Recently I have come across a very interesting stat posted on the social media page of a defense lover group. It is about the crash record of Bangladesh Air Force since Bangladesh came into being. With one or two minor omissions it includes all the crash data of BAF. BAF suffered 90 aircrafts loss in 48 years.Suddenly the idea of finding the link between governance and crash record came to my mind. Common perception is that under "representative regime" , there would be better governance in every sector and they would run smoothly. There would be better maintenance of aircrafts, better storage facility and better training.
BAF lost 11 aircrafts in 2013-2018, a period many described as "hybrid regime".Having a closer look at the breakdown of crash incidents reveals that crashes under a "representative government" (27) surpasses that(19) under a "non-representative government".
But I hold the common perception and did a statistical analysis.
Though the result is a bit contradictory, it is interesting. The 48 years of Bangladesh is divided into 9 distinct periods. Depending on the nature of the government, the regime is dubbed as "representative" or "not representative". The representative regime, which is assigned to value 1, is assumed to come to power through free and fair poll. I did not decide the type of regime , rather its recognition by local and international bodies decided its type.
Since there are too many regime changes in the 70s, I termed 1972-80 as one distinct period. Then I counted the crash incidents took place in that period. BAF's worst moment came in 1991 when tidal wave rendered 40 fighter jets and 4 helicopters useless.Aircraft loss due to war or natural disaster however was not taken into consideration. For this reason, I ignored the 1971 trainer crash and loss of BAF 44 aircrafts in 1991 cyclone.
To test my conviction that "representative regime" has lower crash record, I ran a Poisson regression. For that, I first tried to find out the total number of combat aircrafts at BAF's disposal at the end of the respective period and proportion of crashed aircrafts to the total number of aircrafts in that period, which I called "fraction" here. Since it is difficult to get the exact number of BAF inventory, I took assistance of Wikipedia to get an approximate number of aircrafts in 2018 and then scaled down the total to glean information for earlier periods.Another thing to note is that I also took into consideration the aircrafts of Army Aviation Group but discarded the ones operated by the Navy while counting the total. Proportion of crashed aircraft to total aircrafts will give a better reading because loss of 1 aircraft from a total of 2 means a loss of 50%. Similarly, a loss of 1 aircraft from 20 means a 5% loss. Absolute number of loss could lead to misleading information. That is why I rest my trust on proportion, "the fraction" in this case.
While running the regression I dropped the observation for 1986-90, because no crash took place during that period. And exposure does not count any zero.
I first ran the regression with exposure( ) and then without the exposure( ) but including the variable that would normalize exposure.
The first run says combat aircraft crash under "representative regime" is 1.185 times larger than that of "non-representative regime". However, since it falls within the 95%confidence interval, I cannot reject the hypothesis that "representative regime" has lower crash records. The second run, without exposure, yields similar results. It is still 1.185 times larger and I cannot rule out that crash record under "representative regime" is lower.
Due to the results from regression, I cannot draw a definitive conclusion between BAF aircraft crash and governance.
[Please send your comments, contested views at mrh4478@gmail.com]
BAF lost 11 aircrafts in 2013-2018, a period many described as "hybrid regime".Having a closer look at the breakdown of crash incidents reveals that crashes under a "representative government" (27) surpasses that(19) under a "non-representative government".
Though the result is a bit contradictory, it is interesting. The 48 years of Bangladesh is divided into 9 distinct periods. Depending on the nature of the government, the regime is dubbed as "representative" or "not representative". The representative regime, which is assigned to value 1, is assumed to come to power through free and fair poll. I did not decide the type of regime , rather its recognition by local and international bodies decided its type.
Since there are too many regime changes in the 70s, I termed 1972-80 as one distinct period. Then I counted the crash incidents took place in that period. BAF's worst moment came in 1991 when tidal wave rendered 40 fighter jets and 4 helicopters useless.Aircraft loss due to war or natural disaster however was not taken into consideration. For this reason, I ignored the 1971 trainer crash and loss of BAF 44 aircrafts in 1991 cyclone.
To test my conviction that "representative regime" has lower crash record, I ran a Poisson regression. For that, I first tried to find out the total number of combat aircrafts at BAF's disposal at the end of the respective period and proportion of crashed aircrafts to the total number of aircrafts in that period, which I called "fraction" here. Since it is difficult to get the exact number of BAF inventory, I took assistance of Wikipedia to get an approximate number of aircrafts in 2018 and then scaled down the total to glean information for earlier periods.Another thing to note is that I also took into consideration the aircrafts of Army Aviation Group but discarded the ones operated by the Navy while counting the total. Proportion of crashed aircraft to total aircrafts will give a better reading because loss of 1 aircraft from a total of 2 means a loss of 50%. Similarly, a loss of 1 aircraft from 20 means a 5% loss. Absolute number of loss could lead to misleading information. That is why I rest my trust on proportion, "the fraction" in this case.
While running the regression I dropped the observation for 1986-90, because no crash took place during that period. And exposure does not count any zero.
I first ran the regression with exposure( ) and then without the exposure( ) but including the variable that would normalize exposure.
[Please send your comments, contested views at mrh4478@gmail.com]
No comments:
Post a Comment