I recently completed an analysis for TRAFFIC International to describe trends in the illegal ivory trade from 1996 – 2014. These results are being presented this week at the CITES Standing Committee as part of the document on “Elephant conservation, illegal killing and ivory trade” (in particular see Figures 5-9).
Clearly, it is the results that are of most interest to the audience (of policy makers and NGOs) to help inform decision making about elephants but here I’d like to reflect on the analysis.
The analysis uses quite a complex statistical methodology to turn data about seizures of illicit ivory shipments (from the Elephant Trade Information System) into information about trends in the illegal ivory trade. The methodology is complex because of the inherent biases in seizures data. Specifically, countries differ in their ability to make and report seizures – so an increase in seizures might be because of increased law enforcement not because trade is increasing (as described in this blog post, this set of videos, this PLoS One paper and, if you are a statistician and want more detail, in this Technical Report).
The methodological framework was developed with another statistician (Bob Burn) as part of Darwin Initiative Project 17-020 and we first used it to analyse ivory seizures data for the CITES Conference of the Parties (CoP) in 2013 using data from 1996 – 2012. In this most recent analysis which includes data up to 2014 we use the same methodology and refitted the models so that the trend could be updated with the additional years of data. We did the same for a previous analysis with data up to 2013. For the next CoP analysis the plan is to use the same general methodological framework to update the trend (with 2015 data) and review the variables which account for differences in the seizure and reporting rates between countries and over time.
The complex methodology is becoming more routine
What I find pleasing is that the methodological framework is becoming more routine in its use which is great as this was one of the aims of our Darwin project. This routine use, or standardisation of the methodological approach, has a number of advantages:
The first is that the ability to “turn the handle” and update the results relatively straightforwardly helps given that these analyses are almost always carried out under quite tight time pressure. Note that we are still thinking carefully about what needs to be done and it still takes (quite a bit of) time and a statistician to obtain the results but it is much quicker than if we had to create a new strategy every time data is added,
Second, because the method is being used more routinely people are more familiar with it. Similarly the reporting of results (in this case the Transactions and Weights Indices) is also familiar and so people can get straight on to understanding the message. Note that I am not saying that the indices are difficult to understand just that familiarity can aid understanding. In fact we spent a long time making sure that we outputs that are simple to understand, even though the methodology to get these outputs is complex.
Third, because the methods are routine and have already been documented and peer-reviewed then the reports do not need to contain lengthy annexes detailing the methodology. This means that people don’t need to spend time and resources reviewing them or worrying about them and debating their legitimacy.
These are great pay-offs for a methodological framework that we thought about and developed over several years. None of this means that we are becoming complacent or that the methodology remains fixed for ever. There is always space to make minor adjustments and learn and improve on previous analyses within the existing framework. As statistical methods and computational power increases others may develop alternative methods to provide different insights into the data. But we have a solid foundation and framework to build on and a strategy for producing standard analyses for reporting purposes whilst new methodologies and adaptations and improvements of the existing methodology can be carried out alongside this.