I once read that meteorologists are better calibrated forecasters than stock pickers. By calibrated, I do not mean they are better or worse at forecasting, but that their level of confidence matches their accuracy. When they believe they are 80 % likely to be right, they are right about 80 % of the time. When they are uncertain, their forecasts are indeed less accurate. Stock pickers, and I suspect some economists, tend to be less well calibrated. They are not necessarily poor forecasters, but they are often overconfident, and that is a different problem altogether.

The point came from Superforecasting by Philip Tetlock and Dan Gardner, which explored how some individuals become remarkably well-calibrated forecasters through constant feedback and disciplined self-assessment. One reason meteorologists perform better is feedback. Stock pickers and economists can always point to exceptional events that explain why forecasts went wrong. Who knew there would be a drought, or that Russia would invade Ukraine, or that COVID-19 would shut down economies? There is always a convenient exogenous shock. Weather forecasters do not have that luxury. They live by feedback. They issue forecasts every day and are judged immediately by outcomes they cannot explain away. Their models evolve constantly because their results are tested constantly.

That contrast came to mind when I first encountered Computable General Equilibrium (CGE) models about a decade ago. They were intriguing. A little too clever, perhaps, and impressive in their scale. Thousands of equations, countless parameters, and from all of that emerges the predicted effect of an intervention on GDP, employment, or wages—sometimes quoted to two decimal places. At the time, it seemed remarkable, but I could not see the transparency I wanted. There was little evidence of how these models were calibrated, how reliable they were, or how confident one should be in their results.

I came across them again recently. I am not saying CGE models are unhelpful or that their results are wrong. They can be useful for exploring direction and mechanism – how substitution between sectors or between labour and capital might unfold. They are elegant frameworks for thinking through equilibrium interactions. What concerns me is the way they are sometimes used. There is often little effort to quantify uncertainty, measure accuracy, or report diagnostics that would tell us how well the model is performing. That makes me uneasy.

I would like to see genuinely robust studies that take a CGE model, generate predictions, and check them against what actually happens. Yes, there will always be surprises, but if the models barely outperform a simple trend line, the enormous complexity and opacity seem hard to justify.

Even before that, I would want to see what the model predicts with no intervention at all. CGE models are built to find a general equilibrium, but there is no guarantee the economy we feed in is itself in equilibrium. The first step should surely be to see what the model does on its own. If its natural equilibrium is very different from the starting point, that tells us something. Either the real economy is far from equilibrium, or the model’s structure is not reflecting it well.

I am not suggesting this problem occurs frequently, but the diagnostic results are rarely reported, and that makes me wonder whether it sometimes does. If the model is not allowed to reach its internal equilibrium before the intervention is applied, the change in GDP or other results could be partly driven by the model’s own adjustment toward equilibrium. It would be useful to know.

I would also like to see how stable the results are when the model is run with successive years’ data. If we input the economy as it stood each year, do the parameters and balances remain consistent? If not, that raises questions about the model’s stability. For example, can it predict the next year’s social accounting matrix (the SAM) with any meaningful accuracy? If it cannot, then I am not sure how much confidence we can place in the counterfactual results.

As I understand it, CGE models are calibrated rather than statistically estimated. They rely on factual data inputs, some parameters solved to achieve equilibrium, and elasticities imported from previous studies. Those elasticities differ across the literature, and the uncertainty around them is often considerable. How confident are we in those numbers? What happens if we vary them within plausible ranges? If small changes produce very different equilibria even before any intervention, then those elasticities are critical. If the results remain fairly stable, that is reassuring.

We could also test how much the estimated impact of a policy or intervention depends on these elasticity choices. If the results swing widely, confidence in the model’s precision should be low. And if the model’s internal uncertainty interval for GDP, wages, or productivity is wider than the intervention effect, that should be stated clearly.

In actuarial work, we routinely perform analysis of surplus to understand why results differ from expectations. We ask whether deviations stem from data, assumptions, or the model itself. That discipline drives improvement and professional scepticism.

