Integration - Episode 9 - Maths

You know, usually when we talk about mathematical calculations, there's this um inherent expectation of precision that you get through isolation, >> right? Yeah. Like you want to isolate variables. >> Exactly. It's uh it's very much like taking a photograph. You click the shutter, you freeze the action and you capture just one single static moment in time >> and you [clears throat] look at the resulting photo and you can say with absolute certainty like there it is. That is exactly what was happening at that precise millisecond >> which is a very natural analytical instinct. You know when you're faced with a really complex moving system our first operational reflex is to just stop the clock >> right because we want to get our bearings. >> Exactly. We want to isolate the variables so we can actually understand what we're looking at without the uh the parameters constantly changing on us. >> Yeah. Uh it's a comforting methodology and if you are an actuarial student sitting down to tackle the mathematics of risk and finance you are intimately familiar with that instinct. >> Oh absolutely. >> In a previous deep dive we spent a lot of time looking at differentiation right the mathematics of the precise moment. >> Yeah. Freezing time to find that instantaneous rate of change. >> Right. But the core argument we are exploring today is that the math of pinpointing a precise moment is really um it's only half the story. It is. If you only look at the frozen frames, you might completely blind yourself to the actual movie. >> Exactly. So, the mission for today's deep dive is to explore the exact opposite of that isolation. We are going to take everything you know about breaking a mathematical world apart and we're going to examine the philosophy of putting it all back together >> which is integration. Yes, we are diving into integration today and we have a really fascinating excerpt from our source material called uh the calculus of accumulation, the soul of integration. >> And this specific text is just exceptionally valuable for our discussion. It doesn't just review the calculus you already know. It serves as this um philosophical and historical roadmap. >> Yeah. It bridges the highly abstract world of calculus with the real continuous realities of actuarial science. So welcome to this deep dive. >> To set the stage here, the text argues that integration requires a genuine paradigm shift because I mean in early mathematical training, the focus is entirely on dissection, >> slicing things up, >> right? You take a continuous function and you slice it into those infinitely small frozen moments. But integration demands that you orient your thinking toward the grand constructive act of accumulation. >> Right? The text sets up a really stark dichotomy right from the start. It says differentiation is slicing the world up into frozen slices of time and integration is the grand act of putting those slices back together. >> The mathematics of accumulation, it's not just finding an area on a graph. It's about building a cohesive total out of fragmented data. >> So, let me test a mental model here based on that. If differentiation is hitting the pause button on a film, examining one single frame to see a car mid crash, then integration is uh hitting play. >> That's a great way to put it. Yeah. Like you take thousands of static frames, feed them through the projector and watch how they accumulate into a continuous film. >> Exactly. Differentiation yields the isolated data point and integration yields the cinematic experience. >> Okay. But let me push back a little. Reading through the first chapter, the author treats this accumulation as a monumental conceptual hurdle. And I want to dig into why. >> Okay. Sure. >> If you are an actuarial student, you know how to add things up, right? If integration is fundamentally just putting pieces together, why is this such a profound mountain to climb? Isn't it just addition? >> Well, that's a valid critique, but you have to look closely at what is actually being added. You're right that humans are incredibly adept at discrete addition, >> right? Like adding three apples and four apples. >> Exactly. Or modeling a portfolio with three discrete assets, then adding four more. Accumulating that total is trivial. We're hardwired for discrete finite quantities. >> Okay. So, where's the leap? >> The leap. The reason this requires a total rewiring of how you perceive reality is that integration is not discrete addition. It is the summation of an infinite number of infinitely small quantities. >> Ah, and this is where my movie frame analogy kind of breaks down, doesn't it? >> It does. Yeah. Because a movie reel has a finite number of frames, like 24 frames a second. You could physically sit down and count them. But the realities you're modeling in actuarial science, time, lifespans, interest, they don't operate at 24 frames a second. >> No, reality is perfectly continuous. Between any two seconds on a clock, there isn't just one frame or a 100 frames. There are an infinite number of intermediate moments. >> So, you aren't just