I’ve written before about my use of glysophate (or RoundUp if you prefer the brand name) in my garden. Although I have since moved from Denmark to the Netherlands, the stuff is still in my shed. I very occasionally use it for those few long-rooted weeds that are nigh impossible to remove by hand because they’re stuck between a wall and the pavement. Yes, it’s potentially carcinogenic, but so are coffee, red meat and alcohol, and as with those: the dose matters. Moreover, the alternatives aren’t all that positive either, as I pointed out in that post.

A recent study in this regard has been raising steam in the Netherlands (1) and abroad (2). The main headline it’s been making in the Netherlands is that the use of glysophate in your garden reduces the abundance of house sparrows in gardens by about 25%. Another pesticide against snails (one I do not use incidentally) seems to reduce their abundance by 40%. As a consequence, Arnold van Vliet, a well-known and respected biologist and citizen scientist in the Netherlands calls on us to stop using those substances, and his post on LinkedIn is being liked and reposted. Although I am not necessarily against stopping these the use of these substances my economist brain was tickled: simply stopping seems to me to ignore a bunch of human adjustment effects. Moreover, I wondered what it was that the study had actually investigated, so I decided to read it in full, rather than the summaries in the news articles or just the abstract.

The study

I’ll be the first to admit that it’s a very thorough quantitative study, in fact one I would have loved to conceive and carry out myself. Jealous? Not really, just appreciation. So what did they do? The authors sent a questionnaire to participants in a British garden bird recording scheme, asking them about the surroundings of their garden, their garden and whether they used pesticides, and if so what ones. They matched the participants with the weekly recorded birds in their garden, and then carried out regressions trying to predict the total number of bird species in a garden, or the total number of birds, or the total number of birds of a certain species based on a number of predictors such as urban or rural garden, garden quality, surrounding area quality, presence of pets, latitude and the use of certain pesticides.

There are two things that bug me a bit about the study in the way they have built up their models. First of all their selection of predictors. The paper states:

These variables were tested, using linear models (LMs), against several
independent variables, along with all relevant interactions (see appendix
E). The full model for each dependent variables included all independent
variables and interactions listed in appendix E. […]
We established the most suitable model for each response variable by
successively removing explanatory variables from the complete model,
starting with interactions and the least significant terms, with the best
model being the one with the lowest Akaikes’s Information Criterion (AIC
based model selection)

Tassin de Montaigu & Goulson, 2023, p. 3

Model selection is a murky topic, and economists are also guilty of trying different model types (see e.g. here for an extreme example), but successively removing predictors based on significance alone is frowned upon because it generates a bunch of problems, among others too low p-values and biased parameter estimates. Admittedly that may not exactly be what the authors have done as they also used Akaike’s Information Criterion, but I’m not convinced that is much better.

The second is the use of different predicted variables. All in all they used the following variables as predicted variable:

  1. Total number of bird species in garden
  2. Total average number of birds in garden
  3. Average number of birds of the 10 most common species in garden
  4. Average number of the blue tit (Cyanistes caeruleus) in garden
  5. Average number of the great tit (Parus major) in garden
  6. Average number of the common blackbird (Turdus merula) in garden
  7. Average number of the European robin (Erithacus rubecula) in garden
  8. Average number of the common chaffinch (Fringilla coelebs) in garden
  9. Average number of the greenfinch (Chloris chloris) in garden
  10. Average number of the collared dove (Streptopelia decaocto) in garden
  11. Average number of the dunnock (Prunella modularis) in garden
  12. Average number of the starling (Turnus vulgaris) in garden
  13. Average number of the house sparrow (Passer domesticus) in garden

Take into account further that they applied model selection techniques on all of these models and you understand that they must have run a LOT of models (although probably not two million).

If I look at the full set of reported results a rather mixed bag comes out. The most consistent is the fact that birds are less abundant in urban and sub-urban gardens than in rural gardens, and that there is a relatively strong effect of having a bird-friendly garden and bird-friendly surroundings.

When it comes to pesticides the results are a lot less clear cut. On the first three predicted variables there seems to be very little direct effect of pesticides on the presence of birds. They did find a negative interaction effect between surrounding quality and pesticide use on species richness, meaning that surrounding quality raises total number of species in a garden, but less so if pesticides are used. There is also a complex interaction with urbanisation on total average abundance, where in some cases of urbanisation, pesticide use seems to raise total number of birds, and in other cases it reduces it.

For the species specific results they found a direct effect for house sparrows only, when looking at pesticides overall, but a bit more when looking at specific pesticides (there were in total 70 combinations, with 10 negative and two positive). The most noteworthy was the aforementioned negative relation between the use of glysophate or metaldehyde and abundance of sparrows.

Of all these results and models what made it to the Dutch news was only the effect of glysophate and metaldehyde on house sparrows. Given the sheer number of models tested to me that sounds a lot like the following comic by XKCD:

The economics

All right, enough statistical muttering, let’s go back to economics. Suppose that despite my reservations above I have enough faith in the results and the use of pesticides indeed lowers the presence of (certain) birds in our garden. Would simply stop using them be the solution then?

Putting aside the question as to whether or not we actually want a bird increase in our gardens (I cannot help but wonder if the same outcry would have taken place if this had been about say gulls or doves, aka flying rats in Amsterdam), which is yet another difficult question what would happen if we stop using these pesticides?

Note that the study doesn’t make any claims about toxicity or that the use of pesticides causes a decline in bird numbers. Given its set-up it cannot even do that. That would require an experiment or a quasi-experiment if available. The decline in birds may be caused by the use of pesticides because they are toxic to birds, but also by a third variable that correlates with both the use of the pesticide and the birds presence, e.g. through food availability or garden tidiness. From a theoretical point of view we could even have reverse causality (the decline of bird presence causes us to use more pesticide) although from a practical point this seems unlikely.

The more important part, though, is to ask why people use these pesticides. The answer is: to get rid of certain plants (for glysophate) or snails (for metaldehyde). Stopping the use of pesticides is not going to remove the desire to get rid of these plants and snails and as a consequence people will look for alternatives. This is where the earlier point of causality comes in. If the cause is toxicity then the question becomes if the alternative measures are not equally or more toxic. If the cause is food availability or tidiness then if snails or weeds get removed by hand or other means we’ll still see a lower bird abundance in those gardens. Admittedly it may be a bit more because alternative means to remove plants and snails may not be as effective – or more costly and thus not applied at the same levels.

What then is the solution? Stopping pesticide use in my mind is a 2nd best solution. I would try to remove the desire for weed and pest free gardens. Tell the story that weeds and snails are very necessary for the presence of birds as they provide food and perhaps shelter. In that way you kill two birds with one stone (terrible pun, I’m sorry): you get better birds friendly gardens and you get rid of the pesticides.

How to achieve that though is a different matter, and would probably require more research. All in all I think it at least means that a lot can be gained by environmental and social scientists working together.

Cover image by Beverly Buckley from Pixabay