Actuaries are not perfect. There are areas of our own work where experience analysis or model validation is weak. Some solvency projections (ORSA AvE notably, but there are others) , for example, are never properly reconciled to outcomes. That is disappointing, but at least the mindset of testing and validation is embedded in the profession. The actuarial control cycle is still taught.

We are also learning from “new” data science, which takes validation seriously. Models are trained and tested on different data, evaluated out of sample and out of time, and using techniques such as Monte Carlo Cross Validation (a new discovery for me) or k-fold cross-validation. These approaches help ensure that models perform reliably, not just on the data that built them but on data they have never seen.

I would like to see economists apply similar rigour to CGE modelling. Calibration is not validation. Without testing, backtesting, and clear communication of uncertainty, the apparent precision of these models can be misleading.

CGE models can be valuable tools, but their uncertainties and dependencies are often poorly understood, poorly expressed, and underappreciated.

Perhaps I am missing something. Perhaps these diagnostic checks are done quietly somewhere. I would be glad to be corrected by economists who understand these models better than I do. I am genuinely open to learning more.

Because if these models can withstand that kind of scrutiny, they deserve confidence. If they cannot, they still have value, but we should be honest about what they are: structured thought experiments with useful stories to understand. Not forecasts – and those second decimals places should never be shown.

Nudge Theory

• Concept: Small changes in choice presentation can significantly influence decisions.
• Application: Using social comparison feedback (e.g., “80% of homeowners in your area have flood insurance”) to encourage consideration of additional coverage. Also, implementing gamification elements in health insurance apps to promote healthier lifestyles, potentially reducing health-related claims.
• Replicability: Consistently strong results across various domains, making it a robust finding.

The Watching Eyes Effect

• Concept: Displaying images of eyes or a face can encourage honest behavior.
• Application: Used in various settings, including online insurance, to promote truthfulness.
• Replicability: Mixed results, with effectiveness varying by context.

Default Options

• Concept: Setting beneficial default choices (e.g., higher insurance coverage) can lead to better outcomes.
• Application: Widely implemented in insurance and savings products.
• Replicability: Generally reliable, though effect sizes can vary.

More on this in the excellent Freakonomics podcast episode 572

Queue management at the motor vehicle licence renewal office offers a window into a realm of complexity that extends far beyond the waiting room. Exploring priority queuing and weighted fair queuing (WFQ), we uncover a world of probabilistic models, dynamic rules, and real-world applications that challenge conventional wisdom.

Upon arrival, patrons encounter priority queuing, where elderly individuals receive absolute precedence to accommodate potential mobility challenges. While this prioritisation addresses (quite reasonably!) the needs of a vulnerable group, it can lead to excessively prolonged waits for others, depending on the arrival rate of elderly patrons and the dynamics of the queue. Consideration of different objective functions for various demographics, including the young and old, could unveil alternative optimal solutions.

Actuaries are used to implicitly allowing for objective functions that aim to minimise mean squared errors, but this isn’t the only possibility. Aside from different penalties for waiting times for all participants, explicitly considering different costs to different groups of individuals makes it more interesting.

In contrast to absolute priority queuing, WFQ introduces a probabilistic element, ensuring that even low-priority individuals have a chance of being served at every point in time. The weights can be tuned for those different groups, still achieving prioritisation for the elderly. While wait times for low-priority customers will still be longer than priority customers, they remain finite, underscoring the equitable nature of this approach. I can imagine an algorithm where the weight depends on the wait – the longer the duration of the existing wait the more an individual is prioritised.

The network packets to which WFQ most commonly applies probability aren’t conscious, and therefore their experience of the wait isn’t important. For humans, the conscious experience of knowing that you don’t have to wait until every individual in a priority group is served before you might make the wait more palatable. I’m not a fan of gamification; overall this seems to have result in evil outcomes, greedily addictive algorithms and a worse society overall. However, the introduction of a little randomness and the possibility of a serendipitous early shot at renewing a licence might be a net positive.

