Current Research

Point of View: Artificial intelligence and the productivity conundrum

Wednesday, May 25th, 2016

The world’s first artificially intelligent lawyer has arrived. Called Ross, and built on IBM’s famous cognitive computer called Watson, it has been “employed” by US firm Baker & Hostetler to work in its bankruptcy practice.

According to, Ross can “read and understand language, postulate hypotheses when asked questions, research, and then generate responses (along with references and citations) to back up its conclusions. Ross also learns from experience, gaining speed and knowledge the more you interact with it”. (

It’s not just lawyers who should be looking over their shoulders. All sorts of knowledge workers could see their employment prospects and livelihoods threatened by artificial intelligence, including journalists, accountants, portfolio managers, even surgeons and physicians.

With the aid of the internet and easy access to case law and its interpretation, fewer lawyers may be required to resolve a bankruptcy procedure. Fewer analysts may be required to value a company with the aid of Bloomberg data and its accompanying suite of programmes. Lasers directed by unerringly accurate robots may well help reduce the time in the operating theatre and the dangers of doing so.

What is the impact of these newly adapted technologies on productivity in the industry and the economy as a whole? Let’s use the example of the artificially intelligent bankruptcy lawyer above. The productivity of the lawyers in bankruptcy practice can be defined as the number of cases concluded divided by the number of lawyer hours billed to do so*. Presumably the number of lawyers (and legal hours billed) will decline with the aid of Ross. Thus the surviving lawyers working on bankruptcy law in a legal practice will have become more productive in the sense of an increase in the ratio (cases concluded/hours billed). Measuring productivity in this case seems a simple task.

It becomes much more difficult to measure the productivity of a service provider when real output is much trickier, and sometimes impossible, to measure. One would not wish to measure the productivity of an analyst, journalist, artist or inventor of a new video game by the number of words written and published or number of pictures painted or pixels injected. The quality of the work produced is surely more important than the quantity of output and “quality” is recognised in revenues generated. As is admitted by the calculators of productivity, it is impossible to measure the productivity of government officials, because it is not possible to measure how much they produce. All that can be measured is their employment benefits – an input. In the case of an author, composer, copywriter or game developer, only the value of the royalties they have earned can be measured – their revenue line, not the time spent writing the masterpiece. Measuring productivity requires that inputs and outputs can be independently measured, which is not always the case, especially for service providers.

However, looking at the example of the number of bankruptcy cases (which we would regard as an independent measure of output), what if the quality of advice has improved even as the numbers of hours billed declines? The advice may be superior with the aid of Ross’s deep memory bank. How would we adjust for this quality dimension in our measure of legal productivity? The question is apposite for the service sector generally, where computers and data management (and improved knowledge) have presumably enhanced the quality of service provided by lawyers, analysts and other knowledge professions, including the improved offering of physicians supported by bigger data and better statistics. If the quality of advice has objectively improved, then any hour of consulting service will be delivering more in real terms than a case handled in the same time say 10 years before. The output of the consultant will in effect have increased, even if the input of time is the same. But by how much is the leading question. The physicians may be seeing the same number of patients a day, charging them higher fees, but they (their patients) are likely to be living longer and better lives.

In cases like this we will not be comparing like with like, apples with apples or aspirins with aspirins, making any measure of real output and so productivity over time a very difficult exercise and one subject to significant errors in what is measured. For example, your medical insurance may well have become more expensive – or your cover reduced – but are you not getting a better quality of medical service in return? And exactly how much better? In the case of medical insurance, only what you are paying – not your additional benefits – will find their way into the official price indices.

A further aspect is the impact of improved quality on the broader economy. If the bankruptcy cases are resolved with less billable time spent in court and hence with a larger percentage of debt being recovered with reduced legal expenses, this would be a clear gain to the creditors. Creditors would be better off in real terms, with less spent on legal fees and earlier resolution of their claims, meaning that the creditors could spend more on other goods or services or save more. And lawyers competing with each other for work that has become less costly for them to supply, may well charge you less for their time. It is competition for extra revenue that turns lower costs into lower prices – even in the legal profession – provided they do not collude on fees.

