Using Decision Trees in Ag Decision-Making with Jay Parsons
Center for Agricultural Profitability
Author
08/08/2024
Added
1
Plays
Description
This episode dives more into decision-making in agriculture, focusing on a powerful tool that can help farmers navigate the complexities and uncertainties they face.
Jay Parsons, a professor in Nebraska’s Department of Agricultural Economics and Director of the Center for Agricultural Profiability, joins to discuss his article "Branching Out: Harnessing the Power of Decision Trees." It is co-authored with John Hewlett at the University of Wyoming and Jeff Tranel at Colorado State University, and first published in the July 2024 edition of RightRisk News, which you can find at rightrisk.org. The article delves into how decision trees can aid agricultural producers in making more informed choices amidst risk and uncertainty.
Searchable Transcript
Toggle between list and paragraph view.
- [00:00:00.000]From the Center for Agricultural Profitability at the University of Nebraska-Lincoln, this
- [00:00:14.160]is Nebraska FarmCast.
- [00:00:15.320]I'm Ryan Evans, and on this episode, we'll be diving more into decision-making in agriculture,
- [00:00:20.500]focusing on one tool that can help farmers navigate the complexities and uncertainties
- [00:00:25.720]they face.
- [00:00:26.500]I'm joined again by Dr. Jay Parsons, a professor in Nebraska's Department of Agricultural
- [00:00:32.020]Economics and director of our Center for Agricultural Profitability, to discuss his
- [00:00:37.380]new article on branching out, harnessing the power of decision trees.
- [00:00:42.040]The article is co-authored with John Hewlett at the University of Wyoming and Jeff Tranel
- [00:00:48.000]at Colorado State University, and was first published in the July 2024 edition of Right
- [00:00:54.260]Risk News, which you can find at...
- [00:00:56.440]The article dives into how decision trees can aid agricultural producers in making more
- [00:01:04.040]informed choices amid risk and uncertainty.
- [00:01:07.420]Jay, thanks for being here again.
- [00:01:09.300]Thanks for having me, Ryan.
- [00:01:10.500]So first, can you just explain what the concept is of a decision tree and how that might differ
- [00:01:16.500]from other tools commonly used in decision making in ag?
- [00:01:20.780]Yeah, well, I mean, it gets its name, obviously, because it looks like a tree where you have
- [00:01:25.880]different nodes.
- [00:01:26.420]of decisions or uncertainties and then outcomes from each of those.
- [00:01:30.080]So it branches out and gets, can get fairly complex, fairly fast, depending on how many
- [00:01:34.920]different types of outcomes there are from the uncertainties and how many different choices
- [00:01:39.100]you're considering.
- [00:01:39.900]The way it differs from a lot of the other ways that we do or look at decision making
- [00:01:45.040]in ag is it forces you to think of the sequence of events of how things happen and how information
- [00:01:51.720]is revealed.
- [00:01:52.520]So in that regard, it's a really useful tool.
- [00:01:55.900]So rather than just sitting there at a point in time and thinking of all of the things,
- [00:01:59.300]you kind of think of the sequence in which they happen and just kind of big buckets on
- [00:02:02.880]directions things can go.
- [00:02:04.260]Yeah.
- [00:02:05.400]And when we talk about decision trees, as you mentioned, these are actual things that
- [00:02:09.220]you know, you yourself would sit down and the producer themselves would sit down and
- [00:02:13.380]create kind of write out, chart out this decision making process.
- [00:02:17.060]And we'll get to what that entails here.
- [00:02:18.840]But we definitely want to encourage listeners to check out Jay's article from Right Risk
- [00:02:23.360]News.
- [00:02:23.680]We've linked to it in the podcast.
- [00:02:25.780]Notes here.
- [00:02:26.100]It's on our center's website as well at cap.unl.edu.
- [00:02:29.260]There's some great examples in there of what decision trees can look like.
- [00:02:33.620]And you, as you mentioned, they can get quite complex.
- [00:02:36.380]But you described in the article how decision trees can help farmers navigate uncertainties
- [00:02:42.340]like market prices and weather conditions.
- [00:02:44.380]So wondering if you could just walk us through one of those examples that you put in there,
- [00:02:48.620]the one of a corn producer's decision.
- [00:02:51.560]Yeah, I looked at a corn producer's marketing decision and, you know,
- [00:02:55.660]to begin with, I mapped out that, you know, prior to that,
- [00:02:57.700]there's planting decisions, planting conditions that would contribute to that and so on.
- [00:03:01.620]But the example I gave was really kind of at this point in time, you know,
- [00:03:05.580]midsummer, you're looking at marketing decisions.
- [00:03:07.520]You still don't know exactly how your crop is going to turn out.