Beyond the motor vehicle licence renewal office, the principles of queue management find relevance in diverse fields, including insurance companies. By leveraging similar concepts, insurers can prioritise claims payment or customer service based on factors like severity or urgency, but also explicit objective functions to optimise for target response times within particular bands, aiming for prompt resolution while maintaining fairness and efficiency. Sometimes the maths will demonstrate that based on demand and supply of support calls, failing these objectives is inevitable. Modelling can determine how often that may occur and then you can weigh up the costs of more call centre agents with the unappealing customer experiences.

These waiting processes should ring bells from prior study of queuing processes, of exponential, poisson, negative binomial and gamma distributions. These provide a mathematical framework for understanding and optimising queue management, revealing the hidden complexities behind seemingly simple processes. While you may not think about claims processes as queuing processes, there are some surprising linkages there too.

For me, delving into the dynamics of network switching provided some fun insights. In network architecture, switches play a critical role in directing data packets efficiently across networks. However, the limited capacity of switches can lead to packet loss during times of congestion. Buffers within switches temporarily store packets to alleviate congestion, but if overwhelmed, they are forced to drop packets, resulting in packet loss and network underperformance.

A smarter (yet initially counterintuitive step) is to apply Random Early Detection (RED), a mechanism that deliberately drops packets in anticipation of the switch being overwhelmed. Under RED, these strategically dropped packets provide information to network endpoints to manage the speed of transmission to prevent larger problems down the line. To an extent, this can can be likened to passengers on an overbooked flight being paid to take a different flight. However, unlike RED, this approach does not alert endpoints or other passengers to slow transmission rates, highlighting the unique challenges and solutions in network management.

While queue management may appear remote from your regular work, some exploration might be worth the time spent. By embracing probabilistic models, dynamic rules, and real-world applications, we may see a way to apply our skills in new areas, or even learn from the significant advancements in network modelling in our financial services world.

Given the long silence, there are quite a few topics I’ve been thinking about and discussing with friends and colleagues. You’ll have to check in from time to time to see those discussions translate to a blog.

On the death of blogs and privacy

Blogs have changed since 2006 when this blog started. I don’t know when “peak blog† was, but it seems to be in the past. The audience monopolisers of facebook, twitter, LinkedIn, youtube, Instagram and more have, by turns, elevated and expanded voices, then monetised, controlled and manipulated those same audiences. In some, often ultimate tragic, scenarios, audiences are pushed to extreme views as an inevitable result of revenue-maximising, attention monopolising algorithms.

Brexit and Trump are the two big stories about the role of social media and revenue-maximising, attention-sink algorithms. There are countless more. I found this sobering TED talk about the disconnect between reality and perceptions. The talk uses a small Welsh village as a powerful example to demonstrate the impact of social media and at the same time the lack of transparency that makes these claims difficult to provide and all the more terrifying.

My journey away from the brink

In the last few months I’ve been on a security and privacy journey. Ironically, I harbour suspicions that this journey itself may be the result of the same pressure to extremes that youtube in particular is famed for. But each time as I try to opt out of webs of tracking and analytics cookies and code I am reminded how much tracking and identification is going on in the background and how defenceless almost all internet users are.

Sheepishly, I realise my prior habits of eagerly providing all my contact details and information to google, facebook and LinkedIn for vague, unspecified reasons makes me some of the most naïve out there.  Facebook has been dead to me for several years. Since then, the base of research highlighting facebook’s adverse role in mental health, time-wasting, election meddling and terrorism makes that decision look even better than it feels.

Some unsoliciated privacy advice

If for some reason you are still on facebook, download your data (including contacts and birthdays if that is important) and delete your profile. Then avoid the nagging siren songs for a couple of weeks from facebook to try to entice you back onto the rocks. Your life will be better.

For now, I think I will continue to use Twitter and LinkedIn to post new blogs to reach a greater audience where users have become so used to those being the only places to monitor for anything.