Could the GDP deflator, the price index that converts estimates of GDP in money of the day into a real equivalent, hope to pick this up with a high degree of accuracy? Enough to provide accurate measures of GDP or productivity growth over extended periods of time? The deflator used to convert nominal GDP into real GDP, attempts to adjust for quality improvements in the output of goods and, especially, services produced. Yet in South Africa, 68% of all value added is comprised of services of one kind or another the quality of which may well be changing over time, in ways that are very difficult to measure.

Ours is more of a service economy, than one that produces goods, the output of which is much more easily measured in units of more or less constant quality – for example number of bricks or tons of cement or steel. Thus, if we are underestimating quality improvements in the large service sector, we will be overestimating inflation and so underestimating the growth in real incomes, output and productivity.

Your real incomes and your productivity may well have increased even if you are taking home no more pay or other employment benefits. You may be benefitting from an enhanced quality of service as well as a very different mix of services than was available 10 years before, for example easy internet access that has so changed the way we work and play. This has become a particular problem in the developed world where prices as measured are generally falling. Deflation, rather than inflation, is the greater concern and nominal wages are not rising, even if productivity and the standard of living, differently measured and quality enhanced, is improving (though perhaps poorly recognised, as voters in their frustration at their constant money incomes turn to populists who promise a better standard of living). A mere one or two per cent extra a year factored into GDP or productivity growth measures, well within a range of possible measurement errors, would provide a very different impression of how the developed world is doing. A rising real standard of living, if only we could measure it, might well be accompanying stagnant employment benefits, when calculated in money of the day.

*One of the criteria the World Bank uses for measuring the ease of doing business in any country is outcomes in the bankruptcy courts. The tables below measure ease of doing business across a number of categories, and we show the SA and Australia findings where SA compares quite poorly. The ranking is, for example, 120/189 for ease of starting a business compared to 11 for Australia; and 41 for bankruptcy proceedings compared with 14 for Australia. Both countries rank poorly for trade across borders (130 and 89). Note we do much better than Australia when it comes to protecting minority investors: ranked 14 vs 66; and worse for getting credit, 59 Vs 5.


Corporate Finance

Point of View: A question of (investment) trusts

Tuesday, May 26th, 2015

Understanding investment trusts and how they can add value for shareholders regardless of any apparent discount to NAV.

Remgro, through its various iterations, has proved to be one of the JSE’s great success stories. It has consistently provided its shareholders with market beating returns. Still family controlled, it has evolved from a tobacco company into a diversified conglomerate, an investment trust, controlling subsidiary companies in finance, industry and at times mining, some stock exchange listed, others unlisted. Restructuring and unbundling, including that of its interests in Richemont, have accompanied this path of impressive value creation for patient shareholders.

The most important recent unbundling exercise undertaken by Remgro was in 2008 when its shares in British American Tobacco (BTI), acquired earlier in exchange for its SA tobacco operations, were partly unbundled to its shareholders, accompanied by a secondary listing for BTI on the JSE. A further part of the Remgro shareholding in BTI was exchanged for shares in another JSE-listed counter and investment trust, Reinet, also under the same family control, with the intention to utilise its holding of BTI shares as currency for another diversified portfolio, with a focus on offshore opportunities. Since the BTI unbundling of 2008, Remgro has provided its shareholders with an average annual return (dividends plus capital appreciation, calculated each month) of 23%. This is well ahead of the returns provided by the JSE All Share Index, which averaged 17% p.a over the same period. Yet all the while these excellent and market beating returns were being generated, the Remgro shares are calculated to have traded at less than the value of its sum of parts, that is to say, it consistently traded at a discount to its net asset value (NAV).

The implication of this discount to NAV is that at any point in time the Remgro management could have added immediate value for its shareholders by realising its higher NAV through disposal or unbundling of its holdings. In other words, the company at any point in time would have been worth more to its shareholders broken up than maintained as a continuing operation.