- [00:03:10.300]You don't know how markets are going to turn out in the fall.
- [00:03:12.820]So I just mapped it out as kind of a simple thing where you're trying to decide
- [00:03:17.840]whether to forward price half your crop or not.
- [00:03:19.960]And you know what the current contract price is that's being offered
- [00:03:22.840]by the local elevator for the fall.
- [00:03:25.540]You still need to determine what your actual yields are going to be,
- [00:03:28.380]and you need to determine how prices are going to turn out.
- [00:03:31.140]So I just did a simple thing where yields could go up or down 10%,
- [00:03:35.060]so three different outcomes there, or turn out normal.
- [00:03:37.640]And then same thing on the prices, that they could go up or down 10%
- [00:03:41.440]or turn out normal.
- [00:03:42.400]So you map all that out, and you get like nine different outcomes
- [00:03:45.860]for each decision, whether you leave it cash open
- [00:03:48.900]or forward price half of it, given the current price that's offered to you.
- [00:03:53.460]From there, you just
- [00:03:55.420]got a bunch of numbers that you can calculate expected values
- [00:03:57.940]and do different things that look at different outcomes
- [00:04:01.000]that could possibly turn out.
- [00:04:02.180]And when it comes to making the decision trees
- [00:04:07.160]for agricultural decisions, are there some benefits
- [00:04:10.600]and challenges that might come along with that?
- [00:04:12.420]Well, the challenges are definitely the complexity, because even the example I put in there,
- [00:04:16.400]even though it's very simple and nowhere near capturing
- [00:04:19.460]everything that could possibly happen for a producer,
- [00:04:22.120]because reality is you've got a spectrum of yields, you've got a spectrum of price,
- [00:04:25.300]so the complexity is there, no matter what decision it is,
- [00:04:30.660]it gets complex really fast.
- [00:04:32.280]So that's the challenge, but the thing that's a real advantage
- [00:04:36.580]is it forces you to think through, like I said, big buckets,
- [00:04:40.340]going up, going down, you know that there's probably a spectrum in between there,
- [00:04:44.740]and once you have that template built of the things that can happen,
- [00:04:48.460]you can reuse it.
- [00:04:50.080]So the decision I described would be like you're in the mid-summer,
- [00:04:55.460]but another two months from now you know a little bit more about the market conditions
- [00:04:59.880]that could happen in the fall, you know a little bit more about your production conditions,
- [00:05:03.180]what your yields might be, and you have that template there.
- [00:05:06.000]If you're making another marketing decision at that point,
- [00:05:08.840]let's just say you're looking at maybe forward pricing a quarter of your crop at that point,
- [00:05:13.540]you can just, you know, repopulate that same tree with the new numbers and take another look at the
- [00:05:17.940]decision that you're considering. And one thing you do mention in the article that you write about
- [00:05:23.700]calculating the expected value of these different outcomes using decision trees. So how can
- [00:05:29.240]producers practically apply these calculations to make more informed decisions? Yeah, well, the
- [00:05:34.940]expected value, you know, mathematically is just a weighted average of what what could happen to
- [00:05:40.100]you. And it's not so much that, you know, you're going to make a big deal about, say, a hundred
- [00:05:44.860]dollar difference or something like that. It's it's what you're looking for. There are some
- [00:05:48.340]bigger differences where you definitely prefer one over the other. And then sometimes when the
- [00:05:53.040]expected values are close, you might look at the distribution of the possibility, say, the best
- [00:05:57.080]case, worst case scenario and kind of how those are distributed. And in that regard, pick one of
- [00:06:02.680]the choices as being much preferred to the other choice.
- [00:06:06.720]And then how can they be adapted?
- [00:06:09.720]Adapted in the end or expanded even to accommodate the unique needs that everybody has. There's
- [00:06:15.540]different conditions, different needs on every type of operation out there.
- [00:06:18.920]Yeah, well, there's plenty of places where you can maybe apply it. But it basically boils down
- [00:06:25.520]to where you have a finite set of choices that you're trying to decide from. And then you have
- [00:06:30.760]a lot of uncertainties that can happen and determine the outcomes. And it could be a
- [00:06:37.900]situation where you actually map together a sequence of choices that you're trying to
- [00:06:39.700]of a couple of decisions. So you may be making a choice today, and then that will determine the
- [00:06:44.480]choices that you can make down the line too. So if you want to get really complex, you can expand
- [00:06:49.400]it out quite a bit. But the ones that are easiest for a lot of people to do are the marketing
- [00:06:55.920]decisions where you think of the different marketing outcomes that can occur and whether
- [00:06:59.360]pricing cash or forward contracting will work better for you and what kind of distribution it
- [00:07:04.360]produces. But it could be a production decision deciding which crops to plant and you have
- [00:07:09.680]outcomes for the various crops, different price outcomes for the various crops,
- [00:07:13.000]depending on the mix you actually put out there. It could determine your distribution of revenue
- [00:07:18.540]that you can realize or profits, depending on how you structure it. So that'd be an example.