I still watch clips on youtube, but I do so without logging in and with tracker blocking extensions loaded up in full force. I am aggressively migrating away from Gmail (about 90% complete) and now only check Gmail via an email client that doesn’t require me to log in to google. It’s easier than you might think. I hope you too will ditch Gmail, so that my emails to you aren’t also put into the machine for their monetary purposes.

I’ve ditched Chrome and am fortunate to have an iPhone and, generally, Apple seems less utterly compromised on this front.

Whatsapp remains a problem. Facebook owns WhatsApp and collects tonnes of metadata. Download signal and use that and help me slowly wrench the network benefits away from them.

Born again without google trackers

So that’s all background to this technical update on this blog, which is actually the primary purpose of this post. Almost since inception, I have used google analytics to track users and interest on my blog. It’s useful to understand which posts are most useful and interesting. However, I feel I owe you at least as much privacy protection as I’m trying to give myself. As of a couple of days ago, I have removed google analytics from this site and am using a self-hosted tool which keeps your data out of evil hands.

You are very welcome to block this tracker too.

Future blog topics?

I will explore thoughts on privacy (and longevity, and AI and risk and capital and hyper-selection and fairness and more as usual) in coming blogs.

Possibly helpful resources on privacy

Here are some of the resources I’ve discovered and have influenced my thinking and the tools I’m using as widely as possible.

• Firefox with appropriate extensions and add ons (ublock origin, canvas blocker)
• (or if you want to go all out, TOR browser…)
• Book to get you motivated – Ten arguments to delete your social media accounts now by Jaron Lanier
• Podcast for those with long commutes with Sam Harris and Jaron Lanier (author of 10 arguments to delete your social media now)
• Signal messaging app
• If you must keep Gmail, use something like thunderbird email client (open source, free, cross-platform. There are slicker options available too. There are many more private email options compared to Gmail, but I’m not going to recommend one as suitable across a wide range of use cases and individual preferences.

My training and experience has given me high confidence that this is the right way forward for typical actuarial data.  Here I’m talking in force policy data files, movements, transactions, and so on.  This is really well structured data that will be used many times by different people and can easily be processed once, “on write”, stored in the data warehouse to be reliably and simply retrieved whenever necessary.

The particular issue was effectively about a single transaction, a single claim, being recorded as two separate items because of administration system limitations that required the claim to be processed as two equal amounts rather than a single claim. I can understand wanting to possibly also store the two separate transactions for audit trail purposes, but in terms of the claim for movements and analysis runs and checking against policy rules, it needs to reflect the product, business, actuarial model and customer reality of a single claim.

If the data is permanently stored as two separate, equal transactions, the end consumer of that data will always need to know to combine them, that they’re not duplicates (well they are, but they’re not – see how confusing it can be?). It really is standard practice for data warehouses to have “schema on write” in place. This talks to the structure of the data, but also the cleaning and transformation, application of business and validation rules to the data to fit the data definitions exactly.

And then I heard about Data Lakes. At first, these same a little like a new buzzword for data warehouse. But in fact it is quite different. The idea behind a data like is to store large amounts of unstructured data, as it comes, and figure out what to do with it later.

This is “schema on read”.

If you can imagine an actual lake, full of different things, sand and mud and fish and seaweed, but also boats and piers and old anchors and cool water and warm water and maybe the occasional Loch Ness Monster. Some of it may be structured, most of it not, and you don’t need to define what you will store ahead of time. You just throw it all in the data lake and “figure out what to do with it later”

The value of a data lake is that it is flexible. Nothing is lost from the original data. If a brand new purpose for the data arises that wasn’t even imagined when the data was originally stored, no avenues have been closed off. This is not true for a data warehouse.

However, for the types of data used for a simple actuarial valuation, locking down the exact data to be used to ensure results are consistent, that all valid policies are valued, that the analysis of experience and surplus work correctly, a rigid, reliable data source is required.

If a data lake is to be used, then another store of the highly structured, cleaned and transformed and validated data will be required to support the time-sensitive standard valuation processes. Whether this is  data warehouse feeding off the data lake, or just a simple data base, I suppose if up for debate. But a data lake on its own holds too many monsters to be used raw.

