Produce Pesticide Rankings

My post Details on the Dirty Dozen on EWG’s Shopper’s Guide to Pesticides™ led me to dive into the 2008 USDA data to see just how contaminated (or not) our produce really is. There’s so much information that it’s a little difficult to work with, but with perseverance and the right software (JMP is the best!*), I was able to re-do the EWG analysis but with the newest available data.

Below you can find my results with a through explanation of what I’ve done and why. The results are posted without all the commentary at Produce Pesticide Rankings which has all of the results and Pesticide Produce Rankings Tables which has comparisons of my results to the EWG results. You can download the original USDA data yourself or check out the Latest PDP Findings of Interest to Consumers.

Concentration and LOD

My first step was to compare the detected Concentration to the Limit of Detection. The LOD seems to have been ignored by EWG. The LOD is the smallest concentration of the chemical you are looking for that will give a positive signal with the method used. Every method/chemical combination has a different LOD that can be found by comparing a blank (no chemical) to smaller and smaller concentrations of the chemical. If the detected concentration is at or below the limit of detection, it does not indicate the chemical is present – which is not the same as saying the chemical is not present. The chemical could be there, but the amount is so small that it can not be detected with the method being used.

Let’s put some numbers on it. There were 1,780,365 tests conducted on 13,381 samples (including fruits, vegetables, fish, nuts, and water), with 33,426 of those tests having a concentration listed (1.88%). Of those, 273 were equal to the LOD leaving 33,153 positive concentrations (1.86%). Not a big difference, but still, it would be incorrect to include the concentrations that are below the LOD. In a lot of experiments a blank is subtracted from the results and I can’t think of a reason why that wouldn’t be appropriate here. So, I created a column of Concentration minus LOD and used these numbers for my calculations.

One drop of water is 2 ppm of a bathtub full of water. Image from the Alaska Department of Environmental Conservation.


Most of the tests have a unit of ppm (parts per million), but some are ppb (parts per billion) or ppt (parts per thousand). I converted ppb to ppm (ppb/1000=ppm) and ppt to ppm (ppt*1000=ppm) so all of the average residue values would be in the correct units.

It would be inappropriate to average values with different units. As illustrated by the Alaska Department of Environmental Conservation, ppm is drops per bathtub while ppb is drops per swimming pool!

Comparing the 2008 data with EWG

Because this investigation was inspired by EWG, let’s go through their Spreadsheet column by column to compare the top five values of each. You can find this information in Pesticide Produce Rankings Tables. The 3 types of water tested by USDA top most of the lists in the 2008 data, but since this discussion is on produce, they aren’t included here.

Percent of samples tested with detectable pesticides

This isn’t really a good metric because it doesn’t take into account which of the detected residues are above or below the EPA tolerance level and the EWG numbers don’t take the LOD into account (the numbers I report are all Concentration – LOD), but nonetheless here’s how they stack up.

  • % of samples with 1 or more residues: 95.78 Peaches, 95.55 Celery, 95.24 Nectarines, 94.06 Strawberries, 92.75 Catfish.
  • % of samples with 2 or more residues: 89.74 Celery, 88.66 Strawberries, 86.04 Peaches, 80.65 Nectarines, 72.22 Blueberries.
  • EWG % of samples tested with detectable pesticides:  97.20 Plums, 96.20 Peaches, 95.10 Bell Peppers, 95.00 Celery, 93.60 Apples.
  • EWG % of samples with two or more pesticides: 85.70 Peaches, 84.70 Celery, 82.30 Blueberries, 80.60 Bell Peppers, 74.40 Apples.

As you can see, the percentages don’t vary much from the collection of data used by EWG to the 2008 only data. Some of the foods tested in previous years weren’t tested in 2008 (apples, bell peppers).

A lot of the samples for each commodity have 1 residue, fewer have 2, fewer have 3, and so on. For some perspective, consider the percentage of all tests done on all samples for each commodity that had one or more residue.

  • % of tests with 1 or more residues: 1.18 Nectarines, 0.92 Collard Greens, 0.90 Summer Squash, 0.83 Kale, 0.79 Almonds.

Average number of pesticides found on a single sample

This is a little more useful than the percent of samples with one or more residues, but not by much, since we’re still leaving out consideration of the EPA tolerance.

  • Mean residues detected per sample: 5.15 Celery, 4.94 Strawberries, 3.61 Blueberries, 3.50 Peaches, 2.46 Spinach.
  • EWG Average number of pesticides found on a single sample: 3.79 Celery, 3.08 Peaches, 3.00 Blueberries, 2.90 Strawberries, 2.75 Apples.

The USDA lets us know in their Latest PDP Findings of Interest to Consumers that the number of samples with pesticides and number of pesticides per sample doesn’t correlate to pesticides per serving size because the sample sizes were a lot more than a serving. “Sample size ranges from 16 ounces to 5 pounds depending on food tested. For example, for peaches and celery, the sample size is 5 pounds; for strawberries and blueberries is 3 pounds and 1 pound respectively.”

In regards to number of pesticides per sample, the USDA states: “There may be many more pesticides available for use by food producers, but 20 years of testing show that no food has ever been treated with all available pesticides.”

Average amount of pesticides found in ppm

This might be the worst metric of all because it averages pesticides that have very different toxicity levels. One ppm of one pesticide can be very different from one ppm of another pesticide! Still, here’s where we start to see some real differences!

  • Mean ppm residue by commodity: 0.8 Potatoes, 0.61 Spinach, 0.37 Rice, 0.35 Nectarines, 0.33 Sweet Potatoes.
  • EWG Average ppm of all pesticides found: 1.602 Potatoes, 1.373 Spinach, 1.200 Plums, 1.066 Peaches, 0.906 Red Raspberries.

The EWG shows average ppm of pesticides that are twice what I’ve got from the 2008 data! What’s happening here? One possibility is that EWG didn’t convert the ppt to ppm, but surely they’d notice the different units in the data, so it must be something else. We could have done the averages differently, but that’s unlikely too, it’s just averaging.

The only other thing I can think of is that there were high levels of residues in the past, high enough to skew the overall averages. If this is true, then we have something to celebrate – there have been great reductions in pesticide residues over the years!

Still, this brings up a question: why would the EWG tell people that produce has such high amounts of pesticide residues when produce today actually has much less? If the goal is to tell people what are the safest foods to buy for their families today, why include old data?

The USDA states specifically in their Latest PDP Findings of Interest to Consumers that there have been significant changes over the years, with reduced number of samples with pesticides and reduced ppm of pesticides. Specifically, there have been reductions in the most harmful pesticides as safer alternatives have been approved for use.

Maximum number of pesticides found on a single sample

Again, this metric does not take the EPA tolerances into consideration, and the results are about the same..

  • Maximum residues detected per sample: 14 each Strawberries and Celery, 12 Blueberries, 11 Catfish, 10 each Spinach, Collard Greens, and Peaches.
  • EWG Maximum number of pesticides found on a single sample: 13 each Blueberries, Strawberries, and Celery, 11 Bell Peppers, and 10 Kale.


*Thanks to my husband for explaining that it makes a lot more sense to keep the test data and the sample data in two separate tables that you join when needed based on the sample number. Having all the data in one JMP file is about 8mB which doesn’t work all that well even on a good computer.