- [00:07:23.660]It could be something deciding whether or not to replace a piece of equipment, looking at the
- [00:07:28.460]probability of a breakdown and the actual, so you buy something new, it's probably not going to
- [00:07:33.340]break down, but you can have some fairly significant depreciation that first year and so on. Some
- [00:07:39.660]outcomes attached to that. If you use an older machine, probability of breakdown might be higher.
- [00:07:44.260]You got repair costs involved, some downtime involved. There's lots of different things that
- [00:07:48.760]you can just kind of map out. You don't know the numbers exactly, but you can take a guess at kind
- [00:07:53.580]of them. Hey, this is probably about a one in a four chance we're going to have to replace something
- [00:07:58.020]on that machine this year. What does that look like if that happens versus actually buying a
- [00:08:02.620]new machine? Lots of different examples that you can come up with. It's limited only by your
- [00:08:09.640]thinking it through. And it seems like with any of these decisions, just the act of writing it
- [00:08:15.120]down and having it in front of you, getting it out of your mind and onto paper so you can see it,
- [00:08:20.520]reflect back on it, that must be something extremely helpful in the process, it would
- [00:08:25.880]seem like, as to why you're putting this out there and recommending these as a possible tool
- [00:08:30.800]for people to use. Yeah, by far, that's the biggest benefit is it forces you to think it
- [00:08:34.840]through. Because even in the best case scenario, there's some serious guesses in what you put in
- [00:08:39.620]a table. So it's not, like I said, it's not a matter of seeing $100 difference and saying,
- [00:08:43.080]wow, I definitely want to do A over B. Sometimes it's a matter of realizing what it is that you
- [00:08:47.860]don't know that maybe you could find out some information about, you know, you don't know the
- [00:08:51.580]probability of this happening. But you know what, I think I could talk to some people who,
- [00:08:55.520]you know, my machine example, you could talk to somebody who has a similar machine or somebody
- [00:09:00.200]with just call up the repair shop and say, how often do the how long do these things last?
- [00:09:04.640]Well, you know, and they can give you a lot of times, pretty accurate numbers in terms
- [00:09:09.600]of how long, you know, before some part wears out on a particular machine that you're worried
- [00:09:13.720]about.
- [00:09:14.020]So there's a lot of things that can come out of it.
- [00:09:17.500]But a lot of it is just the benefit of mapping it out and seeing how things could piece together
- [00:09:22.060]over time.
- [00:09:22.640]And again, we encourage you listening to take a look at the article from RightRisk News
- [00:09:28.920]here in July that we've been discussing.
- [00:09:30.320]It's on our website at cap.unl.edu, rightrisk.org as well.
- [00:09:35.460]It's linked in the podcast notes here if you're listening on your favorite platform out there
- [00:09:39.580]and you can see them in action and kind of coming to life.
- [00:09:42.560]So that's Dr. Jay Parsons, professor in the Department of Agricultural Economics
- [00:09:46.840]and director of the Center for Agricultural Profitability
- [00:09:49.560]here at the University of Nebraska-Lincoln.
- [00:09:51.580]Jay, thanks for your time.
- [00:09:53.120]Thanks for having me on, Ryan.
- [00:09:54.340]Nebraska FarmCast is a production of the Center for Agricultural Profitability
- [00:09:58.820]at the University of Nebraska-Lincoln.
- [00:10:00.980]For the latest research-based information and education resources
- [00:10:05.440]to manage your farm or ranch operation, visit our website,
- [00:10:09.760]at cap.unl.edu.
- [00:10:12.280]That's cap.unl.edu.
- [00:10:16.060]♪ music playing ♪
- [00:10:19.240]♪ music playing ♪
- [00:10:23.240]♪ music playing ♪
- [00:10:25.260]you
- [00:10:25.820]you
The screen size you are trying to search captions on is too small!
You can always jump over to MediaHub and check it out there.
Log in to post comments
Embed
Copy the following code into your page
HTML
<div style="height: 5.62em; max-width: 56.12rem; overflow: hidden; position:relative; -webkit-box-flex: 1; flex-grow: 1;">
<iframe
style="bottom: 0; left: 0; position: absolute; right: 0; top: 0; border: 0; height: 100%; width: 100%;"
src="https://mediahub.unl.edu/media/22682?format=iframe&autoplay=0"
title="Audio Player: Using Decision Trees in Ag Decision-Making with Jay Parsons"
allowfullscreen
></iframe>
</div>
Comments
0 Comments