WEBVTT 00:00:00.000 --> 00:00:03.000 OK, let's unpack this. Imagine a single three 00:00:03.000 --> 00:00:05.559 digit number, maybe, you know, of four digits, 00:00:05.599 --> 00:00:07.900 depending on where you are on the planet, that 00:00:07.900 --> 00:00:11.320 quietly holds this immense, almost career defining 00:00:11.320 --> 00:00:14.019 power over your life. We're talking about the 00:00:14.019 --> 00:00:18.739 ultimate ubiquitous global report card. Your 00:00:18.739 --> 00:00:21.059 credit score. Right. I mean, this metric determines 00:00:21.059 --> 00:00:23.879 whether you can rent that apartment, get that 00:00:23.879 --> 00:00:26.199 crucial phone contract without a massive deposit 00:00:26.199 --> 00:00:28.760 or, you know, secure the capital to launch your 00:00:28.760 --> 00:00:31.399 own business. It's it's huge. It is huge. And 00:00:31.399 --> 00:00:34.179 the power of that number is, I think, often invisible 00:00:34.179 --> 00:00:36.899 and just taken for granted until that one moment 00:00:36.899 --> 00:00:39.159 you truly need to use it. Exactly. And at its 00:00:39.159 --> 00:00:42.799 core, a credit score is defined as a numerical 00:00:42.799 --> 00:00:44.960 expression. It's based on a really sophisticated 00:00:44.960 --> 00:00:47.539 analysis of a person's financial files. Files 00:00:47.539 --> 00:00:49.280 that come. from credit bureaus usually. Usually, 00:00:49.460 --> 00:00:51.500 yeah. Information sourced from credit bureaus 00:00:51.500 --> 00:00:53.320 to represent that person's creditworthiness. 00:00:53.399 --> 00:00:56.259 It's really designed to be a simple, quantifiable 00:00:56.259 --> 00:00:59.000 prediction of risk that a lender can easily digest. 00:00:59.600 --> 00:01:01.899 But the moment you cross an international border 00:01:01.899 --> 00:01:04.540 or even just dive into the regulatory fine print, 00:01:04.700 --> 00:01:07.799 that simple quantifiable number becomes anything 00:01:07.799 --> 00:01:10.359 but universal. Oh, absolutely. We are diving 00:01:10.359 --> 00:01:14.099 deep today across systems in 14 different countries 00:01:14.099 --> 00:01:17.959 to see these massive differences in, well, core 00:01:17.959 --> 00:01:21.319 philosophy, in the data they are willing to use, 00:01:21.379 --> 00:01:24.180 and crucially, in the consumer rights built around 00:01:24.180 --> 00:01:26.780 that all -important number. And our mission today 00:01:26.780 --> 00:01:30.700 is, I'd say, twofold. First, we want to extract 00:01:30.700 --> 00:01:33.140 the crucial insights that explain not just what 00:01:33.140 --> 00:01:35.700 your score is, but why the approach varies so 00:01:35.700 --> 00:01:38.079 wildly across the globe. And second, we want 00:01:38.079 --> 00:01:40.079 to show you this fundamental philosophical split. 00:01:40.239 --> 00:01:42.060 We're going to see these highly commercialized 00:01:42.060 --> 00:01:44.560 systems built purely on proprietary statistical 00:01:44.560 --> 00:01:47.140 prediction right alongside. Well, these really 00:01:47.140 --> 00:01:49.159 rigid systems built on government enforcement 00:01:49.159 --> 00:01:52.180 and stringent privacy controls. It's a fascinating 00:01:52.180 --> 00:01:54.500 and often contradictory. Look at how different 00:01:54.500 --> 00:01:57.620 societies define financial trust. It really is. 00:01:57.739 --> 00:02:00.379 So. Let's start with the basics of function, 00:02:00.560 --> 00:02:04.040 but immediately expand the scope. When most people 00:02:04.040 --> 00:02:06.099 think about their credit score, they think about 00:02:06.099 --> 00:02:08.539 applying for a mortgage or maybe getting a new 00:02:08.539 --> 00:02:11.039 credit card. That's the lender's primary use, 00:02:11.139 --> 00:02:13.819 right? A risk mitigation strategy. That is certainly 00:02:13.819 --> 00:02:17.780 the primary stated function. Lenders, we're talking 00:02:17.780 --> 00:02:19.759 banks, credit card companies, other financial 00:02:19.759 --> 00:02:23.099 institutions, they use these scores to evaluate 00:02:23.099 --> 00:02:26.159 the potential risk posed by lending money to 00:02:26.159 --> 00:02:29.879 consumers. They need this data to mitigate losses 00:02:29.879 --> 00:02:32.680 from bad debt, to determine who qualifies for 00:02:32.680 --> 00:02:34.960 a loan, and then to set the appropriate interest 00:02:34.960 --> 00:02:37.479 rates and establish credit limits. It's a system 00:02:37.479 --> 00:02:39.840 built, you know, ostensibly to protect their 00:02:39.840 --> 00:02:41.900 capital. But there's a crucial secondary function 00:02:41.900 --> 00:02:43.879 we often overlook, one that shifts the focus 00:02:43.879 --> 00:02:47.020 entirely from purely risk avoidance to, well, 00:02:47.060 --> 00:02:49.340 to revenue maximization. That's the key commercial 00:02:49.340 --> 00:02:51.979 insight. Lenders absolutely use credit scores 00:02:51.979 --> 00:02:54.099 to determine which customers are likely to bring 00:02:54.099 --> 00:02:56.479 in the most revenue for the institution over 00:02:56.479 --> 00:02:58.780 the long term. So it's not just, is this person 00:02:58.780 --> 00:03:01.240 going to pay me back? No, not at all. Consider 00:03:01.240 --> 00:03:04.259 two hypothetical consumers. One has a perfect, 00:03:04.419 --> 00:03:07.500 you know, 800 plus score and never really uses 00:03:07.500 --> 00:03:09.620 their available credit, pays cash for everything. 00:03:09.819 --> 00:03:12.419 Very responsible, low risk. Extremely low risk. 00:03:12.580 --> 00:03:14.699 But then you have a second consumer who has a 00:03:14.699 --> 00:03:19.060 very good but not perfect score, say a 720, but 00:03:19.060 --> 00:03:21.500 they... Actively use their credit limits. Maybe 00:03:21.500 --> 00:03:24.020 they carry a revolving balance occasionally and 00:03:24.020 --> 00:03:26.060 they frequently utilize high interest products. 00:03:26.259 --> 00:03:29.319 The 720 consumer is a better profit center. Precisely. 00:03:29.340 --> 00:03:31.479 Even if they pose a slightly higher theoretical 00:03:31.479 --> 00:03:34.479 risk. The scoring model, particularly in these 00:03:34.479 --> 00:03:36.800 commercially driven markets like the U .S., is 00:03:36.800 --> 00:03:39.460 often a finely tuned tool for maximizing commercial 00:03:39.460 --> 00:03:42.120 gain, not just a simple measure for preventing 00:03:42.120 --> 00:03:44.539 defaults. I see. So they're using your score 00:03:44.539 --> 00:03:47.819 to segment you. Exactly. Are you a prime customer 00:03:47.819 --> 00:03:50.020 who gets the lowest rate and costs them very 00:03:50.020 --> 00:03:52.560 little? Or are you a profitable customer who's 00:03:52.560 --> 00:03:55.259 going to generate interest revenue? Or are you 00:03:55.259 --> 00:03:58.039 a subprime customer whose risk just outweighs 00:03:58.039 --> 00:04:01.000 any potential profit? It's all about categorization. 00:04:01.280 --> 00:04:04.280 And what's fascinating here is how far this predictive 00:04:04.280 --> 00:04:07.039 technology has expanded outside of traditional 00:04:07.039 --> 00:04:10.219 banking and lending. It's truly become a universal 00:04:10.219 --> 00:04:12.740 vetting tool now. It has become deeply integrated 00:04:12.740 --> 00:04:15.900 across the entire consumer economy. I mean, we 00:04:15.900 --> 00:04:19.480 see mobile phone companies using scores to decide 00:04:19.480 --> 00:04:22.240 if you qualify for a premium service plan without 00:04:22.240 --> 00:04:24.939 needing a huge upfront deposit. Right. That's 00:04:24.939 --> 00:04:26.879 a common one. Insurance companies employ credit 00:04:26.879 --> 00:04:29.040 -based insurance scores heavily in underwriting 00:04:29.040 --> 00:04:32.079 policies. They argue that financial habits correlate 00:04:32.079 --> 00:04:35.160 strongly with risk profiles, whether you're insuring 00:04:35.160 --> 00:04:37.860 a car or a home. Landlords, too. They use them 00:04:37.860 --> 00:04:40.519 relentlessly to vet tenants. Relentlessly. And 00:04:40.519 --> 00:04:42.600 even government departments are known to employ 00:04:42.600 --> 00:04:45.079 these techniques in various capacities where 00:04:45.079 --> 00:04:47.199 financial reliability is necessary, like for 00:04:47.199 --> 00:04:50.180 licensing or certain regulatory compliance roles. 