., Terpenning, I. (1990). STL: Seasonal-trend decomposition procedure based on Loess. Journal of Official Statistics, 6(1), 3-73. Dong, K., Gao, Y., Wang, N. (2013). EMD method for minimizing the effect of seasonal trends in detrended cross-correlation analysis. Mathematical Problems in Engineering, Vol. 2013, Article 493893. Friendly, M., Denis, D. (2005). The early origins and development of the scatterplot. Journal of the History of Behavioural Sciences, 41(2), 103-130. Rosales, L. F., Krivobokova, T. (2012). Instant Trend-Seasonal Decomposition of Time Series with Splines. Georg-August-Universitat Gottingen. Read More