How then is it possible to reconcile the fact that a share that consistently outperforms the market should be so consistently undervalued by the market? It should be appreciated that any business, including a listed holding company such as Remgro, is much more than the estimated value of its parts at any moment in time. That is to say a company is more than the value of what may be called its existing business, unless it is in the process of being unwound or liquidated. It is an ongoing enterprise with a presumably long life to come. Future business activity and decisions taken will be expected to add to the value of its current activities. For a business that invests in other businesses, value can be expected to be added or lost by decisions to invest more or less in other businesses, as well as more or less in the subsidiary companies in which the trust has an established controlling interest. The more value added to be expected from upcoming investment decisions, the higher will be the value of the holding company for any given base of listed and unlisted assets (marked to market) and the net debt that make up the calculated NAV.

Supporting this assertion is the observation that not all investment trusts sell at a discount to NAV. Some, for example the shares in Berkshire Hathaway run by the famed Warren Buffet, consistently trade at a value that exceeds its sum of parts. Brait and Rockcastle, listed on the JSE, which invest in other listed and unlisted businesses, are currently valued at a significant premium to their sum of parts. Brait currently is worth at least 45% more than its own estimate of NAV while Rockcastle, a property owning holding company offers a premium over NAV of about 70%. PSG, another investment holding company, has mostly traded at a consistently small discount to NAV but is now valued almost exactly in line with its estimated NAV.

It would appear that the market expects relatively more value add to come from the investment decisions to be made by a Brait or Rockcastle or PSG, than it does from Remgro. The current value of the shares of these holding companies has risen absolutely and relatively to NAV to reflect the market’s expectation of the high internal rate of returns expected to be realised in the future as their investment programmes are unveiled. Higher (lower) expected internal rates of return are converted through share price moves into normal risk adjusted returns. The expected outperforming businesses become relatively more expensive in the share market – perhaps thereby commanding a premium over NAV – while the expected underperformers trade at a lower share price to provide the expected normal returns, so revealing a discount to NAV.

The NAV of a holding company however is merely an estimate, subject perhaps to significant measurement errors, especially when a significant proportion of the NAV is made up of unlisted assets. Any persistent discount to NAV of the Remgro kind may reflect in part an overestimate of the value of its unlisted subsidiaries included in NAV. The NAV of a holding company is defined as the sum of the market value of its listed assets, which are known with certainty, plus the estimated market value of its unlisted assets, the values of which can only be inferred with much less certainty. The more unlisted relative to listed assets held by the holding company, the less confidence can be attached to any estimates of NAV.

The share market value of the holding company will surely be influenced by the same variables, the market value of listed assets and the estimates of the value of unlisted assets minus net debt. But there will be other additional forces influencing the market value of the holding company that will not typically be included in the calculation of NAV. As mentioned, the highly uncertain value of its future business activities will influence its current share price. These growth plans may well involve raising additional debt or equity, so adding to or reducing the value of the holding company shares, both absolutely and relative to the current explicit NAV that includes only current net debt. Other forces that could add to or reduce the value of the holding company and so influence the discount or premium, not included in NAV, are any fees paid by subsidiary companies to the head office, in excess of the costs of delivering such services to them. They would detract from the value of the holding company when the subsidiary companies are being subsidised by head office. When fees are paid by the holding company to an independent and controlling management company, this would detract from its value from shareholders, as would any guarantees provided by the holding company to the creditors of a subsidiary company. The market value of the subsidiaries would rise, given such arrangements and that of the holding company fall, so adding to any revealed discount to NAV.

It should be appreciated that in the calculation of NAV, the value of the listed assets will move continuously with their market values, as will the share price of the holding company likely to rise or fall in the same direction as that of the listed subsidiaries when they count for a large share of all assets. But not all the components of NAV will vary continuously. The net debt will be fixed for a period of time, as might the directors’ valuations of the unlisted subsidiaries. Thus the calculated NAV will tend to lag behind the market as it moves generally higher or lower and the discount or premium to NAV will then decline or fall automatically in line with market related moves that have little to do with company specifics or the actions of management. In other words, the market moves and the discount or premium automatically follows.