00:04:50.459 --> 00:04:52.660 And this expansion, it feels like it's accelerating 00:04:52.660 --> 00:04:55.480 because of digital finance. The online lenders 00:04:55.480 --> 00:04:57.860 and fintech companies operate purely in the digital 00:04:57.860 --> 00:04:59.889 realm, which lets them tap into tools. totally 00:04:59.889 --> 00:05:01.769 different kinds of data. That is the crucial 00:05:01.769 --> 00:05:05.029 next frontier. Digital finance companies are 00:05:05.029 --> 00:05:07.589 increasingly utilizing what they call alternative 00:05:07.589 --> 00:05:11.389 data sources to calculate creditworthiness. So 00:05:11.389 --> 00:05:13.829 we're moving beyond the traditional bank reported 00:05:13.829 --> 00:05:16.170 three digit score. We really are. We're moving 00:05:16.170 --> 00:05:18.860 to systems that. Analyze things like your utility 00:05:18.860 --> 00:05:22.060 payments, your rent payments, transactional data 00:05:22.060 --> 00:05:24.720 from digital wallets, and even in some regions, 00:05:24.879 --> 00:05:28.480 social media usage. And this trend expands the 00:05:28.480 --> 00:05:31.079 scope of what counts as relevant financial behavior, 00:05:31.439 --> 00:05:34.079 making the scoring process simultaneously more 00:05:34.079 --> 00:05:37.019 complex and frankly, far more reaching into your 00:05:37.019 --> 00:05:39.860 daily life. To truly appreciate the global complexity, 00:05:40.100 --> 00:05:41.720 I think we have to look at the nuts and bolts, 00:05:41.819 --> 00:05:44.819 the underlying technical foundations. These models 00:05:44.819 --> 00:05:46.870 are not simple addition and subtraction. Not 00:05:46.870 --> 00:05:48.970 at all. Let's use Australia as an example because 00:05:48.970 --> 00:05:51.350 the sources dive into the specific statistical 00:05:51.350 --> 00:05:54.009 methods they employ. This is where we see just 00:05:54.009 --> 00:05:56.230 how mathematically advanced these systems have 00:05:56.230 --> 00:05:58.769 become. They are not built on simple algorithms. 00:05:58.930 --> 00:06:01.029 They are built on advanced predictive statistics. 00:06:01.509 --> 00:06:03.790 While standard logistic probability modeling 00:06:03.790 --> 00:06:05.730 is still popular in Australia for developing 00:06:05.730 --> 00:06:08.550 scorecards. Which is good for predicting a binary 00:06:08.550 --> 00:06:10.389 outcome, right? Like default or non -default. 00:06:10.449 --> 00:06:12.810 Exactly. It's very effective for that. But advanced 00:06:12.810 --> 00:06:15.589 methods are also widely used by the bureaus and 00:06:15.589 --> 00:06:17.290 the lenders themselves. And this is where the 00:06:17.290 --> 00:06:19.829 terminology gets a little jarring for the casual 00:06:19.829 --> 00:06:22.769 listener. We are talking about complex statistical 00:06:22.769 --> 00:06:27.589 tools like Mars, CART, CHAID, and random forests. 00:06:27.870 --> 00:06:29.649 I know. It sounds like a graduate course in data 00:06:29.649 --> 00:06:32.500 science. It really does. You're right. The names 00:06:32.500 --> 00:06:34.579 themselves are overwhelming, but the concept 00:06:34.579 --> 00:06:36.540 is actually quite accessible. Think of it less 00:06:36.540 --> 00:06:39.439 like simple division and more like using sophisticated 00:06:39.439 --> 00:06:42.199 machine learning, a kind of financial AI, to 00:06:42.199 --> 00:06:45.319 sort through thousands of data points simultaneously 00:06:45.319 --> 00:06:48.379 to find these hidden connections that influence 00:06:48.379 --> 00:06:50.720 your final score. So it's looking for patterns. 00:06:51.019 --> 00:06:53.980 Deep patterns. Mars, for instance, that's multivariate 00:06:53.980 --> 00:06:56.800 adaptive regression splines. It allows the models 00:06:56.800 --> 00:07:00.240 to handle extremely complex nonlinear relations. 00:07:00.269 --> 00:07:02.970 So it's not just that high credit card usage 00:07:02.970 --> 00:07:05.209 is risky. These systems are looking for much 00:07:05.209 --> 00:07:08.170 more nuanced conditional risk profiles. That's 00:07:08.170 --> 00:07:10.810 it exactly. They look for interactions that simple 00:07:10.810 --> 00:07:13.129 models would just completely miss. The model 00:07:13.129 --> 00:07:15.649 isn't just saying you have high debt. It might 00:07:15.649 --> 00:07:18.470 be saying high utilization combined with a specific 00:07:18.470 --> 00:07:20.850 pattern of credit card balance transfer usage 00:07:20.850 --> 00:07:24.089 focused on a certain type of retailer debt suggests 00:07:24.089 --> 00:07:27.250 a high probability of going 90 days past due. 00:07:27.389 --> 00:07:29.980 Wow. These advanced techniques allow them to 00:07:29.980 --> 00:07:32.639 segment and price risk with much higher precision 00:07:32.639 --> 00:07:35.959 than ever before. This complexity dramatically 00:07:35.959 --> 00:07:38.920 increases the predictive power, which, of course, 00:07:38.920 --> 00:07:41.480 is what the lenders are paying for. And that 00:07:41.480 --> 00:07:44.100 shift towards superior precision was heavily 00:07:44.100 --> 00:07:46.639 enabled in Australia by a major policy change, 00:07:46.819 --> 00:07:48.939 the move from negative to positive reporting. 00:07:49.199 --> 00:07:51.600 That's right. And this change is a defining moment 00:07:51.600 --> 00:07:53.480 in the modern history of credit scoring globally. 00:07:54.199 --> 00:07:56.620 Australia previously operated under a negative 00:07:56.620 --> 00:07:59.339 reporting system. Meaning it only tracked the 00:07:59.339 --> 00:08:01.579 bad stuff. Pretty much. It only tracked applications 00:08:01.579 --> 00:08:03.759 for credit and adverse listings, so things like 00:08:03.759 --> 00:08:06.139 defaults or bankruptcies. If you paid all your 00:08:06.139 --> 00:08:08.500 bills on time, that positive behavior wasn't 00:08:08.500 --> 00:08:11.160 recorded or used to boost your score. And that 00:08:11.160 --> 00:08:13.500 inherently disadvantages people who are financially 00:08:13.500 --> 00:08:15.779 responsible but maybe don't have a long track 00:08:15.779 --> 00:08:18.759 record. Absolutely. When they introduced positive 00:08:18.759 --> 00:08:21.879 reporting, the credit files suddenly contained 00:08:21.879 --> 00:08:25.180 much richer information. Crucially, a history 00:08:25.180 --> 00:08:27.939 of timely payments, the types of loans you hold, 00:08:28.100 --> 00:08:30.379 and how well you manage them. So all that good 00:08:30.379 --> 00:08:33.539 behavior started counting. It did. This influx 00:08:33.539 --> 00:08:36.259 of positive data drastically increased the predictive 00:08:36.259 --> 00:08:39.320 utility of the scores and led to a greater uptake 00:08:39.320 --> 00:08:42.179 in their usage, especially for risk -based pricing, 00:08:42.360 --> 00:08:44.419 where the interest rate you receive is directly 00:08:44.419 --> 00:08:48.039 tied to your individual score. This move really 00:08:48.039 --> 00:08:50.659 cemented the statistical predictive model in 00:08:50.659 --> 00:08:53.269 Australia. Before we jump into the U .S. system, 00:08:53.389 --> 00:08:55.210 let's just make a quick technical distinction 00:08:55.210 --> 00:08:57.529 that the source material drew. We are focusing 00:08:57.529 --> 00:09:00.169 on credit scores here, not credit ratings. Correct. 00:09:00.269 --> 00:09:02.169 We need to be clear for the listener. The score 00:09:02.169 --> 00:09:04.769 we are discussing is for individuals, a numerical 00:09:04.769 --> 00:09:06.870 expression of their personal creditworthiness 00:09:06.870 --> 00:09:09.850 and risk. Got it. A credit rating, conversely, 00:09:09.850 --> 00:09:12.929 is the application of this concept to organizations, 00:09:13.370 --> 00:09:16.210 governments, or debt instruments like bonds. 