Steve Savage: “A Rational Analysis of the USDA Pesticide Residue Data”

A guest post by agricultural scientist Steve Savage,
proprietor of Applied Mythology.
(Updated 1/6/16 to reconnect missing illustrations. This post first appeared on Sustainablog 6/15/11)


When the Environmental Working Group (EWG) makes its annual “Dirty Dozen List” of fruits and vegetables with pesticide residues, it does so without paying any attention to which chemicals were found or what level was detected. This is why it is so misleading.  To do the analysis properly does take a lot more “work” – it took me much of the last two days to do it.

First I had to download the raw data which comes as a 5.5 MB ZIP File that expands to a 83 MB text file.  My son wrote a little Ruby On Rails script that sifts through the millions of rows of data to find the 30,000 actual “detections” of pesticide residues that the USDA reported for 2009.  That list has the identity of the pesticide and its concentration in parts per million, billion or sometimes trillion.  Next, I searched for anMSDS for each of the 300 or so different chemicals to get the specific acute toxicity (this is usually in section eleven of each document).  The acute toxicity is expressed as an LD50 – the milligrams of chemical it would take per kilogram of body weight to kill 1/2 of the rats in a feeding study (Oral LD50).  These are publicly available documents which are usually easy to find except for old, discontinued pesticides and some of the metabolites.  Dividing the LD50 by the detected amount gives you the multiple of its own body weight that the rat would have to eat to reach a toxic dose.

An Example You Can Blame on McDonald’s

As an example, an old, extremely toxic pesticide, aldicarb (Temik) has an LD50 of 1 mg/kg.  This is exactly what you imagine when you hear the word, “pesticide.”  In one sample of fresh potatoes, the USDA scientists detected 0.01 mg/kg of aldicarb sulfoxide – a metabolite which is just as toxic as the aldicarb.  For the rats to die from eating such potatoes would require that they rapidly consume 90 times their own weight of those particular potatoes.   The most toxic potato sample had 1.5mg/kg of the aldicarb sulfoxide which means that the rates could die by eating just one times their own body weight.  A rat might be able to do that.

The EWG essentially treats every one of the 30,000 detections as equal in risk to these worst-case potato values.  Because most pesticides are far, far less toxic than aldicarb, the average residue found by the USDA on potatoes has a safety margin of 595,163.   The only reason that aldicarb is still used on potatoes (and it will be phased out soon), is that for purely brand protection reasons, MacDonald’s asked it’s fry suppliers not to give them any more GMO potatoes (they had been using them for several years, and they still cook them in GMO soybean oil and serve up GMO sodas with corn sweetener).  Still, McDonalds killed the Bt-potato.  That is why potato growers plant their potatoes into a furrow with granular aldicarb so that the roots pick up the insecticide for ~60 days, protecting them from the Colorado Potato Beetle.  Still, potatoes are in about the middle of the pack in terms of average safety margin.  Oranges have a safety factor of nearly 1.5 million.

Sweet corn, which makes the “Clean Fifteen” list for EWG is actually the crop with the lowest average safety margin (8,909).  This demonstrates the meaninglessness of the Dirty Dozen list.

Beyond Averages

Of course, averages can be misleading.  It is more instructive to look at the full distribution of results.  The graph below summarizes all the sample results for fruit crops.  For this graph, values on the right side of the graph represent extremely low risk while those to the left represent relatively higher risk.  As you can see, even though these crops had many pesticide residues, they almost all were present at vanishingly levels meaning extremely minor risk.  People just don’t eat one hundred, one thousand or several million times their body weight of one food in a sitting!

The same is true for vegetables.  The few residues detected on lettuce have safety margins in the million-fold range and the non-Organic lettuce was actually a little better than the Organic lettuce.  By this methodology potatoes and spinach come out the worst with some safety margins in the thousand range.  Still, the message from the real data is completely different than what one gets from the EWG’s “analysis.”

When the EWG list is reported by the unsophisticated media, they say things about it which are completely false.  For instance said, “The “dirty dozen” list of the twelve fruits and vegetables with the highest amount of pesticide residue was released Monday by the Environmental Working Group (EWG).”  That would be impossible because the EWG did not access the raw data which would be necessary to identify how “high” the residues actually were.  They say that cherries “dropped off the list” without mentioning that cherries were not even one of the crops tested in 2009.

