It's a fantastic question, and a really important one. It seems so counterintuitive, no? I'm suppose to train dramatically slower than my goal HM pace (or better yet my current fitness HM pace), and yet on race day, I'm going to be able to go significantly faster. In 2015, you would have also counted me as a skeptic before I took up this methodology of training.
Short answer - Easy runs build the physiological adaptations necessary for endurance running, allow you to do more training more often, keep you injury free more often, and thus more consistent with your training.
I've made a few posts on similar topics before that cover the general ideas:
Eureka! The Quintessential Running Post
Train slow to race fast: Why running more slowly and capping the long run at 2.5 hours may dramatically improve your performance
The Marathon is 99% Aerobic (and 95% for HM and so on): So how to train for it!
And my favorite resource (runners connect blog):
How Fast Should Your Easy Runs Be?
Why Running Harder Won’t Help You Get Faster
Are You Sabotaging Your Long Run by Running the Wrong Pace?
How to Know if You are Running Easy Enough on Your Recovery Days
How Running 80% Easy Could Make You 23% Faster
Jeff Galloway- Go Slow To Go Fast (podcast)
3 Simple Ways to Determine if You are Running Easy Enough: Matt Fitzgerald (podcast)
Hansons Easy Days
In its most basic form, a running training plan is meant to provide a "stimulus" to the body and mind. The "stimulus" depends on the type of training you do. You do speed training and you elicit the potential for changing your muscle fibers (can take years to make this transition). You do lactate threshold pace training and your ability to run a certain pace for about an hour becomes faster. You do a run in super hot conditions over many days and you're body thins out your blood so the blood can travel throughout the body easier and the heart doesn't have to work as hard. Running multiple days in a row causes your body to be in a state of mild fatigue which over time puts you at a different "starting position" for different runs (i.e. instead of starting at mile 1 for your long run, you're actually starting at mile 8, OR said another way if the fatigue build-up in your body is zero going into your long run you might never teach your body what it's like to run beyond your long run mileage **However, with that being said I still place a max of 2.5/4.5 hrs on training runs dependent on style**. It's why the Hansons plan can go to 16 miles max (for some people) and still enable you to be capable of completing a 26.2 mile race. I've personally had runners successfully run as little as 10 miles (150-180 min) and complete a marathon as expected when appropriately balanced with the rest of the week. These long runs allow you to stave off fatigue for longer.
All of these stimuli lead to an adaptation in the body in response to that stimulus (not everything is listed here as adaptations, but a short example)
Speed -> Fiber change
Lactate Threshold -> Fatigue Clearance and tolerance of fatigue
Hot/Humid Conditions -> Thinned out blood
Easy runs -> Increased capillaries
Long Runs -> Mitochondrial density and volume
Glycogen Depletion Training -> Use fat more readily as a fuel source
These stimuli are important. Because the adaptations to these stimuli mostly occur during the recovery portion of the training plan. So if you're constantly training too fast, you'll get stuck in a feedback loop of:
Stimulus -> Some recovery -> Stimulus -> Some recovery
Whereas the runner that does more slow running more often will see a feedback loop of:
Stimulus -> Recovery -> Adaptation -> Stimulus -> Recovery -> Adaptation
So, now think of everyone's pace spectrum. The universal truth is everyone has a VO2max. Everyone also has % VO2max that can be attributed to different race distances. When you train at different paces along the %VO2max spectrum it elicits different responses (like stated above) based on how you accomplish a workout. So to be a well-rounded runner you want to touch bases on a lot of different paces across the spectrum. This will allow you to elicit lots of different benefits on different days. As the race gets closer, you specialize towards your goal. All the while keeping the large majority (an often quoted value is 80%) easy.
So let's circle all the way back to the original question, why should your average pace be so much slower than your goal pace? Because a HM is about 95% aerobic and 5% anaerobic. So it's not a speed race, it's an endurance race. So how do you build endurance or the aerobic system? By providing a low level stimulus, lots of times. This low level stimulus will encourage the body to become more efficient at transporting oxygen, recruiting bone development, increase capillary density, and efficiency in glycogen storage/fat usage just to name a few. Beyond the benefits of the easy runs directly, they also indirectly boost the rest of your training. You're generally a lot less fatigued going into each run, such that you can run more often. And the research will back me up that in general the combination of your balanced running volume with the effort at which you do it will be a determining factor in performance. So if you try to run faster all the time, but then train less often, you won't reap as much benefits as running slower and more often. The primary driver for this discrepancy will be the occurrence of injuries. The slow/higher volume runner is probably (not definitely) going to see less injuries than the faster/lower volume runner (with fast and slow being relative to one's own fitness). This lack of injuries is going to enable the runner not only to reap more adaptations along the way in the training, but also be more consistent with the training over the long term. So over the course of months/years, the slow/high runner will likely (not definitely) end up at a better fitness level.