00:09:16.700 --> 00:09:18.899 We are focused entirely on the personal report 00:09:18.899 --> 00:09:21.639 card that defines your access to capital. OK, 00:09:21.700 --> 00:09:24.100 let's pivot to the model that defines the system 00:09:24.100 --> 00:09:27.759 for much of the world. The highly commercialized 00:09:27.759 --> 00:09:30.840 multi -score approach epitomized by the United 00:09:30.840 --> 00:09:34.299 States. This is the global benchmark, but it's 00:09:34.299 --> 00:09:36.440 far more fragmented than most people realize. 00:09:36.720 --> 00:09:39.139 It really is. The U .S. system is defined by 00:09:39.139 --> 00:09:41.299 its massive scale and its inherent complexity. 00:09:41.870 --> 00:09:44.610 It relied heavily on the big three credit bureaus, 00:09:44.649 --> 00:09:49.309 Experian, TransUnion, and Equifax. They are the 00:09:49.309 --> 00:09:51.590 centralized gatekeepers of the underlying raw 00:09:51.590 --> 00:09:54.360 data used to calculate the scores. And we must 00:09:54.360 --> 00:09:56.419 reiterate a crucial point that confuses a lot 00:09:56.419 --> 00:09:58.639 of people. When we talk about the core FICO score 00:09:58.639 --> 00:10:00.960 calculation, what information is specifically 00:10:00.960 --> 00:10:03.440 excluded from that three -digit number? Yeah, 00:10:03.480 --> 00:10:05.059 this is so important. The source material is 00:10:05.059 --> 00:10:07.700 very clear. Income and employment history, or 00:10:07.700 --> 00:10:10.039 the lack thereof, are not considered by the major 00:10:10.039 --> 00:10:11.899 credit bureaus when calculating the credit score 00:10:11.899 --> 00:10:13.820 itself. So your salary doesn't matter for the 00:10:13.820 --> 00:10:16.320 score itself? If not for the score. Now, a lender 00:10:16.320 --> 00:10:18.379 will certainly look at your income and job stability 00:10:18.379 --> 00:10:20.379 separately when they're underwriting a loan. 00:10:21.230 --> 00:10:23.190 But the three digit number itself is strictly 00:10:23.190 --> 00:10:26.470 based on your credit behavior reported to the 00:10:26.470 --> 00:10:28.690 bureaus. Things like payment history, amounts 00:10:28.690 --> 00:10:31.470 owed. Right. Length of credit history, new credit 00:10:31.470 --> 00:10:35.570 and your credit mix. That's it. Now I get to 00:10:35.570 --> 00:10:38.830 the aha moment on FICO, which stands for Fair 00:10:38.830 --> 00:10:42.169 Isaac Corporation. People often talk about their 00:10:42.169 --> 00:10:45.610 FICO score as if it's this single fixed number, 00:10:45.710 --> 00:10:47.870 maybe something that updates once a month. That 00:10:47.870 --> 00:10:50.470 is the fundamental misconception we need to dismantle 00:10:50.470 --> 00:10:52.929 for every listener. If a lender told you they 00:10:52.929 --> 00:10:54.870 pulled your FICO score, you would still need 00:10:54.870 --> 00:10:57.850 to ask which one. Okay. As of 2018, there were 00:10:57.850 --> 00:11:00.610 29 different versions of FICO scores in use in 00:11:00.610 --> 00:11:03.850 the United States alone. 29? That's incredible. 00:11:04.850 --> 00:11:06.529 Why on earth do we need that many different numbers 00:11:06.529 --> 00:11:09.110 to describe the same person? It boils down to 00:11:09.110 --> 00:11:12.600 predicting specific types of debt failure. This 00:11:12.600 --> 00:11:14.940 is the concept of industry -specific scoring. 00:11:15.379 --> 00:11:18.399 An automotive lending FICO score is formulated 00:11:18.399 --> 00:11:21.320 differently than a credit card lending score. 00:11:21.539 --> 00:11:23.840 And both of those are different from a mortgage 00:11:23.840 --> 00:11:26.879 score. So the behavior that makes me risky for 00:11:26.879 --> 00:11:28.840 a credit card doesn't necessarily make me risky 00:11:28.840 --> 00:11:32.139 for a fixed auto loan. Precisely. The statistical 00:11:32.139 --> 00:11:35.000 weights are adjusted. The behaviors that predict 00:11:35.000 --> 00:11:37.639 risk for a five -year fixed installment loan 00:11:37.639 --> 00:11:40.080 on a car are different from the behaviors that 00:11:40.080 --> 00:11:42.559 predict risk for a revolving line of credit that 00:11:42.559 --> 00:11:45.059 you can run up and down. The model is designed 00:11:45.059 --> 00:11:47.860 to maximize predictive power for that specific 00:11:47.860 --> 00:11:50.700 lending market. Give us a concrete example of 00:11:50.700 --> 00:11:52.720 how that difference plays out for the consumer 00:11:52.720 --> 00:11:55.429 when those weights shift. Okay, let's go back 00:11:55.429 --> 00:11:57.929 to our examples. Take a consumer who has a solid 00:11:57.929 --> 00:12:00.970 history of several paid in full car loans. This 00:12:00.970 --> 00:12:02.990 shows they're really reliable with installment 00:12:02.990 --> 00:12:05.330 debt, but maybe they have very little credit 00:12:05.330 --> 00:12:07.929 card payment history. That individual will generally 00:12:07.929 --> 00:12:10.870 score significantly better on an automotive enhanced 00:12:10.870 --> 00:12:13.289 FICO score than they would on a bank card enhanced 00:12:13.289 --> 00:12:16.210 FICO score. Because the automotive model is rewarding 00:12:16.210 --> 00:12:18.730 the specific relevant positive behavior in its 00:12:18.730 --> 00:12:21.490 wheelhouse. It sees those paid off car loans 00:12:21.490 --> 00:12:24.750 and gives them more weight. The credit card model 00:12:24.750 --> 00:12:27.830 sees the lack of revolving credit history and 00:12:27.830 --> 00:12:30.529 might see that as a negative or at least a neutral. 00:12:30.710 --> 00:12:32.990 And the ranges themselves even vary depending 00:12:32.990 --> 00:12:35.450 on which version you're looking at. Yes. The 00:12:35.450 --> 00:12:37.850 general purpose FICO scores typically range from 00:12:37.850 --> 00:12:40.610 300 to 850, which is the range most people are 00:12:40.610 --> 00:12:43.179 familiar with. Right. However, the industry specific 00:12:43.179 --> 00:12:47.200 scores range from 250 to 900. This wider, more 00:12:47.200 --> 00:12:50.080 granular scale allows lenders to assess very 00:12:50.080 --> 00:12:52.820 specific risk profiles at the edges of the lending 00:12:52.820 --> 00:12:55.519 market. This inherent complexity is even built 00:12:55.519 --> 00:12:57.600 into the most common lending scenario in the 00:12:57.600 --> 00:12:59.840 U .S. obtaining a mortgage. The bank doesn't 00:12:59.840 --> 00:13:02.220 just pull one score, do they? No, not at all. 00:13:02.299 --> 00:13:04.379 For most U .S. mortgages originated, the lender 00:13:04.379 --> 00:13:06.379 obtains three separate scores, one from each 00:13:06.379 --> 00:13:08.519 of the big three bureaus. So one from Equifax, 00:13:08.620 --> 00:13:11.019 one from Experian, one from TransUnion. Correct. 00:13:11.139 --> 00:13:13.120 And they're often specific versions. The Beacon 00:13:13.120 --> 00:13:16.419 5 .0 score from Equifax, the FICO Model 2 score 00:13:16.419 --> 00:13:19.559 from Experian, and a Classic 04 score from TransUnion. 00:13:19.639 --> 00:13:22.659 Why the redundancy? Why pull three different... 00:13:22.879 --> 00:13:25.759 Named scores. They pull three scores for two 00:13:25.759 --> 00:13:28.419 main reasons. First, it's about redundancy and 00:13:28.419 --> 00:13:31.000 error mitigation. Each bureau might hold slightly 00:13:31.000 --> 00:13:33.440 different data or weight the variable slightly 00:13:33.440 --> 00:13:36.019 differently based on how the FICO formula was 00:13:36.019 --> 00:13:38.480 applied to their raw file. Okay, that makes sense. 00:13:38.679 --> 00:13:41.200 And second, they are mandated to use the middle 00:13:41.200 --> 00:13:44.039 score of the three for the primary underwriting 00:13:44.039 --> 00:13:46.340 decision. The middle score, not the average. 