Humane Insecticides: A Cost-Effectiveness Calculation

Abstract. It is not known whether insects can feel pain, but even if we assign this possibility a low probability, the large number of insects alive makes the expected value of their suffering considerable. In particular, in this piece I investigate the cost-effectiveness of trying to reduce the expected amount of suffering by agricultural pests through promotion of less painful insecticides. While I don't necessarily think pesticides cause net harm, since they may prevent the existence of lots of insects that would endure net suffering in the wild, either way, killing insects more humanely would prevent suffering. An inefficient brute-force strategy would be to directly pay farmers to use pesticides that cause faster and more benign deaths. Under plausible assumptions, each dollar used for this purpose would prevent the expected equivalent of 250,000 experiences of being killed by the original, more painful pesticide. Of course, funding research or advocating policy changes would likely be more cost-effective than this direct approach.

Background

The evidence is mixed on the question of whether insects can feel pain. A fair number of papers have investigated this topic, but two in particular that have good reviews of the literature are Jane A. Smith, A Question of Pain in Invertebrates, ILAR Journal, and Jeffrey A. Lockwood, The Moral Standing of Insects and the Ethics of Extinction, Florida Entomologist (JSTOR link). As Smith notes, "The well-being of invertebrates used for research is being taken increasingly seriously," with V.B. Wigglesworth, "Do Insects Feel Pain?", Antenna, suggesting that we should assume insects can suffer unless we have evidence proving otherwise. I would discourage this precautionary-principle approach in favor of a more conservative Bayesian expected-value approach, but it seems clear to me that it would be wrong to completely ignore the possibility of insect pain until we have more information.

Suppose we decide on a relatively conservative probability that insects feel pain of 0.01. At any given moment, the earth contains 10^18 ("a billion billion") insects, so even multiplying by 0.01, there are 10^16 "expected insects." Considering that insects live very short lives, and that most insect offspring die long before reaching maturity, it's plausible that insect lives contain far more suffering than happiness. See "The Predominance of Wild-Animal Suffering over Happiness: An Open Problem (pdf)." Given the difficulty of finding a satisfactory solution to wild-insect suffering, it may be tempting to throw up one's hands and say, "Yes, wild insects probably endure a lot of unnecessary pain, but there's nothing we can do about it." This may or may not be the case, but either way, such an attitude ignores areas in which we could clearly make a net positive reduction in insect suffering--regardless of whether wild insect lives are worth living.

As suggested by Wigglesworth, one area where insects may suffer preventably is in the laboratory; Smith reviews principles for improving care and treatment of such insects. But the numbers of insects involved here is relatively small by comparison to, say, the insects killed by pesticides. One's immediate reaction to this fact might be to try to limit use of pesticides on farms and lawns so that humans kill fewer insects. However, if insect lives aren't worth living, this may be precisely the wrong thing to do, since the insects killed by pesticides would have died in other (probably painful) ways, and pesticides prevent--at least temporarily--the existence of new insects that would otherwise have lived miserable lives. Of course, promoting increased pesticide use may be problematic on account of human-health impacts, as well as perhaps on religious or other (non-utilitarian) philosophical grounds.

The issues raised above need to be explored, but they do not automatically imply that we should throw up our hands and move on to other issues. In the remainder of this piece, I'll explore the cost-effectiveness of trying to promote use of pesticides that kill insects more quickly and less painfully. While quite arguably not optimal, this approach to relieving insect suffering avoids the thorny issue of whether insect lives are worth living and whether we should kill more or fewer of them. As such, the numbers here should represent a lower bound on the cost-effectiveness of using resources to address insect pain. If these numbers compete in cost-effectiveness with other possible uses of time and money, then even more efficient measures to reduce insect suffering could only be more urgent.

The pesticides referred to here needn't be chemical in nature; they could, for instance, be biological control agents, often used in organic and integrated-pest-management systems. To give one example: Ichneumon wasps are sometimes used to control flies and beetles. However, these parasites eat their prey while still alive out from the inside over an extended period. If the hosts do feel pain during this process, it is likely severe and protracted, which suggests trying to dissuade farmers from using this method. On the other hand, the fact that the host doesn't respond aversively during the process (I don't think) may imply that it lacks pain receptors to its internal organs, similarly to the way in which human viscera have few such neural connections. (This is just a conjecture on my part.) In this case, perhaps ichneumon wasps would be a humane alternative, unless the final stages of consumption of the host, affecting its external parts, is still painful. Ichneumon wasps are just one of several pest-control parasites that consume their prey in the same fashion -- and which may therefore be potential targets for replacement.