This has nothing to do with the Famous Five. While it is most common in the general insurance space, it is relevant across the spectrum of risk management and assumption setting.

Events Not In Data or “ENID” is the forgotten cousin of “what to do with outliers in your data”.

Outliers and where to find them

Outliers are observed values substantially different from others in a sample. Some more formal definitions include:

“An outlier is an observation that lies an abnormal distance from other values in a random sample from a population”

“an outlier is an observation point that is distant from other observations”

How to deal with outliers?

Simple question, complex answer. It depends a great deal on the context.

Ultimately you need to make the judgement call “are these outliers under- or over-represented in the data”.

If you believe you have a 1 in 100 year event in your 3 year data set, then leaving it in unadjusted will skew your average (as well as the rest of the distribution) away from the true underlying population distribution.

However, it is also key to ask yourself if there are Events Not In Data (ENID) that would be part of the underlying population distribution. Most 3 year samples won’t include a 1 in 100 year event and therefore might understate the average and bias the distribution.

I believe it can be good practice to consider the data with and without the data point. And without could mean capping or omitting entirely. The cut-off point for outliers can be expressed as a multiple of assumed standard deviations based on how much data you have in your sample. It’s obviously also circular in that your estimate of standard deviation depends on whether or not you include the outlier in the estimate of standard deviation!

I said at the start, it depends on the context and the underlying variable.

Putting Events Not In Data into the data

Outliers may be hard to pin down an define exactly, but they usually appear because they make the results look worse than hoped for and there is plenty of pressure to “fix it”. When experience has been good, because those outlying events did not occur, there is naturally less pressure to go looking for trouble.

Except that’s exactly what one should do.

Just because events are slightly away from the mean or occur somewhat less often doesn’t mean they shouldn’t have been considered prior to the pricing / reserving / business forecasting / assumption setting / risk management process. So in theory, the analyst should already have a view (however accurate) about what to expect away from the expectation. That can be a useful starting point to assess whether an allowance for ENID should be made.

Over time, the views on the shape of the distribution and the reasonability of ENID at various levels will be refined. That is after all, the basics of the Actuarial Control Cycle.

ENID and skewness

The world we live in seems to be naturally asymmetrical.  So many of these ENID are on the down side. Including them contributes to greater variance, greater kurtosis and also negative skewness.

Examples of variables exposed to ENID (aka where to find them)

The list is endless and it might be more practical to think of examples where it could not apply. However, just to make sure you are thinking broadly enough, here are a tiny number of examples.

• general insurance claims (the standard example)
• stock market returns
• corporate defaults
• lapse assumptions (upwards spikes in lapses due to economic conditions or regulatory changes are not every year occurrences)
• expense assumptions (where “one off” costs relating to restructuring or retrenchment are very likely to occur over a 15 year projection period, but hopefully didn’t occur in the last 3 year expense investigation period.)
• rainfall and drought (Cape Town planners, I’m looking at you)

Remember this list is incomplete, so there will be ENID in this list of ENID affected variables.

Events not in data – guidance from the Bank of England

Many firms use reserving methods that project forwards
from historical data. On its own, this is unlikely to satisfy the
Directive requirement for a probability-weighted average of
future cash-flows, since not all possible future cash-flows — or
the events that cause them — may be represented in the data.

Although these events are sometimes referred to as
‘binary events’ or ‘extreme events’, such terms suggest that
events not found in the data are necessarily extreme or rare.
This is not the case, so the PRA prefers to use the term ‘events
not in data’, or ENID.

Firms should take ENID into account when calculating
technical provisions. Applying a simple percentage uplift
without justification is not an adequate method.

Where outliers are removed from the data as part of the
reserving process, this removes events from the data. Firms
should make an allowance for this in the technical provisions
calculation unless they have shown that it would not be
possible for these, or similar, events to occur again in future.

A recent example relates to the common non-life practice of not discounting claims reserves.  Solvency II, SAM and IFRS17 moves to require discounting aside, it is still more a common GAAP approach to not discount than to discount claims reserves.