00:13:46.460 --> 00:13:48.879 A middle score. Yeah. This redundancy is built 00:13:48.879 --> 00:13:51.460 into the system to buffer against potential data 00:13:51.460 --> 00:13:54.440 errors or peculiarities from a single source, 00:13:54.720 --> 00:13:57.279 ensuring a more conservative and thorough risk 00:13:57.279 --> 00:14:00.120 assessment. If we zoom out again, what is the 00:14:00.120 --> 00:14:03.600 stated objective of this entire predictive framework, 00:14:03.759 --> 00:14:05.940 specifically the FICO risk score? What is it 00:14:05.940 --> 00:14:08.340 actually trying to predict? The stated design 00:14:08.340 --> 00:14:11.039 objective is highly specific. And it's narrowly 00:14:11.039 --> 00:14:13.679 focused on lender risk profiles. It's designed 00:14:13.679 --> 00:14:15.440 to predict the likelihood that a consumer will 00:14:15.440 --> 00:14:18.460 go 90 days past due or worse in the subsequent 00:14:18.460 --> 00:14:20.500 24 months after the score has been calculated. 00:14:20.740 --> 00:14:22.799 So it's a two -year forward -looking prediction. 00:14:23.200 --> 00:14:26.080 Exactly. It's a purely forward -looking risk 00:14:26.080 --> 00:14:28.059 prediction based on past financial behavior. 00:14:28.600 --> 00:14:31.240 A high score means a low probability of that 00:14:31.240 --> 00:14:33.700 failure. A low score means a higher probability. 00:14:34.120 --> 00:14:36.720 Now, FICO is dominant, but they are not the only 00:14:36.720 --> 00:14:39.500 game in town. We have alternative models and, 00:14:39.580 --> 00:14:42.240 more importantly, proprietary ones. That is correct. 00:14:42.440 --> 00:14:44.740 We have established alternatives like Vantage 00:14:44.740 --> 00:14:46.940 Score, which is often offered to consumers by 00:14:46.940 --> 00:14:49.700 the bureaus themselves. TransUnion, for example, 00:14:49.700 --> 00:14:52.860 offers Vantage 3 .0. But perhaps more impactful 00:14:52.860 --> 00:14:55.620 are the proprietary scoring models developed 00:14:55.620 --> 00:14:58.820 internally by many large lenders, especially 00:14:58.820 --> 00:15:01.519 major credit card issuers and auto finance companies. 00:15:01.840 --> 00:15:03.960 So they take the bureau data but then run it 00:15:03.960 --> 00:15:05.740 through their own secret sauce. That's a great 00:15:05.740 --> 00:15:07.820 way to put it. They mix in their own customer 00:15:07.820 --> 00:15:10.460 history and statistical weights, making them 00:15:10.460 --> 00:15:12.559 true trade secrets that give them a competitive 00:15:12.559 --> 00:15:15.259 edge. Let's quickly talk about consumer rights, 00:15:15.379 --> 00:15:17.340 especially concerning accessing this information. 00:15:17.799 --> 00:15:20.200 We need to clearly define the difference between 00:15:20.200 --> 00:15:23.600 a hard and soft inquiry here. Yes, this is a 00:15:23.600 --> 00:15:26.539 point of frequent confusion for consumers. A 00:15:26.539 --> 00:15:28.820 hard inquiry occurs when you apply for a new 00:15:28.820 --> 00:15:31.919 loan or a new credit card. It signals to other 00:15:31.919 --> 00:15:34.159 lenders that you are actively seeking credit. 00:15:34.509 --> 00:15:36.929 and multiple hard inquiries in a short period 00:15:36.929 --> 00:15:39.269 can negatively affect your score. Okay, that's 00:15:39.269 --> 00:15:41.570 the one to watch out for. Right. A soft inquiry, 00:15:41.730 --> 00:15:43.909 however, is typically done by you checking your 00:15:43.909 --> 00:15:46.950 own score or by a lender pre -screening you for 00:15:46.950 --> 00:15:49.289 an offer. It does not impact your score at all. 00:15:49.409 --> 00:15:52.870 So how does the U .S. regulate access to the 00:15:52.870 --> 00:15:55.879 data that generates these inquiries? In the United 00:15:55.879 --> 00:15:58.620 States, Americans are legally entitled to one 00:15:58.620 --> 00:16:01.779 free credit report annually from each of the 00:16:01.779 --> 00:16:04.639 three bureaus via annualcreditreport .com. One 00:16:04.639 --> 00:16:07.620 free report. Yes, but here's the catch. They 00:16:07.620 --> 00:16:09.559 are not automatically entitled to a free score 00:16:09.559 --> 00:16:11.779 unless adverse action is taken against them, 00:16:11.960 --> 00:16:14.240 a right that was mandated by the 2010 Wall Street 00:16:14.240 --> 00:16:16.539 Reform Bill. So you get the report for free, 00:16:16.600 --> 00:16:18.299 but you often have to pay for the score itself. 00:16:18.700 --> 00:16:20.519 Or sign up for some kind of credit monitoring 00:16:20.519 --> 00:16:22.860 trial, yes. But wait, if the Canadian system 00:16:22.860 --> 00:16:25.299 lets you order a free printed report any number 00:16:25.299 --> 00:16:27.779 of times in a year and the U .S. system restricts 00:16:27.779 --> 00:16:30.779 you, why are both using the same big three bureaus? 00:16:31.159 --> 00:16:34.080 Is the difference purely regulatory or is there 00:16:34.080 --> 00:16:36.100 a commercial incentive the U .S. bureaus are 00:16:36.100 --> 00:16:38.799 protecting? The difference is primarily regulatory, 00:16:38.899 --> 00:16:40.919 but it's certainly driven by commercial protection. 00:16:41.360 --> 00:16:44.419 In Canada, the consumer may order a free printed 00:16:44.419 --> 00:16:46.820 copy of their full credit report any number of 00:16:46.820 --> 00:16:49.399 times in a year, provided the request is submitted 00:16:49.399 --> 00:16:51.929 in writing and delivered by mail. Any number 00:16:51.929 --> 00:16:54.710 of times. Any number of times. The bureaus must 00:16:54.710 --> 00:16:57.350 provide it. And crucially, they confirmed that 00:16:57.350 --> 00:16:59.769 this request is classified as a soft inquiry 00:16:59.769 --> 00:17:03.009 and does not affect the score. This contrasts 00:17:03.009 --> 00:17:05.089 sharply with the U .S. system, where selling 00:17:05.089 --> 00:17:07.230 the score is a significant revenue stream for 00:17:07.230 --> 00:17:09.670 both the bureaus and for FICO. The Canadian model 00:17:09.670 --> 00:17:12.269 offers far more transparency regarding the data, 00:17:12.349 --> 00:17:15.049 even though they use the same 300 to 900 score 00:17:15.049 --> 00:17:18.049 range as the U .S., using scores like Equifax 00:17:18.049 --> 00:17:21.289 Beacon and TransUnion Empirica. Exactly. It shows 00:17:21.289 --> 00:17:24.430 that the mechanism, the score calculation, can 00:17:24.430 --> 00:17:26.970 be commercially driven, but the access rights 00:17:26.970 --> 00:17:29.609 are fundamentally determined by national regulation 00:17:29.609 --> 00:17:32.170 and the philosophy toward consumer information. 00:17:32.450 --> 00:17:34.970 So we see this commercialized, bureau -driven 00:17:34.970 --> 00:17:37.670 model spreading globally. Let's look quickly 00:17:37.670 --> 00:17:40.029 at India and Brazil, which have adopted this 00:17:40.029 --> 00:17:42.809 structural DNA, but adapted it to unique local 00:17:42.809 --> 00:17:45.940 challenges. India presents fascinating challenges 00:17:45.940 --> 00:17:49.279 due to its rapid economic growth and its history 00:17:49.279 --> 00:17:52.180 of cash transactions. They have four licensed 00:17:52.180 --> 00:17:54.579 credit information companies, with TransUnion 00:17:54.579 --> 00:17:57.019 Sybil being the most popular, along with Experian 00:17:57.019 --> 00:18:00.079 and Equifax. The Sybil score also runs from 300 00:18:00.079 --> 00:18:02.789 to 900. What's unique about the Indian system 00:18:02.789 --> 00:18:04.950 is how it handles people who are just entering 00:18:04.950 --> 00:18:06.970 the formal banking structure, the so -called 00:18:06.970 --> 00:18:09.349 thin file. That's the crucial adaptation for 00:18:09.349 --> 00:18:11.930 financial inclusion. India has specific scores 00:18:11.930 --> 00:18:14.609 for consumers with very little history. An individual 00:18:14.609 --> 00:18:16.950 with no credit history gets a negative one score. 