Approach 1: Paying Farmers to Use Different Pesticides

Suppose we think that pesticide A causes death to the insects that it kills twice as quickly as pesticide B and that the two pesticides are equally painful per unit time. Suppose also that both are equally potent. If farmers currently use pesticide B because pesticide A is more expensive, then we could set up a program to compensate farmers for switching to use of pesticide A. This scheme seems highly inefficient by comparison to other approaches, but it is direct and avoids concerns that some might have about whether simply doing research or advocacy would really make any difference.

Define the following variables:

• p: our subjective probability for how likely it is that insects feel pain. This is likely to vary from species to species, but I ignore that complication here.
• n: the average number of insects (and other small sentient organisms, if any) that inhabit a hectare of crop land during the growing season. In practice, this is likely to vary from one type of crop to another and one region to another.
• f: the fraction of the above insects and other organisms that are killed (or otherwise suffer significantly) as a result of applying pesticides.
• s: the average ratio of the painfulness of the more humane pesticide to the less humane pesticide (e.g., s = 0.5 in the previous example, because pesticide A is half as painful as pesticide B). The percent by which suffering of death is reduced by using the more humane pesticide is (1-s)*100%.
• c: the dollar amount by which it's more expensive to spray a hectare of cropland once with the more humane pesticide rather than of the less humane one.

Suppose we spend c dollars to compensate a farmer for using a more humane pesticide on a single hectare. The expected number of insects benefiting is pnf, and they each avoid suffering an amount of suffering equal to 1-s times the painfulness of death by the more painful pesticide. Our expected reduction in suffering per dollar R is then

R = pnf(1-s)/c

expressed in units of "expected equivalent number of experiences of death by the more painful pesticide prevented per dollar."

Next, I suggest plausible subjective probability distributions for each of the five parameters. With that, I compute a numerically approximated probability density function for R. (Note that it's formally incorrect to assign a probability distribution to p, but I wanted the results to convey a sense of how cost-effectiveness changes depending on whether p is set high or low.) Readers can recompute the results using their own subjective probabilities using this Excel workbook.

• ln p is uniform on [-6, -0.5]

  I think 0.6 ~= exp(-0.5) is a high estimate for p. 0.002 ~= exp(-6) seems like a low estimate.

• ln n is normal with mean 15 and standard deviation 2

  An expert friend of mine estimates that a hectare of insect-infested crop land would have between 500,000 (natural log ~= 13) and 50,000,000 (natural log ~= 17) insects, depending on the type of crop and type of insect.

• f is normal with mean 0.8 and standard deviation 0.05

  The same friend estimates 0.8 as reasonable. "The Estimation of Insect Density and Instar Survivorship Functions from Census Data" by Martin Birley (JSTOR link) assumed 10-20% survivorship when a sugar-cane pest was sprayed with residual insecticide (p. 503).

• s is uniform on [0.3, 0.9]

  This is speculation.

• c is uniform on [1, 8]

  I don't have excellent data on this. What I did was to look at the cost of "chemicals" per acre for various types of crops using the USDA's surveys of Characteristics and Production Costs. Roughly, chemical costs per acre for wheat ranged from ~$3 to ~$9, for soybeans from ~$22 to ~$29, and for corn from ~$20 to ~$30. I'll assume a uniform distribution between $3 and $30. But this is in $/acre. Noting that one acre is 0.4047 hectares, this becomes a uniform distribution between $7.4 and $74. It's not clear how many chemical sprayings per year this represents; I assume 1.5. It's also not clear how many other chemicals (fertilizers, herbicides, fungicides) this includes; I'll assume 1/3 of all chemicals are insecticides. I also assume that switching to the more humane insecticide would increase total chemical costs by 50%. The result, after liberal rounding, is a uniform distribution on [1, 8].

The expected number of death-experiences prevented per dollar is around 250,000. If death by pesticides is as bad as spending, say, 2 days in a factory farm, this is like preventing 1,369 years of suffering for animals in factory farms per dollar. The cost to prevent a year of suffering is then $0.0007. This competes strongly with donating to Vegan Outreach (at least if only direct suffering of factory-farmed animals is considered); in that case, the cost to prevent a year of suffering is between $0.02 and $3.65.

What would the figures be if a very conservative value were used for p? Take p = 0.001. Then the expected value is roughly 2,100, which still translates into a cost of $0.09 to prevent a year-equivalent of factory-farm suffering.

Approach 2: Fund Research on More Humane Pesticides

[not completed]