Discounting or fiddling with inflation has some obvious implications for analysing actual vs expected analysis, reserve run offs, and reserve adequacy analysis. That some non-life reserving actuaries trip over because it’s more natural in the life space.

But, first, why are non-life reserves so often not discounted? There are several reasons typically given:

• (1) Claim cash flows are too uncertain, therefore it’s not possible to know what term discount rate to apply.

This is mostly a weak reason. In insurance, all cash flows are uncertain to some or other extent. We are discounted expected values with an implied or explicit assumed probability distribution and will usually have some sense of the timing of cash flows. Large claims will cause noise in the analysis of timing and investment returns, but large claims will always cause more noise in the underlying underwriting performance anyway.

There is a link to point #5, which has merit in my mind.

• (2) Claim payment periods are usually short, so it doesn’t make a material difference to the result, while increasing complexity.

I’ll consider this combined with #3 below as they are closely related.

• (3) Interest rates are low, so it doesn’t make much difference.

Yes, discounting increases complexity. And where the claim development periods are truly short and interest rates are moderate or low, it might not be worth it.

Accounting has “materiality”, Solvency II and SAM have “proportionality” and most actuaries would apply some level of judgement around “significance”. Interestingly, none of these terms is usually well defined or, maybe more concerningly, consistently applied.

However, operating in a range of developing markets with inflation and discount rates anywhere from 3% to 25% means that even fairly short development periods can have a significant impact. The answer surely has to be “perform some quick testing to demonstrate the potential magnitude of the difference to inform a decision” rather than always just ignore it as a rule?

• (4) We want to be conservative and this introduces an element of conservatism.

This is an argument I have accepted before, but still isn’t the best way to handle this. If conservatism or profit deferral or market value margins are required, then the magnitude and run-off of these margins should be deliberate rather than arbitrary.

I’ve also seen versions of “We haven’t allowed for ULAE so these two offset, another version where the end result might not be that different but reflects a weak methodology subject to producing materially incorrect results if discounting and ULAE allowances begin to diverge.

• (5) Our actual investments need to be extremely liquid given the need to pay out large claims at short notice. Therefore we aren’t able to earn the yield implied by the yield curve. We don’t want to incur strains in future as our assets earn less than the unwind of the liabilities.

As I highlighted in #1 above, this is a real issue.  The operational and risk management necessity of maintaining sufficient liquidity, particularly in markets where the instruments used to derive the “risk free yield curve” are not necessarily liquid at all, causes a departure between the rates that theoretically (and naively) should be used to discount and what can actually be reasonably earned on backing assets.

How this issue is dealt with practically is a matter of applicable regulation and standards, professional guidance, past practice and documented policies, the purpose for which the results are being produced, and a good dose of judgement.

• (6) A related point to (5) is that insurers don’t want to show interest rate sensitive liabilities, especially when assets are not well matched.

Assets may not be well matched due to bad reasons, but also the need for liquidity, which in certain markets means duration = 0.

In part 2, how discounting and inflation adjusting interact. Part 3 will finally address the AvE, reserve adequacy analysis and impact on P&L.

Check out part 2

Check out part 3 (not yet available)

I’d go as far as to say that many of the “Big Data!” stories and analysis currently performed is really plain old statistical analysis with a few new touches from the ever-expanding list of R libraries.

For example, it seems that papers with shorter titles get more citations by other researchers.  Although the research considered 140,000 papers, there is nothing especially “Big Data!” about the analysis. The paper and authors suggest several possible causes related to the quality of the journal, period of time etc. Disappointingly, they don’t seem to have modelled these possible effects directly to understand whether there is any residual effect.

There is scope for great analysis without “Big Data!” and plenty of scope of poor analysis with all the data in the world.

To paraphrase Lehohla, he disagreed with their methodology, their data and their values and ethics:

First, I need to engage the author on methods. Second, I address the facts. Third, I focus on the morality of political systems and, finally, I question the integrity of the luminaries of DaMina and ask them to come clean.