00:18:17.130 --> 00:18:19.769 A negative one. A negative one. And those with 00:18:19.769 --> 00:18:22.710 less than six months of history get a zero. This 00:18:22.710 --> 00:18:24.710 is an explicit acknowledgement that a lack of 00:18:24.710 --> 00:18:28.250 history isn't necessarily a high risk, but rather 00:18:28.250 --> 00:18:31.200 a lack of data. It forces the system to treat 00:18:31.200 --> 00:18:32.880 these people differently than someone who has 00:18:32.880 --> 00:18:35.519 a history of default. It does. It then takes 00:18:35.519 --> 00:18:38.839 18 to 36 months or more to build a satisfactory 00:18:38.839 --> 00:18:41.960 score once data starts accumulating. It's a system 00:18:41.960 --> 00:18:44.079 designed to onboard people into the credit economy. 00:18:44.319 --> 00:18:46.680 And Brazil, which has transitioned rapidly from 00:18:46.680 --> 00:18:49.460 an older blacklist model to this predictive system. 00:18:49.599 --> 00:18:51.799 Brazil's system is very similar to the U .S. 00:18:51.799 --> 00:18:53.859 structure now, which was a significant regulatory 00:18:53.859 --> 00:18:56.859 shift. Their scores range from zero to a thousand. 00:18:57.400 --> 00:19:00.019 It analyzes mandatory factors like timely payments, 00:19:00.180 --> 00:19:02.240 negative debts, and financial relationships. 00:19:02.500 --> 00:19:04.579 So a pretty big change for them. A huge change. 00:19:04.819 --> 00:19:07.259 This transition was complex, but it was necessary 00:19:07.259 --> 00:19:09.759 to modernize their lending market and shift from 00:19:09.759 --> 00:19:12.319 purely reactive debt enforcement to proactive 00:19:12.319 --> 00:19:15.650 risk prediction. It underscores the global consensus 00:19:15.650 --> 00:19:18.230 that predictive scoring, when coupled with positive 00:19:18.230 --> 00:19:20.650 reporting, provides a more efficient mechanism 00:19:20.650 --> 00:19:23.609 for capital allocation. Okay, we are now pivoting 00:19:23.609 --> 00:19:25.890 dramatically. We've established the U .S. system 00:19:25.890 --> 00:19:29.789 is about fragmentation, proprietary models, and, 00:19:29.869 --> 00:19:32.049 you know, maximum predictive power for commercial 00:19:32.049 --> 00:19:35.380 profit. Now let's travel to Europe. Where the 00:19:35.380 --> 00:19:38.500 core philosophy shifts 180 degrees. It really 00:19:38.500 --> 00:19:41.200 does. Away from prediction and towards strict 00:19:41.200 --> 00:19:44.599 legal enforcement, public transparency, and robust 00:19:44.599 --> 00:19:48.579 consumer control. Let's start with Austria. This 00:19:48.579 --> 00:19:51.119 represents a fundamental philosophical difference 00:19:51.119 --> 00:19:54.299 in how trust is defined. In Austria, credit scoring 00:19:54.299 --> 00:19:56.759 is primarily structured as a blacklist model. 00:19:56.900 --> 00:20:00.319 A blacklist? Yes. The system focuses almost exclusively 00:20:00.319 --> 00:20:02.799 on identifying consumers who have failed to pay 00:20:02.799 --> 00:20:05.240 their bills, and those names are compiled onto 00:20:05.240 --> 00:20:08.460 lists held by various credit bureaus. So trustworthiness 00:20:08.460 --> 00:20:10.859 isn't measured by how well you manage credit, 00:20:11.019 --> 00:20:13.279 but by whether you've ever failed to meet a basic 00:20:13.279 --> 00:20:15.819 obligation. That's the distinction. The score 00:20:15.819 --> 00:20:17.720 is often less predictive and more a reflection 00:20:17.720 --> 00:20:20.599 of documented unresolved debt failure. An entry 00:20:20.599 --> 00:20:22.700 on the blacklist can lead to the immediate denial 00:20:22.700 --> 00:20:25.420 of contracts, even for non -banking services. 00:20:25.680 --> 00:20:28.500 Like a phone contract. Exactly. Telecom carriers, 00:20:28.680 --> 00:20:31.539 for instance, use this list daily to vet new 00:20:31.539 --> 00:20:34.299 customers and avoid the risk of bad debt. While 00:20:34.299 --> 00:20:37.160 banks still use these lists, for larger loans, 00:20:37.279 --> 00:20:39.420 they typically focus more heavily on traditional 00:20:39.420 --> 00:20:42.279 factors like collateral security and verified 00:20:42.279 --> 00:20:44.869 income. What's most remarkable about the Austrian 00:20:44.869 --> 00:20:47.789 system, however, is the strength of the consumer 00:20:47.789 --> 00:20:50.569 privacy protections governed by law. This is 00:20:50.569 --> 00:20:53.130 the antithesis of the U .S. approach. Absolutely. 00:20:53.210 --> 00:20:55.529 The Austrian Data Protection Act provides some 00:20:55.529 --> 00:20:57.470 of the strongest consumer controls we've seen 00:20:57.470 --> 00:21:00.529 globally. The law specifies that consumers must 00:21:00.529 --> 00:21:03.250 actively opt in for their private data to be 00:21:03.250 --> 00:21:05.549 used for credit scoring or any third -party assessment 00:21:05.549 --> 00:21:07.730 purpose. You have to opt in so it's not automatic. 00:21:08.069 --> 00:21:10.509 Not at all. If you don't consent, your data cannot 00:21:10.509 --> 00:21:13.660 be shared or used. Period. That is a massive 00:21:13.660 --> 00:21:17.000 level of control. Imagine being able to retroactively 00:21:17.000 --> 00:21:19.299 revoke permission for your data to be used by 00:21:19.299 --> 00:21:22.579 credit bureaus after the fact. You can. And if 00:21:22.579 --> 00:21:25.359 you withhold permission later, any further distribution 00:21:25.359 --> 00:21:27.960 or use of that collected data becomes explicitly 00:21:27.960 --> 00:21:31.519 illegal. This speaks directly to a privacy -first 00:21:31.519 --> 00:21:34.339 approach, prioritizing the individual's control 00:21:34.339 --> 00:21:37.339 over their personal financial information above 00:21:37.339 --> 00:21:39.579 the commercial efficiency of the lending market. 00:21:39.819 --> 00:21:42.240 So what else does that act guarantee? Beyond 00:21:42.240 --> 00:21:44.279 the opt -in requirement, consumers in Austria 00:21:44.279 --> 00:21:46.579 have the right to receive a free copy of all 00:21:46.579 --> 00:21:49.539 data held by credit bureaus once a year, and, 00:21:49.579 --> 00:21:52.359 crucially, any wrong or unlawfully collected 00:21:52.359 --> 00:21:55.480 data must be deleted or corrected quickly. The 00:21:55.480 --> 00:21:58.000 emphasis is on accuracy and consent, and it's 00:21:58.000 --> 00:22:00.839 all enforced by law. Now let's move to Sweden, 00:22:00.920 --> 00:22:03.160 which takes this debt enforcement approach to 00:22:03.160 --> 00:22:05.819 an even stricter state -driven level. It ties 00:22:05.819 --> 00:22:08.119 creditworthiness directly to failure to pay taxes 00:22:08.119 --> 00:22:10.700 or bills, managed through the Swedish Enforcement 00:22:10.700 --> 00:22:13.400 Authority. Sweden's system is highly rigorous 00:22:13.400 --> 00:22:16.279 and focused on financial integrity. It aims to 00:22:16.279 --> 00:22:18.440 swiftly identify people who have a history of 00:22:18.440 --> 00:22:20.720 neglect to pay any obligation, whether it's a 00:22:20.720 --> 00:22:23.460 commercial bill or a government tax. This system 00:22:23.460 --> 00:22:26.299 is driven by the Kronofagden. The Swedish Enforcement 00:22:26.299 --> 00:22:29.140 Authority. Right, which collects debts. When 00:22:29.140 --> 00:22:31.519 a debt case is forwarded to this authority, that 00:22:31.519 --> 00:22:34.519 single action generates a specific record. And 00:22:34.519 --> 00:22:36.559 that record carries the weight of the state. 00:22:36.779 --> 00:22:39.380 It does. It is called a betelnings and marketing. 