adding up a neat stack of data points. You're attempting to sum an infinity of moments. And each one has a duration of essentially zero. >> And that specific paradox paralyzed the greatest mathematical minds for centuries. How do you build a tangible finite total out of building blocks that are infantesimally small >> without the result just blowing up to infinity, >> right? Or collapsing into zero. That's why the text emphasizes this isn't just advanced edition. >> Okay, so to really illustrate why this is so difficult, the source text takes us on a historical detour. Long before actuaries were building continuous mortality models, we go back uh over 2,000 years to ancient Greece. It's a necessary detour to grasp the modern integral. You really have to understand the geometric problem that forced its invention. >> So the text introduces us to Archimedes and he was operating in a math landscape governed entirely by straight lines. >> The straight edge and the compass, >> right? If you gave him a triangle or a rectangle, he could calculate the area perfectly. But he realized a really frustrating truth. Nature is rarely made of straight lines. >> Yeah. Nature is almost entirely curves. a thrown projectile, a coastline, a planet. The tools of antiquity had no formal language to quantify a continuous curve. >> And Archimedes focused his frustration on one shape in particular, the circle. He wanted the exact area. But how do you measure a curve when you're only measuring stick as straight, >> right? And the text outlines his methodology, the method of exhaustion, >> which is aptly named. So, let's walk through this because it lays the absolute foundation for the definite integral. He starts with a circle, but he only knows street shapes. So he inscribes a square perfectly inside it. >> Right. Where the four corners just touch the inner boundary. >> Yeah. >> And calculating the area of that square is trivial for him. >> But the square is a terrible approximation of the circle's area. There are these massive curved gaps between the flat edges of the square and the sweeping curve of the circle. >> Exactly. It severely undercalculates the true area. So he iterates. He doubles the sides from a four-sided square to an eight-sided octagon. >> And mathematically, the area of that octagon is much closer. The straight edge is pushed further out into those empty curved spaces. >> Mhm. But there's still gaps. So he doubles it again to 16 sides than 32 sides. >> And reading this, it immediately made me think of early computer graphics, like an 8-bit video game from the 1980s trying to render a round steering wheel. >> Oh, that's a perfect modern analog, >> right? It looks terrible because the screen is made of large square pixels. The curve looks incredibly blocky. It's straight edges trying to mimic a circle. >> Archimedes was essentially manually increasing the resolution of his polygon. >> Yes. As graphics technology improves, you get more pixels and they get smaller. The straight edges become so tiny and densely packed that it begins to hug the true curve tighter and tighter. And eventually at a high enough resolution, the human eye can't even perceive the straight square pixels. It just registers as a perfectly smooth curve. >> And our committees did exactly this in the sand. He calculated the area of a 96-sided polygon. The sides were incredibly short and the perimeter hugged the curve with astonishing precision, >> giving humanity its first really accurate approximation of pi. >> But the crucial point the text makes is that he didn't just stop at 96 sides and call it close enough. The true conceptual leap, the one that echoes into modern actuarial modeling, was a leap of theoretical abstraction. >> Right. He hypothesized the limit. He realized that a polygon with a finite number of sides, even a billion sides, would always have microscopic gaps. >> It's still just an approximation. >> Exactly. The only way to perfectly map the continuous curve was to imagine a shape with an infinite number of sides. And if a shape bounded by a finite circle has an infinite number of sides, the length of each of those sides must be infinitely small. They have to approach zero. >> Precisely. The geometric gaps shrink to zero and the infinite straight lines seamlessly become the curve. >> The text describes this so beautifully. It says he solves the curve by chopping a complex curved problem into tiny manageable straight pieces and adding them all up. And the author declares this mental leap is the absolute soul of integration. >> That is the essence of it. Yeah. When you face a reality too chaotic or curved to measure discreetly, you chop it into an infinity of tiny linear pieces and accumulate them. >> But as the text points out, Archimedes only gave us half the puzzle. He gave us the geometry of slices. What he lacked was the realization that this geometric accumulation was intimately connected to motion and speed. >> Right? Which correct? The derivative of the distance function is the speed function. you extract the rate of change from the accumulated total. >> But the breakthrough is the