Demining: good for increasing economic productivity, cost-ineffective for avoiding injury
If you are interested in funding demining work, i.e. work to
clear landmines and similar ordnance, there are two possible motivations:
Outcome
|
Cost-effectiveness
|
Metric
|
Increase economic
productivity
|
Reasonable
|
20% rate of return on investment
|
Prevent injury or death
|
Poor
|
$450,000 per person saved from injury
|
0. About demining and landmine clearance
The two highest profile UK-based landmine-clearing charities are Halo Trust and MAG, and the analysis here is largely based on Halo Trust.
1. Increasing economic productivity
Comparative analysis
The social return on investment (i.e. amount of $ going to the beneficiaries per $ invested by the donor) is around 20% per annum. Note that there are a number of uncertainties around this number, which are explained below. First a couple of comparisons:
Note that there is sufficient uncertainty in the calculations that Landmine clearance, so it should be considered in the same ballpark as Bees for Development.
Bees for development is a charity which trains people in the developing in apiculture so that they can sell the honey. The SROI of c 30% is based on assumptions provided by the charity themselves.
Calculations and assumptions
In coming up with this figure the following calculations were made. Note that several of the inputs are supported with very little evidence, meaning that the overall confidence in the figure should be low.
Notes on assumptions
(A) Minefield area cleared + Battle Area Cleared (source: Halo Trust 2014-15 accounts: http://apps.charitycommission.gov.uk/Accounts/Ends13/0001001813_AC_20150331_E_C.pdf)
(B) Total cost of running Halo Trust in 2014-15 was £27,607,000 (source accounts http://apps.charitycommission.gov.uk/Accounts/Ends13/0001001813_AC_20150331_E_C.pdf) This has been converted from £ to $
(C) The cost of the work is driven by the amount of area cleared, rather than the number of mines (although estimates are also available for the cost per mine cleared)
(D) Calculated based on ratio of GDP to acres -- so this represents the average. This is not the same as the marginal effect of an extra acre. Note that an important assumption here is that people actually *do* start using the land again -- this will only happen if there is trust that the demining work has been effective.
(E) Assumes diminishing marginal returns on land. In particular, assumes that returns will diminish quite substantially in this context, since in many countries, lots of the GDP is focused on high-density GDP generating areas in large cities, and the rate at which economic productivity will arise will be much lower in these more rural areas.
(F) Covers things like paths being cleared giving children a shorter walk to school
(G) Assumes that when impoverished people are competing for resources against people who have more power, the less powerful people are likely to lose. Also assumes that the corruption effect applies often in the post-conflict areas where landmines are found, but not always
(H) Assumes that the charity doing the work to demine will work directly with local communities, and check in on those communities to ensure that the land ends up in the right hands. Halo Trust, for example, state that they do this. Also assumes that this process is not always successful
2. Preventing injury and death
Comparative analysis
The cost per injury averted is high. For comparison, GiveWell estimates the cost per life saved at below $10,000 for the Against Malaria Foundation; this estimate puts the cost per injury averted at $200,000 at a minimum, and on average probably over $400,000. Note also that more than half of those injuries will not lead to death, so using this comparison is in fact favourable to the landmine charity.
(A) Stated consistently in several different internet sources:
http://www.care.org/emergencies/facts-about-landmines,
https://www.unicef.org/sowc96pk/hidekill.htm
http://www.landminefree.org/2017/index.php/support/facts-about-landmines http://www.un.org/en/sections/issues-depth/demining/index.html(B) Different sources on the internet give different answers for this, however most are higher than this (c 15,000-20,000) and when traced back, they date to a period some decades ago when mine injury incidence rates were higher. The most credible source seems to be http://the-monitor.org/en-gb/reports/2016/landmine-monitor-2016/casualties-and-victim-assistance.aspx, which gives the number of casualties as 6,461. We then multiplied this by 1.25 to get the figure used. This is because the site says: ""As in previous years, the mine/ERW casualties identified in 2015 only include recorded casualties, not estimates. Based on the Monitor research methodology in place since 2009, it has been estimated that there are up to approximately 25–30% additional casualties each year that are not captured in the Monitor’s global mine/ERW casualty statistics, with most occurring in severely affected countries and those experiencing conflict. ""
(C) Straight calculation of the ratio
(D) The situation is a little bit complex -- for a landmine, the lifetime is several decades until the battery runs out, but when the battery runs out, it's not that the probability of contact with the mine causing injury goes to zero, it's just that it reduces significantly. However it's long enough that assuming that any reasonable level of discounting is happening, it is a reasonable approximation to treat it as indefinite.
(E) Several internet sources give the cost as somewhere in the range $300-$1,000, e.g.
https://www.unicef.org/sowc96pk/hidekill.htm
http://www.landminefree.org/2017/index.php/support/facts-about-landmines
(F) Discount rate: this is intended to capture the fact that if a benefit arises at some point in the future, it would be possible to invest the money first, and then disinvest it and achieve other good things in the future, but have more money available to do so. It also reflects the reality that we are less confident in the accuracy of the assumptions the further out we go; in particular the relevant assumption here is the propensity for a mine to injure someone. (G) The is calculated as the sum of an infinite series of the form p + p[q/(1-r)] + p[q/(1-r)]^2 + p[q/(1-r)]^3 + ..., where p is the probability a mine will activate/injure someone, q=(1-p), r = discount rate. The most accurate formula for this is = (C)/(1-(1-(F))(1-(C))). However for small (C), this is indistinguishable from (C)/(F).
(H) As can be seen by following through the calculations, this is more accurately referred to as the expected present value of the cost per person saved.
Acknowledgements: Thanks to Matt Neave, who provided some useful input based on actual experience in the field.