00:22:40.009 --> 00:22:43.230 or a non -payment record. The mere appearance 00:22:43.230 --> 00:22:44.970 of a case being forwarded to the enforcement 00:22:44.970 --> 00:22:47.349 authority, even if it is technically a company 00:22:47.349 --> 00:22:50.089 debt, results in this record appearing on your 00:22:50.089 --> 00:22:52.549 file with private credit bureaus. And how long 00:22:52.549 --> 00:22:55.329 does that stay there? For an individual, this 00:22:55.329 --> 00:22:58.190 record is stored for mandatory three years. For 00:22:58.190 --> 00:23:00.309 a company, it's five years. And the consequences 00:23:00.309 --> 00:23:02.930 of having a Betel Nings and Merkning are incredibly 00:23:02.930 --> 00:23:05.609 difficult to overcome. They create massive hurdles 00:23:05.609 --> 00:23:08.240 across your life. Having this record makes it 00:23:08.240 --> 00:23:11.019 challenging to obtain loans, secure phone subscriptions, 00:23:11.440 --> 00:23:14.599 run an apartment, or even in some cases obtain 00:23:14.599 --> 00:23:17.640 certain jobs where you might handle cash or sensitive 00:23:17.640 --> 00:23:20.160 financial data. So it's a major red flag. An 00:23:20.160 --> 00:23:22.519 automatic, non -negotiable red flag for most 00:23:22.519 --> 00:23:24.799 consumer services. While banks still consider 00:23:24.799 --> 00:23:27.640 assets and income for very large loans for day 00:23:27.640 --> 00:23:30.579 -to -day life, it's a huge obstacle. The system 00:23:30.579 --> 00:23:33.000 is so rigid that it doesn't leave much room for 00:23:33.000 --> 00:23:35.579 simple administrative error, as demonstrated 00:23:35.579 --> 00:23:37.819 by the famous case of the Swedish astronaut. 00:23:38.380 --> 00:23:41.319 This anecdote perfectly illustrates the philosophy 00:23:41.319 --> 00:23:43.779 here. It's the ultimate lesson in administrative 00:23:43.779 --> 00:23:46.420 compliance. We're talking about the Swedish astronaut 00:23:46.420 --> 00:23:50.119 Krister Fuglsang. He received a betelnings and 00:23:50.119 --> 00:23:53.220 marketing because a car he owned passed a toll 00:23:53.220 --> 00:23:56.039 station for the Stockholm congestion tax. OK. 00:23:56.119 --> 00:23:58.720 But at the time he was living in the USA training 00:23:58.720 --> 00:24:01.140 for his first space shuttle mission and had an 00:24:01.140 --> 00:24:03.980 old invalid address registered for the car in 00:24:03.980 --> 00:24:06.279 Sweden. So the payment requests and the enforcement 00:24:06.279 --> 00:24:08.819 injunctions never reached him in time. Exactly. 00:24:08.859 --> 00:24:11.119 The notices didn't reach his valid location. 00:24:11.420 --> 00:24:14.259 And here is the philosophical pivot in the Swedish 00:24:14.259 --> 00:24:17.180 system. Failure to actively dispute an injunction 00:24:17.180 --> 00:24:19.680 issued by the enforcement authority is seen as 00:24:19.680 --> 00:24:21.759 tacitly admitting the debt. Even if you never 00:24:21.759 --> 00:24:24.319 saw the injunction? Even then. The debt itself 00:24:24.319 --> 00:24:26.779 may have been minor at all, but the failure to 00:24:26.779 --> 00:24:28.980 comply with the administrative process was the 00:24:28.980 --> 00:24:32.019 offense. The astronaut example is perfect. It 00:24:32.019 --> 00:24:34.400 means in Sweden, being financially trustworthy 00:24:34.400 --> 00:24:37.339 isn't just about debt management. It's about 00:24:37.339 --> 00:24:40.259 perfect administrative compliance, a distinction 00:24:40.259 --> 00:24:42.880 that would be completely alien to a U .S. consumer 00:24:42.880 --> 00:24:45.640 who has numerous rights and channels to dispute 00:24:45.640 --> 00:24:48.339 administrative errors months or even years later. 00:24:48.519 --> 00:24:51.059 Absolutely. And despite the fact that the case 00:24:51.059 --> 00:24:53.019 was eventually appealed and the underlying tax 00:24:53.019 --> 00:24:55.980 issue was retracted, the nonpayment record remained 00:24:55.980 --> 00:24:58.740 on his file for the full three years. For three 00:24:58.740 --> 00:25:01.569 years. For the full three years. Because, according 00:25:01.569 --> 00:25:04.069 to the strict letter of the law, it could not 00:25:04.069 --> 00:25:06.569 be retracted once officially established by the 00:25:06.569 --> 00:25:09.789 authority. His global fame and status were irrelevant 00:25:09.789 --> 00:25:11.690 to the rigidity of the financial enforcement 00:25:11.690 --> 00:25:14.609 system. It emphasizes the absolute power of the 00:25:14.609 --> 00:25:16.789 enforcement authority and the critical importance 00:25:16.789 --> 00:25:18.769 of maintaining a valid registered address in 00:25:18.769 --> 00:25:21.299 Sweden, even if you were floating in orbit. We 00:25:21.299 --> 00:25:23.579 have dissected the American model of commercial 00:25:23.579 --> 00:25:25.980 prediction and the Scandinavian model of state 00:25:25.980 --> 00:25:28.940 enforcement and privacy control. Now we are diving 00:25:28.940 --> 00:25:30.819 into a selection of countries where the rules 00:25:30.819 --> 00:25:34.059 are entirely bespoke, offering philosophies that 00:25:34.059 --> 00:25:36.900 resist standardization entirely. Yeah, these 00:25:36.900 --> 00:25:38.980 are the really unique ones. Let's start with 00:25:38.980 --> 00:25:42.279 the UK, which has perhaps the most shocking deviation 00:25:42.279 --> 00:25:45.619 from the standardized global system. Here's where 00:25:45.619 --> 00:25:48.599 it gets really interesting. The UK stands alone 00:25:48.599 --> 00:25:51.740 in strongly resisting the standardized FICO style 00:25:51.740 --> 00:25:55.000 model. The key takeaway here, and it's a big 00:25:55.000 --> 00:25:57.700 one, is that there is no such thing as a universal 00:25:57.700 --> 00:26:00.359 credit score or credit rating in the UK. Wait, 00:26:00.420 --> 00:26:03.619 say that again. No universal score. None. The 00:26:03.619 --> 00:26:05.819 number the consumer sees is generally irrelevant 00:26:05.819 --> 00:26:09.660 to the lender. So if a listener in the UK. checks 00:26:09.660 --> 00:26:12.359 their score through an agency like Experian, 00:26:12.500 --> 00:26:15.819 Aquafax or TransUnion, that number isn't what 00:26:15.819 --> 00:26:18.019 the lender uses to decide their loan application. 00:26:18.380 --> 00:26:20.880 That's the major point of confusion and frustration 00:26:20.880 --> 00:26:23.700 for British consumers. Those scores provided 00:26:23.700 --> 00:26:25.619 by the credit reference agencies are largely 00:26:25.619 --> 00:26:28.799 marketing tools for consumer awareness. The actual 00:26:28.799 --> 00:26:31.609 decision makers the banks and lenders, use their 00:26:31.609 --> 00:26:34.089 own proprietary algorithms to assess potential 00:26:34.089 --> 00:26:36.289 borrowers. So every bank has its own system. 00:26:36.450 --> 00:26:38.250 And those algorithms are effectively treated 00:26:38.250 --> 00:26:41.569 as trade secrets. That secrecy creates a massive 00:26:41.569 --> 00:26:45.029 commercial black box. The consumer feels powerless 00:26:45.029 --> 00:26:48.529 because they can't possibly fix or game a system 00:26:48.529 --> 00:26:51.470 they can't even see the inputs for. That is the 00:26:51.470 --> 00:26:54.150 practical consequence. The most popular statistical 00:26:54.150 --> 00:26:57.369 technique used by UK lenders is still the foundation 00:26:57.369 --> 00:27:01.220 of modern scoring. Logistic regression. The binary 00:27:01.220 --> 00:27:03.740 outcome predictor. Right. It's designed to predict 00:27:03.740 --> 00:27:07.359 a binary outcome. Bad debt, meaning the borrower 00:27:07.359 --> 00:27:10.299 has defaulted or not. Some banks also try to 00:27:10.299 --> 00:27:12.240 predict the amount of bad debt a customer may 00:27:12.240 --> 00:27:14.900 incur, but that binary default prediction remains 00:27:14.900 --> 00:27:17.480 standard. But the lack of transparency seems 00:27:17.480 --> 00:27:19.900 almost deliberate compared to the U .S. model, 00:27:20.079 --> 00:27:22.960 where despite its complexity, the FICO objective 00:27:22.960 --> 00:27:25.940 is at least clearly stated. In the U .K., lenders 00:27:25.940 --> 00:27:28.079 are under no legal obligation to reveal their 00:27:28.079 --> 00:27:30.480 internal scoring model, nor are they required 00:27:30.480 --> 00:27:32.839 to reveal the minimum score required for acceptance. 