So, what does the USDA data actually tell us?  That we should feel confident that the fruits and vegetables available in our markets are perfectly safe, and we should be consuming them in ever greater quantities to take advantage of all the cancer- and other disease-fighting chemicals they naturally contain.  It also tells us that the EWG should be ashamed of their list, print a full retraction, and refund their ill-gotten financial gains.

If anyone would like a copy of the processed data or graphs I would be happy to email it from  Also if anyone would like to improve on my collection of LD50 values that would be much appreciated.  My website is Applied Mythology.

Crop dusting image from wikimedia

sustainablog (

Steve Savage: “How The USDA Unwittingly Aids EWGs Pesticide Disinformation Campaign”

A guest post by agricultural scientist Steve Savage, 
proprietor of Applied Mythology
(This post first appeared on Applied Mythology 6/19/2012)

Each year, the Agricultural Marketing Service of the USDA (USDA-AMS) conducts an extensive sampling and analysis of items from the actual US food supply to determine what, if any, pesticide residues are present at the consumer level.  This information is published each year, and the actual raw data is also available for download.  The data for 2010 was just recently released.

What Does The Data Actually Tell Us?

For 2010, as for preceding years, the data demonstrates is that pesticide residues are only present at very low levels, usually dramatically below the conservative ‘tolerances’ set during the risk analysis by the EPA.  This is quite remarkable considering that pesticide use on crops depends on many thousands of independent decisions by many thousands of individual farmers both in the US and in dozens of countries from which we import food.

This year the USDA provided several summaries in an effort to be clear about what they have found.  In this years press release one finds the following, unambiguous statements:

The 2010 PDP report confirms that food does not pose a safety concern 
based upon pesticide residues.  
Statement from the EPA ‘The data confirms EPA’s success in phasing- out pesticides used in children’s food for safer pesticides and pest control techniques.  The very small amounts of pesticide residues found in the baby food samples were well below levels that are harmful to children.’  
Statement from FDA: ‘Based on the PDP data from this report, parents and 
caregivers can continue to feed infants their regular baby foods without being concerned about 
the possible presence of unlawful pesticide chemical residues.’ 

Statement from the USDA: ‘Age-old advice remains the same: eat more fruits and vegetables and wash them before you do so.  Health and nutrition experts encourage the consumption of fruits and vegetables in every meal as part of a healthy diet…’

Unambiguously positive assessments like this can also be found in the main data summary, the ‘What Consumers Should Know‘ highlights, and in the ‘Questions and Answers‘ link.

What Does The Press Tell Us?

As a typical example, CNN starts with the headline, ‘Watch out for the 2012 Dirty Dozen,’ and continues, ‘Apples and celery are still agricultures dirtiest pieces of produce according to the Environmental Working Groups annual ‘Dirty Dozen’ report.’  Its version of the baby food findings are, ‘For the first time this year, the USDA also collected data on pesticide residue in baby food, finding many of the studies samples to be contaminated with organophosphate pesticides.’

How can CNN report something so radically different than what the USDA said?  They simply are repeating what the Environmental Working Group has said in its press release of 6/19 and make no effort to compare it with the official document.

HuffPost Healthy Living starts with the headline, ‘Dirty Dozen: EWG Reveals List of Pesticide-Heavy Fruits and Veggies.’  Nowhere in this article is there even a reference to what the USDA, EPA and FDA said about the data.  It simply passes along the EWG interpretation as if it were gospel. 

This is the mainstream media.  You can well imagine what is said on various organic and Food Movement sites and blogs.

In contrast, Jon Hamilton writing for an NPR blog uses the headline, ‘Why you shouldnt panic about pesticide in produce.’  Jon notes that the EWG sends a ‘mixed message,’ saying ‘you should be concerned about pesticide residues in fruits and vegetables, but not so concerned that you stop eating these foods.’  Rather than simply parroting EWG, this writer demonstrates some journalistic mettle by interviewing a scientist at McGill University who can provide some perspective.  He also demonstrates that he read at least some of the USDA documentation by quoting specific numbers on apples and the non-alarm assessment for the baby food data.  Such efforts at balance are unfortunately rare.

