Data Note: COVID-19, social distancing, and pipeline vandalism in Nigeria

The method

Several methods exist to determine the significance and relationship between groups of data; for example, t-test, ANOVA, etc. In recent times, some scholars have challenged the use of a threshold to declare the statistical significance of the p-value.^13–17 Two main arguments exist. One; research data contain more meaning than is summarized in a P-value and its statistical significance. Two, the concepts are frequently misunderstood and consequently inappropriately interpreted. The abolishment of p-values has been echoed in some articles; an example is Ref. 18.

Traditionally, Researchers examine differences between groups using t-test, ANOVA.^17,19-21 In this study the one-way ANOVA was used; and, percentage differences were added as quantifiers. A one-way ANOVA is a statistical test used to determine whether or not there is a significant difference between the means of three or more independent groups.

The assumptions in the one-way ANOVA are:

The verification of the assumptions used in the ANOVA analysis

Normality - This can check visually or, by the histogram. Normality checks were carried out on all subgroups datasets (Figure 1), prior-COVID, COVID and post-COVID datasets (Figure 2) and on historical datset (Figure 3). Before data can be analysed statistiscally, it must be shown to be “normally distributed”. “Normal” data are data that are drawn from a population that has a normal distribution. For the subgroups these are:

Figure 1. Normality check on subgroups.

Normality checks were conducted on the underlying data sets that formed each group and subgroup A-D to determine whether the sample data have been drawn from a normally distributed population (within some tolerance). When sample data are representative; the conclusions drawn from such data are often valid.²² (The Assumption(s) of Normality, http://www2.psychology.uiowa.edu/faculty/mordkoff/GradStats/part%201/I.07%20normal.pdf, Copyright © 2000, 2011, 2016, J. Toby Mordkoff).

Normality check on the group data:

Figure 2. Normality check on group data.

The data sets in the groups- pre-COVID, COVID and post-COVID, are 6 months before, during and after the COVID lockdown. While the historical data ranged from May 2015-Febuary 2019. These data were also subjected to normality check. A histogram is an estimate of the probability distribution of a continuous variable. The graphs (Figures 1, 2 and 3) approximate the characteristics bell-shape curve²³ (Lund Research Ltd. Descriptive and Inferential Statistics. Available http://www.statistics.lard.com).

Figure 3. Normality check on “historical data” used in the analysis.

Variance

This can be checked with a boxplot. For the subgroups these are:

Figure 4. Variance check on sub groups.

The check on the group data:

Figure 5. Variance check on groups.

Figure 6. Variance check on “Historical data group”.

Independence

There is no formal test to verify that the observations in each group are independent.

The data

The study uses easily accessible and verifiable data. Primary data was collected by the Researchers from first-hand sources. All data used in this study reside in public domains. This is in line with the Authors’ aim to allow ease to the methods, materials, and protocols. It also allows replication. Primary data of the number of cases of pipeline vandalism each month from January 2015 to January 2021were collected and grouped based on date. The names of each group are self-explanatory. The groups are:

Under the land use decree, the oil wealth of the country (Nigeria) resides with the Federal Government. All aspect is controlled by the Nigeria National Petroleum Cooperation (NNPC) or its subsidiaries. Incidences of vandalism of pipeline are ascribed in the Monthly Financial and Operations Reports (MFOR) of the Nigeria National Petroleum Corporation (NNPC).²² Thus, the integrity and veracity of data used cannot be in doubt. These monthly reports are available for free download by the public from the NNPC website link (NNPC; https://www.nnpcgroup.com).²² This information can be accessed by clicking “NNPC Business” and selecting “Business Information”, then “Monthly Performance Data” on their website (https://www.nnpcgroup.com) or through (https://www.nnpcgroup.com/NNPC-Business/Business-Information/Pages/Monthly-Performance-Data.aspx). The underlying data is available for free download by the public, thus; reproducibility of the dataset is facilitated. The information abstracted from the NNPC MFOR was the number of cases of pipeline vandalism per month, the month of vandalism, and the year of vandalism.

Furthermore, we obtained information and dates of major National and International events that may be additional external stimuli in this analysis. This information was collected from national and regional newspapers and web-based publications, and web pages.

These are:

Classifications in group 1, “Historical Data”

“Historical data” span from January 2015 to July 2019, these data represent pipeline vandalism data before the advent of COVID-19 and its restrictions. These data were collected before the outbreak. Consequently, it could be assumed that COVID-19 did not influence the incidences of pipeline vandalism during this time. These data can therefore be used as a “baseline”; a such of “norm”. For in-depth study, this group (spanning approximately 4 years was further divided into subgroups of a year durations each).