00:27:33.140 --> 00:27:35.220 And if you get rejected? If an applicant has 00:27:35.220 --> 00:27:37.440 declined, the lender is not obliged to reveal 00:27:37.440 --> 00:27:40.259 the exact reason why, although industry associations 00:27:40.259 --> 00:27:43.039 do require them to provide a satisfactory reason, 00:27:43.220 --> 00:27:45.819 which is often something vague. like poor payment 00:27:45.819 --> 00:27:48.720 history. So it's a very opaque system. Extremely. 00:27:48.779 --> 00:27:51.720 This radical proprietary secrecy is built into 00:27:51.720 --> 00:27:54.420 the foundation of the UK credit system and is 00:27:54.420 --> 00:27:56.799 the antithesis of the regulatory transparency 00:27:56.799 --> 00:27:59.640 we just saw in Scandinavia. Let's jump across 00:27:59.640 --> 00:28:02.079 the Irish Sea to Ireland, which employs a highly 00:28:02.079 --> 00:28:04.319 organized dual system that attempts to balance 00:28:04.319 --> 00:28:06.740 private commercial data with mandatory government 00:28:06.740 --> 00:28:09.859 oversight. Ireland utilizes two parallel registers 00:28:09.859 --> 00:28:13.059 operating simultaneously. First, there's the 00:28:13.059 --> 00:28:15.759 Private Irish Credit Bureau, the ICB, which is 00:28:15.759 --> 00:28:18.140 financed by its members, the financial institutions 00:28:18.140 --> 00:28:20.980 themselves. OK. And the ICB generates a score 00:28:20.980 --> 00:28:23.680 that runs in a remarkably tight range, only from 00:28:23.680 --> 00:28:27.500 224 to 581. That's a weird range. And is 224 00:28:27.500 --> 00:28:30.279 good or bad? Counterintuitively, 224 is the lowest 00:28:30.279 --> 00:28:32.880 risk and 581 is the highest risk. OK, so a low 00:28:32.880 --> 00:28:35.319 score is good there. Correct. And the government's 00:28:35.319 --> 00:28:37.859 mandatory role in this dual system is the Central 00:28:37.859 --> 00:28:43.240 Credit Register, or CCR, which is main. And lenders 00:28:43.240 --> 00:28:59.480 have to use this. So the government has the final 00:28:59.480 --> 00:29:03.000 say on the data for... Big loans. It creates 00:29:03.000 --> 00:29:05.460 a strong government backed baseline for large 00:29:05.460 --> 00:29:08.500 loan risk assessment, bypassing the need for 00:29:08.500 --> 00:29:10.660 consumer consent because the information is deemed 00:29:10.660 --> 00:29:13.400 critical for financial stability. And information 00:29:13.400 --> 00:29:16.240 is removed from both registers five years after 00:29:16.240 --> 00:29:18.279 the loan is fully repaid. Moving down to the 00:29:18.279 --> 00:29:20.460 southern hemisphere, South Africa is highly interesting 00:29:20.460 --> 00:29:23.019 because they leverage a wide range of data, both 00:29:23.019 --> 00:29:25.539 positive and negative, to achieve exceptional 00:29:25.539 --> 00:29:27.940 predictive power, often in a rapidly developing 00:29:27.940 --> 00:29:30.720 market context. South Africa's system is immensely 00:29:30.720 --> 00:29:33.200 diverse in its application. It's widely used 00:29:33.200 --> 00:29:35.859 not just by banks, but by micro lenders, clothing 00:29:35.859 --> 00:29:38.380 and furniture retailers and insurers. So it's 00:29:38.380 --> 00:29:40.819 very integrated into daily commerce. Very. And 00:29:40.819 --> 00:29:42.900 the data storage is comprehensive, including 00:29:42.900 --> 00:29:45.680 both positive and negative data, which, as we 00:29:45.680 --> 00:29:47.779 noted earlier, drastically increases the predictive 00:29:47.779 --> 00:29:50.059 power of the scores and allows for more nuanced 00:29:50.059 --> 00:29:52.380 pricing. And this is where they have embraced 00:29:52.380 --> 00:29:55.059 advanced scoring techniques, moving beyond just 00:29:55.059 --> 00:29:57.200 payment history and debt levels into something 00:29:57.200 --> 00:30:00.759 truly unique, psychometric assessments. This 00:30:00.759 --> 00:30:03.640 is a major innovation, and it's a direct response 00:30:03.640 --> 00:30:07.039 to the challenge of financial inclusion. Compusk, 00:30:07.059 --> 00:30:09.140 one of the bureaus, introduced the Compuscore 00:30:09.140 --> 00:30:12.900 PSY. This is a three digit psychometric based 00:30:12.900 --> 00:30:16.380 score. A psychometric score. Yes. It is specifically 00:30:16.380 --> 00:30:18.720 used by lenders to make informed decisions on 00:30:18.720 --> 00:30:21.539 people with thin files, those with very little 00:30:21.539 --> 00:30:24.079 formal credit history, or those who might be 00:30:24.079 --> 00:30:26.339 marginal declines based on traditional data. 00:30:26.519 --> 00:30:29.119 So how does a psychometric score work in this 00:30:29.119 --> 00:30:31.400 context? What kind of behavioral traits are they 00:30:31.400 --> 00:30:33.319 trying to measure when the traditional numbers 00:30:33.319 --> 00:30:35.789 aren't there? They are essentially applying behavioral 00:30:35.789 --> 00:30:38.750 economics to lending. These scores use questionnaire 00:30:38.750 --> 00:30:41.230 data or transactional patterns to gauge traits 00:30:41.230 --> 00:30:44.230 related to stability, future orientation, and 00:30:44.230 --> 00:30:45.930 financial discipline. So they're trying to figure 00:30:45.930 --> 00:30:47.930 out if you're a responsible person, even if you've 00:30:47.930 --> 00:30:50.670 never had a credit card? That's the idea. They 00:30:50.670 --> 00:30:52.549 are trying to determine if a person demonstrates 00:30:52.549 --> 00:30:54.690 characteristics that correlate with financial 00:30:54.690 --> 00:30:57.170 responsibility, even if they haven't yet taken 00:30:57.170 --> 00:31:00.339 out a formal bank loan. This helps lenders responsibly 00:31:00.339 --> 00:31:03.259 extend credit to populations traditionally underserved 00:31:03.259 --> 00:31:06.500 by rigid scoring models. It's an attempt to substitute 00:31:06.500 --> 00:31:08.859 character assessment for history when history 00:31:08.859 --> 00:31:12.329 is scarce. That's a fascinating layer using psychology 00:31:12.329 --> 00:31:15.170 to predict financial trustworthiness. It shows 00:31:15.170 --> 00:31:17.529 the incredible lengths the industry will go to 00:31:17.529 --> 00:31:20.430 maximize predictive capacity, especially in emerging 00:31:20.430 --> 00:31:23.170 markets. It is a dynamic response to a practical 00:31:23.170 --> 00:31:26.250 problem. How do you assess millions of people 00:31:26.250 --> 00:31:29.029 moving from a cash economy into a formal digital 00:31:29.029 --> 00:31:31.930 economy? It's a really innovative solution. Now, 00:31:31.930 --> 00:31:33.829 turning to Asia, let's quickly touch on China 00:31:33.829 --> 00:31:35.789 and Sri Lanka and then wrap up our European tour. 00:31:36.329 --> 00:31:38.509 In China, we focus on the privately developed 00:31:38.509 --> 00:31:41.230 credit score systems, primarily Sesame Credit, 00:31:41.470 --> 00:31:44.130 which is provided by Alibaba, Affiliate Ant Financial, 00:31:44.309 --> 00:31:46.809 and also Tencent Credit. So the big tech companies. 00:31:46.970 --> 00:31:50.430 The biggest. These are massive proprietary systems 00:31:50.430 --> 00:31:53.769 tied intrinsically to digital commerce and usage 00:31:53.769 --> 00:31:56.670 patterns. While the source material links them 00:31:56.670 --> 00:31:59.329 to the broader social credit system, our focus 00:31:59.329 --> 00:32:01.589 remains on the financial mechanism that grants 00:32:01.589 --> 00:32:04.250 access to loans and services based on purchasing 00:32:04.250 --> 00:32:07.819 and digital behavior. And Sri Lanka, a more centralized 00:32:07.819 --> 00:32:10.019 system. Correct. In Sri Lanka, the system is 00:32:10.019 --> 00:32:12.480 managed by CRIB, the Credit Information Bureau, 00:32:12.619 --> 00:32:15.420 which is a central authority. Individuals can 00:32:15.420 --> 00:32:18.299 request a self -inquiry credit report, though 00:32:18.299 --> 00:32:21.200 they are subject to a fee. This is a model common 00:32:21.200 --> 00:32:23.440 in many smaller economies where the state or 00:32:23.440 --> 00:32:25.559 a state -approved entity maintains the primary 00:32:25.559 --> 00:32:28.420 database. Finally, let's look at Denmark, Norway, 00:32:28.559 --> 00:32:31.569 and Germany, which reinforce that core privacy 00:32:31.569 --> 00:32:34.130 first philosophy we saw earlier. Denmark has 00:32:34.130 --> 00:32:37.309 a peculiar split in its scoring focus. Danish 00:32:37.309 --> 00:32:39.970 credit scoring is divided rigorously. Private 00:32:39.970 --> 00:32:42.049 scoring assesses the probability of defaulting 00:32:42.049 --> 00:32:44.769 for an individual, while business scoring assesses 00:32:44.769 --> 00:32:46.450 the probability of bankruptcy for a company. 