Why Does The EWG’s Version Get So Much Attention vs The USDA’s?

Although sensationalism, low journalistic standards, and limited scientific background are certainly involved in the largely uncritical magnification of the EWGs message by the press (and particularly the blogosphere), Im afraid that UDSA-AMS is partially to blame.  Dont get me wrong, they do a rigorous collection and analysis of the data.  They are extremely clear in what they conclude with sound reasoning.  They are definitely transparent and unbiased in their presentation.  What what USDA does not do is provide a summary version of the data that is easily digestible by ordinary readers, including typical members of the press.  EWG provides a simple list with a one dimensional ranking.  It is a gross and misleading simplification, but that makes it easy to relay as if it was a real analysis.  Unfortunately, the summaries that the USDA presents are extremely detailed, extremely long, and not easy to interpret even for someone who wants to.  Let me explain.

A Document That I Doubt Many People Really Read and Digest

The annual summary document for each years PDP data is huge, in the range of 200 pages as a pdf (it is hard to tell, the appendices are numbered independently).  It starts with 28 pages of background on methodology and summary of sample types etc.  Then there are 9 pages of a historical appendix.  Finally the actual data begins in a 77 page appendix, but this is organized by chemical – nothing that most people would even begin to relate to.  For each chemical there is no information about whether that product is something toxic or not, nor is there information on any other dimension of its environmental profile.  To get that, someone would have to search for an MSDS and maybe an EPA RED – none of which are easy for any layman to interpret.  All this table lists are the number of samples, the number and percent of ‘detections,’  the range of those detections in PPM, the LOD (limit of detection), and the EPA tolerance (an extremely conservative level set by an elaborate risk analysis). Most people would have no idea what to do with those numbers.  In fact they show that the vast majority of ‘pesticide detections’ are at levels well below the tolerances, but it is tedious to do the comparisons by eye and there is nothing in the table to give the message of how far below any level of concern the vast majority of samples are actually shown to be.  Im not surprised that no independent journalistic interpretation of these data occur.

Below the chemical-by-chemical summary there are dozens of pages summarizing commodities other than fruits and vegetables.  That is followed by an appendix J which unhelpfully simply compares the ‘percent detections’ for imported and domestic samples (a summary level as seriously unhelpful as that by the EWG).

Finally, at the very end of this huge document there is a crop-by-crop summary with the same data columns as for each chemical (#samples, #detections…) which has the same tedious requirement to compare detections and tolerance that are all usually numbers to several decimal points.  I have never seen anyone in the press do much if anything with this data set.

What Does This Report Need?

To be fair, USDA-AMS has amassed such a huge body of data.  It is difficult to summarize it in a way that is intuitively meaningful.  Their bottom line conclusion, ‘produce is safe,’ is a perfectly valid, but some visual representation could go a long way towards getting that message across.  I acknowledge that this is difficult.  I have made some attempts to do so in the past.  I plan to do so again with this years data, but that will take time.  The same would be true for even an ambitious reporter, while simply reporting what the EWG says allows less ambitious reporters to keep up with the instant news cycle.

A Mountain of Data

It is a wonderful thing that the USDA makes the effort to analyze so many samples of so many crops and looks for so many different chemicals.  The downside is that this generates a database that is beyond what most of us even know how to process.  The file that one can download with the raw data has been getting bigger every year and has now reached 85MB.  Im used to dealing with large files, but none of my ordinary software can deal this this.  Ive gotten my son to write a program in RUBY to parse the data and only give me the tiny fraction which contains ‘detections’ and discard the millions of rows of data that effectively say, ‘we didnt find this chemical in this sample.’  Im going to ask him to modify the program this year with some additional summaries.  

When and if I get this done, I plan to make this available to anyone interested in doing an actually meaningful analysis of the data and to explore ways to present it graphically in ways that can compete with the egregiously trivialized ‘analysis’ done by EWG.  

You are welcome to comment here, and/or email me at

Produce stand image from Steve Savage

(Via Steve Savage Applied Mythology.)