The sub-groups are:

An analysis of the sub-groups would reveal if the use of GIS had any impact on cases of pipeline vandalism.

Classifications in “groups 2-4”

Data from groups 2-4 were arbitrarily set within a duration of 6-months each. Pipeline vandalism during the time frames could be imparted by COVID-19. A comparative analysis of data six months before lockdown (group 2, Table 1) and six months after lockdown (group 4, Table 1) would reveal if COVID-19 had any impact on cases of pipeline vandalism.

To minimize/remove seasonal variations due to the weather (wet and dry season) data were compared only with data from the corresponding seasonal frame. This is logical, in temperate zones, data of pipeline vandalism in the summer should be compared against summer data; winter against winter in the colder zones; similarly data of pipeline vandalism in the rainy season should be compared only against data of pipeline vandalism in another rainy season in the tropic.

Software used in the data analysis

For ease of accessibility, the software used for analysis was the MS office Excel 2013 with the Analysis ToolPak add-in.

Reproducibility and replication of results

The reproducibility of data determines if similar results or conclusions could be attained by a different research team, using the same methods. The results in this study, are not artifacts of the unique setup, therefore any researcher using any statistical tool should lead to the same/similar results.

Replication, on the other hand, refers to the repetition of a research study, usually with different situations and different subjects. This determines if the basic findings of the original study can be applied to other participants and circumstances. It can be considered as a “re-run” study; aimed to confirm results. The severe acute respiratory syndrome (novel coronavirus COVID-19 or SARA-CoV-2) and its associated lockdown offered a unique opportunity that may not be replicable on the same scale.

In all statistical analyses in this project, an alpha = 0.05 as the significance threshold was set. In line with best practice for transparency in data analysis, our research hypotheses were clearly articulated; and null and alternative hypotheses were established. This means that the null hypothesis would be rejected if the p-value is less than or equal to 0.05 and the alternative hypothesis would be accepted.

The programmed MS Excel spreadsheet

The programmed MS Excel spreadsheet was used in the calculation of the time series analysis using a moving average.

The programmed Excel spreadsheet consist of rows and column. Each column was given a unique identifier ranging from 1 to 10 (Figure 7) and a column heading which is the formula used for the calculation in the column (Figure 7).

The third and fourth columns are the date and cases of pipeline vandalism for the period. The fifth column is the moving average. In the excel spreadsheet, the moving average, MA was calculated by:

The moving average, $\mathit{MA}=\mathit{AVERAGE}\left(D_t,:,D_{t+1}\right)$

Where D_t = is the cases of pipeline vandalism at time, t

And

D_(t+1) = is the cases of pipeline vandalism at time, t+1

In columns 6-9 the seasonal and irregularity components are handled and the data deseasonalized.

Figure 7. The parts of the programmed MS Excel spreadsheet.

Analysis of the groups (1 August 2019 to 31 January 2021)

For the group data, the total incidences during the time frame covered by the group, average, and standard deviation were established. The grouped data were subjected to an ANOVA analysis, A null and alternative hypothesis were set as followed:

Analysis of the sub-groups (May 2015-April 2019)

For the sub-groups, the total cases in each subgroup, the mean, and the standard deviation were calculated. A null and an alternative hypothesis were set.

The sub-grouped data were also subjected to an ANOVA analysis, and a time series analysis (after the data were smoothened by moving average).

Allowances made to control bias or unwanted sources of variability

Seasonal confounds

There are two principal seasons, the wet rainy season and the hot dry season in Nigeria. Pipeline vandalism takes place in remote locations near isolated, rural roads and footpaths; not readily accessible during adverse weather conditions. We, therefore, assume that the rainfall affects the number of cases of pipeline vandalism. However, rainfall patterns are fairly predictable and torrential rainfall occurs in the mid of the rainy seasons. Seasonal confounds were eliminated by comparing data for the same months in each group. This implies that the rainfall season data (in one group or year) were compared only with the rainfall season data (in another group or year); with similar arguments for the dry season data.

Data points in each group or sub-group

For all analyses, the number of data points was of uniform length to reduce any possible bias due to unparalleled data points.

In each group (prior, during, and post COVID-19 lockdown groups) the number of data points was six. In the four sub-groups of the historical data (i.e., sub-group “A”, sub-group “B”, sub-group “C”, and sub-group “D”) each sub-group had 12 data points.

The data was assembled over an even interval and ordered chronologically with equal time-frequency.

Exclusion of data

All cases/incidences of vandalism of pipelines that fall before or after the time frame under review (1 January 2015 to 31 January 2021) as ascribed in the MFOR were removed from the analysis.