00:32:46.589 --> 00:32:49.369 So two very separate tracks. Very separate. The 00:32:49.369 --> 00:32:51.829 models rely primarily on applicant -provided 00:32:51.829 --> 00:32:54.549 information and publicly available data, but 00:32:54.549 --> 00:32:56.970 the entire system is heavily restricted by legislation 00:32:56.970 --> 00:32:59.589 compared to neighboring countries, which indicates 00:32:59.589 --> 00:33:01.970 a very cautious approach to consumer data use. 00:33:02.210 --> 00:33:04.559 And what about Norway and Germany? They share 00:33:04.559 --> 00:33:06.940 a common feature regarding consumer rights that 00:33:06.940 --> 00:33:09.920 contrasts strongly with the U .S. model. Both 00:33:09.920 --> 00:33:12.279 Norway and Germany ensure that consumers have 00:33:12.279 --> 00:33:14.880 the robust right to receive a free copy of all 00:33:14.880 --> 00:33:17.920 data held by credit bureaus once a year. This 00:33:17.920 --> 00:33:20.279 mirrors the privacy focused elements we saw in 00:33:20.279 --> 00:33:23.349 Austria. But Norway's system has a twist. It 00:33:23.349 --> 00:33:26.450 does. Norway's scoring mechanism relies heavily 00:33:26.450 --> 00:33:29.069 on publicly available information. We're talking 00:33:29.069 --> 00:33:32.009 about tax returns, taxable income, and any non 00:33:32.009 --> 00:33:35.930 -payment records. Imagine if the IRS published 00:33:35.930 --> 00:33:38.410 your income and debts annually. That's the level 00:33:38.410 --> 00:33:41.150 of financial transparency required there. Precisely. 00:33:41.450 --> 00:33:44.710 It links financial transparency to credit access 00:33:44.710 --> 00:33:47.480 in a way that would be entirely foreign. and 00:33:47.480 --> 00:33:49.660 almost certainly illegal under current data privacy 00:33:49.660 --> 00:33:53.279 laws in the U .S. context. Your taxable income 00:33:53.279 --> 00:33:56.339 is an explicit component of the score, contrasting 00:33:56.339 --> 00:33:58.660 sharply with the U .S. model, where income is 00:33:58.660 --> 00:34:00.859 deliberately excluded from the calculation. It's 00:34:00.859 --> 00:34:03.960 a totally different societal choice. It is. It 00:34:03.960 --> 00:34:06.259 reflects a societal prioritization of public 00:34:06.259 --> 00:34:08.699 information and financial integrity over individual 00:34:08.699 --> 00:34:11.469 financial privacy. So we have just traversed 00:34:11.469 --> 00:34:14.030 a financial landscape of 14 countries and the 00:34:14.030 --> 00:34:17.130 differences are astonishing. We've seen three 00:34:17.130 --> 00:34:19.630 completely different philosophies govern how 00:34:19.630 --> 00:34:22.530 financial trust is established globally. So what 00:34:22.530 --> 00:34:24.010 does this all mean for the learner who wants 00:34:24.010 --> 00:34:26.429 to be truly well informed about the global report 00:34:26.429 --> 00:34:28.929 card? It means you have to shed the idea that 00:34:28.929 --> 00:34:31.710 the U .S. model defined by statistical prediction 00:34:31.710 --> 00:34:34.150 and commercial competition is the template. The 00:34:34.150 --> 00:34:36.679 philosophical split is just so stark. On one 00:34:36.679 --> 00:34:38.559 side, you have the U .S. and the commercialized 00:34:38.559 --> 00:34:41.739 systems like Brazil and India focused on maximizing 00:34:41.739 --> 00:34:44.099 predictive capacity through multiple proprietary 00:34:44.099 --> 00:34:47.539 and highly segmented scores, all designed to 00:34:47.539 --> 00:34:50.480 identify both risk and profit centers. Then on 00:34:50.480 --> 00:34:52.599 the second side, you have the Scandinavian and 00:34:52.599 --> 00:34:54.860 Austrian models, which are focused on enforcement 00:34:54.860 --> 00:34:57.579 and control. They prioritize legal debt compliance 00:34:57.579 --> 00:35:00.400 and grant the consumer powerful active consent 00:35:00.400 --> 00:35:03.420 rights over the use of their data. So trust is 00:35:03.420 --> 00:35:06.199 enforced by the state. Not primarily predicted 00:35:06.199 --> 00:35:08.179 by a private algorithm. Yes. On the third side, 00:35:08.340 --> 00:35:10.500 you have the UK model, which is defined by intense 00:35:10.500 --> 00:35:13.559 corporate secrecy. The score you see is a ghost 00:35:13.559 --> 00:35:16.000 score. And the real algorithms are proprietary 00:35:16.000 --> 00:35:18.340 trade secrets that lend themselves to commercial 00:35:18.340 --> 00:35:21.000 efficiency, but offer almost zero transparency 00:35:21.000 --> 00:35:23.480 to the consumer. And the sheer variation in the 00:35:23.480 --> 00:35:25.619 scoring scales highlights the arbitrary nature 00:35:25.619 --> 00:35:28.239 of the number itself. I mean, we've seen scores 00:35:28.239 --> 00:35:30.980 that run from zero to a thousand in Brazil, 300 00:35:30.980 --> 00:35:33.960 to 900. the U .S., and that strangely tight range 00:35:33.960 --> 00:35:38.400 of 224 to 581 in Ireland. The number is just 00:35:38.400 --> 00:35:41.039 a marker relevant to that specific national context. 00:35:41.469 --> 00:35:44.090 And the power of that score is expanding globally, 00:35:44.349 --> 00:35:47.429 being used in incredibly diverse ways from the 00:35:47.429 --> 00:35:49.829 standard mortgage application to psychometric 00:35:49.829 --> 00:35:52.170 assessments being used in South Africa to score 00:35:52.170 --> 00:35:55.250 people with thin files. The difference often 00:35:55.250 --> 00:35:58.250 boils down to a core conflict. How much privacy 00:35:58.250 --> 00:36:00.670 are you willing to surrender for optimized prediction? 00:36:00.949 --> 00:36:04.309 That distinction. Privacy versus prediction is 00:36:04.309 --> 00:36:06.789 the fundamental fault line in global financial 00:36:06.789 --> 00:36:10.250 trust systems. And that leads us to a final provocative 00:36:10.250 --> 00:36:12.510 thought for you, the listener. We saw that the 00:36:12.510 --> 00:36:14.869 generic FICO score's stated design objective 00:36:14.869 --> 00:36:17.730 is narrowly focused on predicting your likelihood 00:36:17.730 --> 00:36:20.750 of going 90 days past due or worse in the next 00:36:20.750 --> 00:36:22.409 two years. Right, the official line. The official 00:36:22.409 --> 00:36:25.110 line. But we also know that many large lenders 00:36:25.110 --> 00:36:27.650 use their own proprietary algorithms, these commercial 00:36:27.650 --> 00:36:30.550 black boxes, that constantly adjust to market 00:36:30.550 --> 00:36:33.000 segments and maximize revenue. So what does it 00:36:33.000 --> 00:36:35.159 mean for your future? And how might your financial 00:36:35.159 --> 00:36:38.059 behavior change if the calculation of your trustworthiness 00:36:38.059 --> 00:36:41.360 relies on a proprietary algorithm that constantly 00:36:41.360 --> 00:36:44.400 adjusts to market demands? The system is designed 00:36:44.400 --> 00:36:47.159 less for universal fairness and more for maximizing 00:36:47.159 --> 00:36:49.480 a lender's profit. It raises an important question 00:36:49.480 --> 00:36:51.920 you should keep mulling over. In a proprietary 00:36:51.920 --> 00:36:54.639 scoring system, are you being evaluated primarily 00:36:54.639 --> 00:36:57.300 as a risk to be avoided or as a profit center 00:36:57.300 --> 00:36:59.239 to be optimized? That's the true nature of the 00:36:59.239 --> 00:36:59.980 Global Report Card.