If you do some digging into the widely available training plans, you'll see that most modern day coaches follow this methodology. That includes Galloway, Fitzgerald, Hansons, Hal Higdon, McMillan, etc. Some do a better job of explaining it than others. But with a keen eye, you can see it's there for all of them. About the only one that comes to mind that is different is FIRST. But that's because FIRST also asks you to do cycling at an aerobic pace multiple times per week (cross-training). If you do FIRST without the cycling, then you won't be sufficiently prepared for your race. From my own personal experience of blending running with cycling, I've found that I need 2x cycling time to equal the same running time. So if I want to move from 10 hrs per week of running down to 5 hrs per week of running, I need ~10 hrs of additional cycling work to supplement my 5 hrs running. So that means run+cycle will take more time than just running alone to equal a certain running fitness level.
So how slow is slow enough? There are a lot of pace calculators out there:
https://lukehumphreyrunning.com/hmmcalculator/https://runsmartproject.com/calculator/https://www.mcmillanrunning.com/http://www.jeffgalloway.com/training/magic-mile/
In general they're going to give you a similar range for an individual fitness level. When I write training plans for others, I generally find the average weekly/monthly pace ends up being about 12% slower than estimated marathon tempo.
So a different question would be, can I train too slow? According to Tom Schwartz from TinMan training he has a dataset (not shared) that shows easy days can be as slow as 5 min/mile slower than 5k pace and still be beneficial as an easy day. Being 5 min/mile from 5k pace is pretty darn slow relatively speaking. So it leaves a fairly large window.
So does it work? Can you really train slow and race fast? I've been writing plans for just over 5 years now (just crossed 500) and over a wide spectrum of paces from sub-3 marathoners to sub-7 marathoners, the idea of train slow to race fast works. It's a fairly universal use idea for endurance running. From my own experience back in 2015:
I used a "PR the day" type mindset from when I started in 2012 through mid-2015. I went out and ran hard every day. If I wanted to be faster, I had to train at faster paces endlessly. So, in the
3 years prior to switching my mindset, this was my improvement progress. I did about 2700 miles in those 3 years, was never seriously injured and trained rather consistently.
5k – *35:00 to 23:36 (33% improvement)
10k – 51:45 to 49:49 (4% improvement)
HM – 2:01:00 to 1:49:24 (10% improvement)
M – 4:50:26 to 4:20:34 (10% improvement)
After
18 weeks of Hansons training with going slow, reducing the long run, increasing the training load and physiologically relevant pacing:
10k – 49:49 to 49:22 (1% improvement) *Occurred during the 2nd half of the marathon
HM – 1:49:24 to 1:46:00 (3% improvement) *Occurred during the 2nd half of the marathon
M – 4:20:34 to 3:38:53 (16% improvement)
After
27 weeks of Hansons training:
5k – 23:36 to 22:03 (7% improvement) **Occurred during the 2nd half of the half marathon
10k – 49:49 to 44:36 (10% improvement) **Occurred during the 2nd half of the half marathon
HM – 1:49:24 to 1:38:47 (10% improvement)
M – 4:20:34 to 3:38:53 (16% improvement)
So in the 3 years before the change in methodology, I cut off 11.5 min from my HM. After just 27 weeks, I cut another 10.75 min from my HM time.
After
3 years (7750 miles, but roughly the same timeframe) of going slow, reducing the long run, increasing the training load and physiologically relevant pacing:
5k - 23:36 to 19:27 (18% improvement)
10k - 49:49 to 39:54 (20% improvement)
HM - 1:49:24 to 1:30:35 (17% improvement)
M - 4:20:34 to 3:14:05 (26% improvement)
Hope that helps! And let me know if something needs a better explanation.