Other assumptions made

The destruction of these pipelines has been a scourge on the national petroleum industry in Nigeria since time immemorial,¹ two groups of people disrupt pipelines in Nigeria. One; the activists, radicals, and militants, to make political statements, and two, the thieves. Thieves, solely for monetary consideration via illegal possession of the fluids therein.

The former, make political statements before any attempted disruptions, often to inform the government and allow negotiation for the fulfillment of their demands; the latter does not. During the lockdown, no activists, radicals, or militants made any political statement; so, we can assume they also heeded the order to “isolate and social distance”. We, therefore, attributed all pipeline vandalism during the COVID-19 lockdown period to thieves.

Analysis of groups and sub-groups

A comparative analysis of data six months before lockdown (group 2, Table 1) and six months after lockdown (group 4, Table 1) would reveal if COVID-19 had any impact on cases of pipeline vandalism.

Figure 11 compares the cases of pipeline vandalism 6 months after the deployment of GIS and other security apparatus and in the months of lockdown. The total lockdown was observed to yield better results. It was observed that a blanket restriction lowered the cases of vandalism the most; to an all-time low of 19 cases in March 2020 after a start of 32 cases in Febuary 2020 it was observed that cases started to rise (Figure 11), although it never reach pre-lockdown levels. The blanket restriction on movement was most effective (Table 3).

Figure 9. Comparative analysis cases of pipeline vandalism with renovations of the methodologies used by NNPC.

Figure 10. Pipeline vandalism analysis by groups.

Figure 11. The effect of improvement of security apparatus and lockdown on cases of pipeline vandalim.

Figure 12. The effect of lockdown implementation on cases of pipeline vandalism.

Lag analysis

A “lag” is a fixed amount of time. In the lag analysis in this paper, two key observations (number of cases of vandalism) are plotted lagged. These are the time periods after the implementation of the GIS into the security apparatus to checkmate pipeline vandalism before changes could be observed. The six-month lag may be the “learning/training and implementation phase” after the media announcements and deployment.

Notice the declining incidences of pipeline vandalism from August after the installation in May with an all-time low in December (Figure 9). It was noted that after the periods of renovations of the methodologies used to checkmate the activities of vandals.

The uncompromising movement restrictions also favored a reduction in cases of pipeline vandalism, as a similar shift was observed in the groups’ data. These data span from six months before lockdown, the COVID-19 pandemics lockdown group and six months after. The lockdown period and the periods immediately after, presented the fewest cases of pipeline vandalism (Figure 10).

As observed from Figures 11 and 12, short term benefits were observed. The restriction of movement led to a reduction of pipeline vandalism when the number of cases of vandalism for any month are compared. However by December, initial ripples were observed.

The implementation of a different security protocol in May 2016 was found to be followed by a reduction in cases of pipeline vandalism. The restriction during the pandemic was found to be followed by a reduction in cases of pipeline vandalism. These methods could be said to be also effective.

Rate of effectiveness

From a security viewpoint, it was therefore desirous to determine if greater success would be attributed to either the use of GIS or blanket restriction. This would enable the design of a more winning approach to vandalism. For this, the cases of pipeline vandalism 6 months after the incorporation and implementation of the GIS systems and the 6 months of total and comprehensive lockdown were compared (Table 4).

To determine this, the rate of effectiveness was considered. This may be defined as:

Rate of effectiveness, $R_e=\left(\frac{\left(B,‒,A\right)}{t}\right)$

Where:

Rate of effectiveness = R_e

Number of cases at start of observation period = A

Number of cases at end of observation period = B

Time duration of observation period = t

Time series analysis

A time series analysis of the data was undertaken to determine the effect of time vandalism on pipeline and allow forecasting using a moving average (MA) model.

In time series analysis a sequence of data points recorded over an interval of time and collected at consistent intervals over the set period of time at consistent intervals is analysised. the data points are not intermittently or randomly selected. The time series analysis of subgroups A, B, C and D are shown in Tables 5, 6, 7, and 8 respectively. In Figures 13, 14, 15, and 16 dataset from each subgroups and generated moving average models are plotted. This shows the degree of fitness of the moving average models in each incidence.

Figure 13. Time series analysis subgroup A.

Figure 14. Time series analysis subgroup B.

Figure 15. Time series analysis subgroup C.

Figure 16. Time series analysis subgroup D.

Underlying data

Harvard Dataverse. Effects of COVID-19 on pipeline vandalism in Nigeria, West Africa. DOI: https://doi.org/10.7910/DVN/8X5KKB.³⁰

This project contains the following underlying data:

Dataset Data for Effects of COVID-19 on pipeline vandalism ingested files:

Data are available under the terms of the Creative Commons Zero “No rights reserved” data waiver (CC0 1.0 Public domain dedication).