reverse. If you take that speed function, a continuous curve showing how fast the car was going at every single fraction of a second, and you integrate it, >> you perfectly reconstruct the total distance traveled. >> Wow. By accumulating all those infinitely tiny moments of speed, >> exactly. What you just described is the fundamental theorem of calculus. The integral of a rate of change gives you the total accumulated amount. >> It fundamentally unifies geometry and motion. The geometric area under the speed curve physically represents the accumulated distance. >> And understanding why this matters is so crucial. Newton and Livveness created a reversible universal language. It means if you could observe how fast something is changing, which is often way easier to measure in real time, you can mathematically project the total outcome. >> Right? We see this all the time. It's much easier to measure the current rate of inflation than it is to intuitively grasp what it'll do to a dollar over 30 years. But if we have the rate, we integrate it to find the total effect >> precisely. >> However, there's a mechanical piece missing here. Archimedes had inscribed polygons. Newton and Lveness gave us the theoretical connection. But how do we actually compute the accumulation of a totally unpredictable curve in the real world? >> Right? Because an economic curve doesn't look like a perfect circle. >> Exactly. So for the operational mechanics, the text brings us into the 19th century to the German mathematician Burnhard Ryman, >> the architect of the Ryman sum. >> What exactly was his contribution here? >> Well, Archimedes used a polygon for a circle. Ryman standardized that concept so it could be applied to any continuous function. He proved you could calculate the exact area under any wild sweeping curve by dropping perpendicular lines down to the x-axis, filling the space with vertical rectangles. >> Back to straight lines. >> Always back to straight lines. Yeah. >> Yeah. >> Because the area of a rectangle is basic geometry, just base multiplied by height. >> So let's clearly define the dimensions of these rectangles because the variables map directly to how actuarial models are built. >> Okay. If I'm looking at a graph, the width of each vertical rectangle sitting on the x-axis represents a tiny increment of time, right? The delta x, >> correct? A discrete finite slice of the x-axis. And the height of that rectangle is simply the value of the function, the y-value at that point. So area is f(x) multiplied by delta x. >> But if you just draw 10 wide rectangles under a curve, you get a really blocky inaccurate estimation. >> Right? So Ryman applied the Archimedian limit. He asked what happens as the width of those rectangles, the delta x approaches zero. >> As the width shrinks to zero, the number of rectangles you need approaches infinity. >> Exactly. And the jagged flat tops of those infinitely thin rectangles perfectly conform to the smooth sweep of the curve above them. >> So mathematically, the definite integral is taking the sum of the areas of those infinite rectangles. You continuously multiply the height by an infinite decimally small width and add them up. And the integral sign itself, that elongated S, literally stands for suma, the continuous sum of infinite parts. >> I love that detail. I want to ground this highly abstract geometry in a really tangible reality. The source text has this brilliant analogy, the VNA waterfront analogy. >> Oh yeah, it's a fantastic way to conceptualize the transition from rate to accumulation. >> Right. So imagine you're standing at the waterfront. You're looking down at the canal and the tide is rushing in. The water is moving continuously. If you look at this through differentiation, you're taking a snapshot. >> You determine that at precisely 2.00 p.m., the water is flowing in at a rate of, say, 5,000 L per minute. You have a rate of change at a frozen instant. >> But if you use integration, your perspective completely shifts. You don't just look at 2.00 p.m. You take the continuously changing rate over the entire afternoon. Maybe it rushes in, then slows to a crawl, then surges again. >> Mhm. You map that chaotic flow rate as a curve. The y-axis is the rate in liters per minute and the x axis is time. >> Then you apply Ryman's logic. You'd shop the afternoon into infinitely thin rectangles of time. You multiply the flow rate at each millisecond by the microscopic duration of that millisecond and accumulate the area >> and the result is no longer a speed. >> No, it outputs a physical reality. It tells you the exact total number of lers of water now sitting in the harbor. I really want to emphasize the profound nature of that. Integration takes an abstract measurement of velocity, something you can't physically hold, and by accumulating it over time, transforms it into a tangible absolute quantity, >> a physical volume of water, it synthesizes a concrete reality. And that precise transformation is the bridge we need to bring this discussion fully into the professional realm for our actuarial students. >> Yes, let's make that pivot. We've