This wasn’t data I had looked at before, but some of Lehohla’s criticisms seemed valid. Using nominal GDP growth data is close to meaningless over periods of different inflation.

Second, the methodology of interpreting economic growth should use real growth instead of nominal growth because this carries with it differing inflation rates. This is to standardise the rates across high and low inflation periods.

I haven’t confirmed the DaMina calculations, but the labels in their table do say “current USD prices” which suggests they have used nominal data. It’s little wonder any period including the 1970s looks great from a nominal growth perspective with nominal USD GDP growth in 1973 and 1974 being 34% and 23%, compared to real growth of 2.2% and 3.8%. The high inflation of the 1970s arising from oil shocks and breakdown of the gold standard distorts this analysis completely.

Lehohla’s other complaint is also important, but less straightforward to my mind –

The methods that underpin any comparison for a given country cannot be based on a currency other than that of the country concerned. The reason is that exchange-rate fluctuations exaggerate the changes beyond what they actually are.

Two problems here – one is that purchasing power adjusted GDP indices are not typically available going far back in history. The other is that if one is using real GDP, the worst of the problems of currency fluctuations are already ironed out. (The worst, certainly not all and it would still be a factor that should be analysed rather than completely overlooked.)

I was disappointed that neither piece mentioned anything at all about real GDP per capita. Does it really matter how much more we produce as a country if the income per person is declining? Income inequality aside, important as it is, more GDP per capita means more earning power per person, more income per person, more things per person. It is a far more useful measure of prosperity for a country, and particularly for comparing economic growth across countries with different population growth rates.

My own analysis, based on World Bank data (available from 1960 to 2013)

I’ve put these numbers out without much analysis. However, it’s pretty clear that on the most sensible measure (real GDP per capita) over the periods the DaMina study considered, post-apartheid growth has been better than during the 1971-1990 period of Apartheid.

The conclusion is reversed if one includes the 1960s Apartheid economy and the latest data to 2013, the picture is reversed on both measures.

This, above all else, should talk to the dangers of selecting data to suit the outcome.

This analysis doesn’t talk to the impact of the gold standard, the low cost of gold mining closer to the surface than it is now, the technological catch-up South Africa should have benefited from more in the past, the impact of international sanctions and expenditure on the old SADF and who knows what else. There are much big monsters lurking there that I am not equipped to begin to analyse.

My overall conclusion? The Apartheid days were not “economically better” even without ignoring the millions of lives damaged. Unfortunately, our economic growth has for decades been too low to progress our economy to provide a better life for all.

Here is the problem:

Despite the theory of “Convergence“, the US has had double South Africa’s per capita GDP growth for over five decades.  Real GDP per capita increased by 72% in South Africa over the entire period from 1960 to 2013, which sounds impressive until you realise that the US managed 189%. That is more than 2.5x our growth Brazil has done even better at 237%. “Even Kenya” outperformed us over this period.

1% per annum real per capita GDP growth is just not good enough.

I can only begin to imagine the data horrors of dealing with multiple insurers, multiple sources, multiple different data problems. The analysis they do is critically useful and, in technical terms, helluva interesting. I enjoyed the presentation at both the Cape Town and Johannesburg #LACseminar2013 just because there is such a rich data set and the analysis is fascinating.

I do hope they agree to my suggestion to put the entire, cleaned, anonymised data set available on the web. Different parties will want to analyse the data in different ways; there is simply no way the CSI Committee can perform every analysis and every piece of investigation that everyone might want. Making the data publicly available gives actuaries, students, academics and more the ability to perform their own analysis. And at basically no cost.

The other, slightly more defensive reason, is that mistakes do happen from time to time. I’m very aware of the topical R-R paper that was based on flawed analysis of underlying data. Mistakes happen all the time, and allowing anyone who wants to have access to the data to repeat or disprove calculations and analysis only makes the results more robust.

So, here’s hoping for open access mortality investigation data for all! And here’s thanking the CSI committee (past and current) for everything they have already done.