Comparative analysis
The social return on investment (i.e. amount of $ going to the beneficiaries per $ invested by the donor) is around 20% per annum. Note that there are a number of uncertainties around this number, which are explained below. First a couple of comparisons:
Option
|
Notional donation amount
|
Beneficiary receives
|
Confidence
|
Landmine clearance
|
£100 now
|
£20 per annum
|
Very Low
|
Give Directly
|
£100 now
|
£91 now
|
High
|
Bees for development
|
£100 now
|
£30 per annum
|
Low
|
Note that there is sufficient uncertainty in the calculations that Landmine clearance, so it should be considered in the same ballpark as Bees for Development.
Bees for development is a charity which trains people in the developing in apiculture so that they can sell the honey. The SROI of c 30% is based on assumptions provided by the charity themselves.
Calculations and assumptions
In coming up with this figure the following calculations were made. Note that several of the inputs are supported with very little evidence, meaning that the overall confidence in the figure should be low.
Notes on assumptions
(A) Minefield area cleared + Battle Area Cleared (source: Halo Trust 2014-15 accounts: http://apps.charitycommission.gov.uk/Accounts/Ends13/0001001813_AC_20150331_E_C.pdf)
(B) Total cost of running Halo Trust in 2014-15 was £27,607,000 (source accounts http://apps.charitycommission.gov.uk/Accounts/Ends13/0001001813_AC_20150331_E_C.pdf) This has been converted from £ to $
(C) The cost of the work is driven by the amount of area cleared, rather than the number of mines (although estimates are also available for the cost per mine cleared)
(D) Calculated based on ratio of GDP to acres -- so this represents the average. This is not the same as the marginal effect of an extra acre. Note that an important assumption here is that people actually *do* start using the land again -- this will only happen if there is trust that the demining work has been effective.
(E) Assumes diminishing marginal returns on land. In particular, assumes that returns will diminish quite substantially in this context, since in many countries, lots of the GDP is focused on high-density GDP generating areas in large cities, and the rate at which economic productivity will arise will be much lower in these more rural areas.
(F) Covers things like paths being cleared giving children a shorter walk to school
(G) Assumes that when impoverished people are competing for resources against people who have more power, the less powerful people are likely to lose. Also assumes that the corruption effect applies often in the post-conflict areas where landmines are found, but not always
(H) Assumes that the charity doing the work to demine will work directly with local communities, and check in on those communities to ensure that the land ends up in the right hands. Halo Trust, for example, state that they do this. Also assumes that this process is not always successful
2. Preventing injury and death
Comparative analysis
The cost per injury averted is high. For comparison, GiveWell estimates the cost per life saved at below $10,000 for the Against Malaria Foundation; this estimate puts the cost per injury averted at $200,000 at a minimum, and on average probably over $400,000. Note also that more than half of those injuries will not lead to death, so using this comparison is in fact favourable to the landmine charity.
Calculations and assumptions
In coming up with this figure the following calculations were made. This is moderately well-evidenced by the standard of most assessments for charitable interventions.
In coming up with this figure the following calculations were made. This is moderately well-evidenced by the standard of most assessments for charitable interventions.
(A) Stated consistently in several different internet sources:
http://www.care.org/emergencies/facts-about-landmines,
https://www.unicef.org/sowc96pk/hidekill.htm
http://www.landminefree.org/2017/index.php/support/facts-about-landmines http://www.un.org/en/sections/issues-depth/demining/index.html(B) Different sources on the internet give different answers for this, however most are higher than this (c 15,000-20,000) and when traced back, they date to a period some decades ago when mine injury incidence rates were higher. The most credible source seems to be http://the-monitor.org/en-gb/reports/2016/landmine-monitor-2016/casualties-and-victim-assistance.aspx, which gives the number of casualties as 6,461. We then multiplied this by 1.25 to get the figure used. This is because the site says: ""As in previous years, the mine/ERW casualties identified in 2015 only include recorded casualties, not estimates. Based on the Monitor research methodology in place since 2009, it has been estimated that there are up to approximately 25–30% additional casualties each year that are not captured in the Monitor’s global mine/ERW casualty statistics, with most occurring in severely affected countries and those experiencing conflict. ""
(C) Straight calculation of the ratio
(D) The situation is a little bit complex -- for a landmine, the lifetime is several decades until the battery runs out, but when the battery runs out, it's not that the probability of contact with the mine causing injury goes to zero, it's just that it reduces significantly. However it's long enough that assuming that any reasonable level of discounting is happening, it is a reasonable approximation to treat it as indefinite.
(E) Several internet sources give the cost as somewhere in the range $300-$1,000, e.g.
https://www.unicef.org/sowc96pk/hidekill.htm
http://www.landminefree.org/2017/index.php/support/facts-about-landmines
(F) Discount rate: this is intended to capture the fact that if a benefit arises at some point in the future, it would be possible to invest the money first, and then disinvest it and achieve other good things in the future, but have more money available to do so. It also reflects the reality that we are less confident in the accuracy of the assumptions the further out we go; in particular the relevant assumption here is the propensity for a mine to injure someone. (G) The is calculated as the sum of an infinite series of the form p + p[q/(1-r)] + p[q/(1-r)]^2 + p[q/(1-r)]^3 + ..., where p is the probability a mine will activate/injure someone, q=(1-p), r = discount rate. The most accurate formula for this is = (C)/(1-(1-(F))(1-(C))). However for small (C), this is indistinguishable from (C)/(F).
(H) As can be seen by following through the calculations, this is more accurately referred to as the expected present value of the cost per person saved.
Acknowledgements: Thanks to Matt Neave, who provided some useful input based on actual experience in the field.