covered the fundamental theorem and Ryman's infinite rectangles, but the crucial question is why is this specific math so non-negotiable for an actuary? Why can't financial models just rely on discrete statistics? >> The text addresses this directly. It centers on the fundamental nature of the risks actuaries quantify lifespans, market fluctuations, catastrophic losses. They do not occur in neat uniform increments. >> Right? The physical world doesn't wait for the end of the fiscal quarter. Accidents don't just happen on the first of the month. >> Exactly. If the world were purely discreet, if liabilities only triggered in set intervals, then simple addition would suffice. But reality is a continuous spectrum. >> And when you are legally required to distill that continuous chaos down into a single concrete number to price a premium, you have to possess the math to accumulate continuous curves. You cannot price a continuous risk using discrete math without introducing massive tracking errors. >> To see exactly how this works, the text opens up the actuarial toolkit. It walks us through four specific applications where integration is the operational engine. >> Let's start with application number one, calculating expected values and variances for continuous distributions, >> right? Where statistical theory meets real world severity modeling. The text uses a commercial liability claim like a multi-car collision. When a massive accident happens, the insurance claim doesn't pull from a neat menu of prices. >> No, it's not binary. The damage exists on a continuous probability density function or PDF. The cause could be $5,000. It could be 5,1.25 or $2.7 million. There's an infinite continuum of possible severities. >> And each severity has its own precise probability. Now, if I'm trying to find the true mathematical average cost of this claim to base the pure premium on, I can't just take a historical sample of 50 claims and average them. >> No, because that just gives you historical sample mean. It doesn't give you the theoretical expectation of the continuous risk profile because the sample space is infinite. Discrete summation completely fails. >> So, we bring in the integral. Let's break down the mechanics of the expected value formula. >> Okay, imagine the graph of the PDF. The x-axis is the magnitude of the loss. every possible dollar amount from zero out to infinity. The y-axis fx is the probability density the likelihood of that exact amount occurring. >> So a highly probable minor bumper repair would be a massive peak on the y-axis. A catastrophic total loss would be a point very low on the curve stretching far out to the right >> precisely. And to calculate the true expected value, you can't just look at the peak. You must take every single possible infinite outcome along the x-axis, multiply it by its mathematical likelihood fx, and then you're accumulate that product across the entire continuous domain, >> usually from 0 to infinity. You're using Ryman's infinitely thin rectangles to accumulate the probability weighted value of every possible financial outcome. >> Exactly. >> And what strikes me is how heavily this relies on integration to capture tail risk. If a distribution has a heavy tail, a small probability of a multi-million dollar loss far out on the right, discrete sampling might miss it entirely. >> That is a crucial insight. Integration forces the actuary to rigorously price the continuous asmtotic tails into the premium. Without the integral sweeping from zero to infinity, catastrophic tail risks would be systemically underpriced. >> Wow. Okay. So, let's transition from pricing a sudden loss to managing the money meant to pay for it. Application number two shifts to financial mathematics, specifically continuous cash flows. >> Right? The circulatory system of asset liability management. >> The author uses a great phrase here. They state that in complex financial systems, money doesn't just move in discrete steps at the end of a ledger period. Often money flows like a river. >> It's a very accurate metaphor. Think about a massive real estate trust managing thousands of properties. The revenue doesn't materialize in a single bank deposit at midnight on December 31st. >> No, it's a constant unceasing stream of transactions happening globally every second. A continuous payment stream. >> Furthermore, the interest acting upon that money is continuous. We model continuous compounding using the mathematical constant E. >> The money is literally earning microscopic fractions of interest on every microscopic fraction of a second. >> Money is a constantly swelling river. Yes. >> But here's the operational challenge. If an actuary is managing a pension fund with this massive river of cash projected into the future, they can't just report a metaphor to the board. >> The board needs the present value. What is the exact value of this infinite future river in today's dollars? >> And calculating that requires integration because you have two continuous forces acting against each other simultaneously. >> Let's unpack that conflict. You have the payment stream flowing continuously forward adding value. But simultaneously you have the force of interest delta constantly pushing against that future value discounting it back to the present. >> Exactly. Because a dollar received 10 years from now is worth far less today. The discount factor applied to any single point in time is an exponential decay function. E to the negative delta t. And the further out into the future the time gets, the heavier that exponential discount factor becomes, >> right? Shrinking the present value. >> So to find the total present value, the actuary must chop that future river into infinitely thin remnant slices of time. They take the exact flow rate at that microcond, multiply it by the continuous discount factor for that microcond, >> and then they integrate. They accumulate all of those discounted slices from time zero out to the end of the term or even infinity. >> It's an elegant piece of mathematical engineering. You take continuously generated future wealth, subject it to the erosive friction of time and interest, and use calculus to boil it down into one highly precise present-day dollar value. >> It allows the actuary to instantaneously price an infinite continuum of future cash. >> Incredibly powerful. But as we move to application number three, the text takes us away from pure finance and applies this to human mortality. We move into survival models. This is arguably the foundational construct of the entire life insurance industry. >> The text asks us to visualize the survival function curve. Let's describe it. The x-axis is time representing the age of a person. The y-axis is probability from 0 to 1 or 0% to 100%. >> Right? At time zero, the very beginning, the curve is at exactly 1.0. There is a 100% probability the person is alive at the exact moment you begin measuring. And as time moves forward, as they age through their 20s, 40s, 60s, the curve inevitably, gracefully begins to slope downward. >> Mhm. The probability of surviving to 50 is lower than surviving to 20. >> And as they enter their 80s and 90s, the slope accelerates downward, eventually hitting zero at the absolute maximum boundary of the human lifespan. >> It's the stark mathematical shape of mortality. >> Now, here is the stunning revelation. If an actuary wants to calculate the exact average number of years this individual has left to live, the complete expectation of life, they don't do a discrete weighted average of dying in year 1 plus year two. >> No, because human life isn't lived in discrete annual intervals. You don't survive your 45th year and instantly teleport to your 46th. Mortality is a continuous force operating at every fraction of a second. >> So the model must accumulate continuously. To find the complete expectation of life, the actuary simply integrates the survival function from time 0 to infinity. >> They calculate the precise geometric area resting entirely underneath that sweeping survival curve. >> Let's consider what that physically represents. By taking the area under the curve, you were slicing a lifetime into an infinity of tiny fractions of a second. You're evaluating the probability of surviving that microscopic slice of time and accumulating all those infinite probability weighted surviving moments together. The total geometric area is mathematically equal to the average number of continuous years lived by integrating infinite probabilities of surviving each passing millisecond. The calculus outputs a lifetime. >> It's a remarkable intersection of abstract geometry and human reality. The integral provides the only true representation of the complete expectation of life. >> Wow. It's the absolute bedrock for pricing any continuous life annuity. It really is genuinely elegant mathematics. >> It is. >> We've accumulated probability, time, money, lifespans. Let's examine the final application. Risk management and contract architecture. How actuaries use integration to construct policy modifications like deductibles and liability limits. >> Right? So if the first three were about modeling natural behavior, this fourth application is about human intervention. How an actuary forces a mathematical boundary onto a continuous risk. >> Let's go back to the PDF we discussed in application one, the curve of infinite possible severities of a commercial claim. A rational insurance company almost never agrees to take on infinite risk. >> No, they don't expose themselves to that extreme right-hand tail where the multi-billion dollar losses live. They modify the coverage. They impose a policy limit. They cover the continuous severity of the loss, but only up to a strictly defined maximum threshold, which we call L. >> And imposing this limit fundamentally alters the math. The actuary is no longer integrating the pure undisturbed curve from 0 to infinity. They're integrating a newly modified peace-wise function. >> The text has a great analogy for this. The dam on a river. >> Yes. The probability density curve is the natural shape of the river valley. The potential financial loss is the water flowing into it. When an insurer sets a limit at L, they're essentially building a rigid concrete dam exactly at point L on the X-axis. >> A perfect extension of the waterfront analogy. So any loss that occurs between zero and the limit L is captured and held by the dam. The insurer pays for that volume entirely. >> But what happens if a massive catastrophic loss exceeds the limit L? The water hits the dam, fills it up exactly to the top and the excess water violently flows over the spillway. And the crucial point is that the insurer is not financially responsible for the water going over the spillway. >> Right? They are only responsible for the precise volume retained behind the dam. So when calculating the new limited expected value to price the modified premium, they run a peace-wise integral. >> They split the calculation into two parts. First, they integrate the standard severity curve, but only from zero up to the limit L. >> That calculates the expected value of all the water that flows perfectly into the reservoir. >> But they can't ignore the probability that a catastrophic flood will happen. If a loss exceeds L, the insurer still has to pay the maximum limit, L. >> Correct? So, the second piece of the equation handles the spillway. The actuary calculates the total cumulative probability that the loss will be greater than L, the area of the tail beyond the dam. >> And they multiply that entire tail probability mass by the maximum payout L. >> Exactly. Then they add the integral of the retained risk to the probability weighted mass of the maximum limit. The calculus lets the actuary surgically slice the continuous risk curve at a specific point. Discard the infinite severity of the tail and reintegrate only the area of the risk constrained by the contract. >> It is precision engineering. >> Yeah. >> By manipulating the upper bound of the definite integral. They instantly calculate exactly how much the expected value drops when they move a policy. ensuring the premium perfectly matches the modified area of risk, which perfectly illustrates why the author calls integration the tool of the builder. >> It really is. >> So, let's synthesize everything we've unpacked here. We started by identifying the instinct to isolate and differentiate to measure the world by breaking it down into frozen slices. >> We journeyed through the geometry of Archimedes, realizing you have to conceptually overwhelm a curve with an infinity of tiny straight lines to map it. We saw how Newton and Linus proved that finding the rate of change and accumulating the area are cosmic mathematical opposites. And we used Ryman's rectangles to see a flow rate transform into an accumulated volume. >> And most importantly, we applied all of that directly to the continuous realities of the actuarial profession. >> We established that if differentiation is the tool for measuring sensitivity and velocity, then integration is the ultimate tool of synthesis. It's the mechanism that lets you face a chaotic system governed by shifting variables, liability severities, cash flows, human mortality, and systematically gather them up. >> You don't just observe the chaos, you accumulate it. You distill the continuous infinite down into solid, highly precise financial metrics. >> And those integrated metrics are the absolute foundation of the reserves and insurance mechanisms that protect global markets. >> The true applied power of the calculus of accumulation. But before we conclude, we need to leave you with a conceptual challenge. Yes, a final thought. >> We have spent this entire session discussing how to mathematically integrate realworld curves to build solid promises. Our framework assumes the probability curve. The rate of the tide, the severity of the crash just exists out there in nature waiting to be measured. >> We treat the curve as an objective physical reality. >> Precisely. But here is the provocative question you have to consider as you move into applied behavioral finance. What happens to the integrity of your integral when the very act of ensuring the risk fundamentally alters the shape of the continuous curve you are trying to measure? >> Ah moral hazard. The observer effect applied to risk accumulation. >> Right? If a corporation knows they are heavily insured up to a massive limit L, they might continuously take on riskier behaviors than they otherwise would have. >> Exactly. The underlying probability density function shifts. The tail gets fatter simply because the insurance contract exists. How do you integrate a reality that actively mutates in response to the fact that you promise to protect it? >> Your underlying curve becomes a moving target. That is a profound complication to the math. I want you to let that systemic challenge sink in. >> Think about the elasticity of the curves you are measuring. >> Yeah. The next time you sit down and trace an interval sign onto a piece of paper, you aren't just summing numbers. You are accumulating a highly reactive continuous reality. Thank you for joining us for this deep dive into the soul of integration. Keep dissecting the concepts, keep accumulating those insights and we will catch you in the next session.