If this updated discount or premium can be shown to revert over time to some predictable average (which may not be the case) then it may be useful to time entry into or out of the shares of the holding company. But the direction of causation is surely from the value attached to the holding company to the discount or premium – rather than the other way round. The task for management is to influence the value of the holding company not the discount or premium.

Yet any improved prospect of a partial liquidation of holding company assets, say through an unbundling, will add to the market value of the holding company and reduce the discount. After an unbundling the market value of the holding company will decline simultaneously and then, depending on the future prospects and expectations of holding company actions, including future unbundling decisions, a discount or premium to NAV may emerge. The performance of Remgro prior to and after the BTI unbundling conformed very well to this pattern. An improvement in the value of the holding company shares and a reduction in the discount to NAV on announcement of an unbundling – a sharp reduction in the value of the holding company after the unbundling and the resumption of a large discount when the reduced Remgro emerged. See figure 1 above.

The purpose of any closed end investment trust should be the same as that of any business and that is to add value for its shareholders by generating returns in excess of its risk adjusted cost of capital. That is to say, by providing returns that exceed required returns, for similarly risky assets. Risks are reduced for shareholders through diversification as the investment trust may do. But shareholders can hold a well diversified portfolio of listed assets without assistance from the managers of an investment trust. The special benefits an investment trust can therefore hope to offer its shareholders is through identifying and nurturing smaller companies, listed and unlisted, that through the involvement of the holding company become much more valuable companies. When the nurturing process is judged to be over and the listed subsidiary is fully capable of standing on its own feet, a revealed willingness to unbundle or dispose of such interest would add value to any successful holding company.

This means the holding company or trust will actively manage a somewhat concentrated portfolio, much more concentrated than that of the average unit trust. Such opportunities to concentrate the portfolio and stay active and involved with the management of subsidiary companies may only become available with the permanent capital provided to a closed end investment trust. The successful holding company may best be regarded and behave as a listed private equity fund. True value adding active investment programmes require patience and the ability to stay invested in and involved in a subsidiary company for the long run. Unit trusts or exchange traded funds do not lend themselves to active investment or a long run buy and hold and actively managed strategy of the kind recommended by Warren Buffett. A focus on discounts to estimates of NAV, to make the case for the liquidation of the company for a short term gain, rather than a focus on the hopefully rising value of the shares in the holding company over the long term, may well confuse the investment and business case for the holding company, as it would for any private equity fund. The success of Remgro over the long run helps make the case for investment trusts as an investment vehicle. So too for Brait and PSG, which are perhaps best understood as listed private equity and highly suited to be part of a portfolio for the long run.




A little light algebra and calculus can help clarify the issues and identify the forces driving a discount or premium to NAV


Let us therefore define the discount as follows, treating the discount as a positive number and percentage. Any premium should MV>NAV would show up as a negative number.


Disc % = (NAV-MV)/NAV ………………………………………..           1

Where NAV is Net Asset Value (sum of parts), MV is market value of listed holding company

NAV = ML+MU-NDt …………………………………………….       (2)


Where NAV is defined as the sum of the maket value of the listed assets held by the holding company. MU is the assumed market value of the unlisted assets(shares in subsidiary companies) held by the holding company and NDt is the net debt held on the books of the holding company – that is debt less cash.

Note to valuation of unlisted subsidiaries MU;

MU may be based on an estimate of the directors or as inferred by an analyst using some valuation method- perhaps by multiplying forecast earnings by a multiple taken from some like listed company with a similar risk profile to the unlisted subsidiary. Clearly this estimate is subject to much more uncertainty than the ML that will be known with complete certainty at any point in time. Thus the greater the proportion of MU on the balance sheet the less confidence can be placed on any estimate of NAV.

The market value of the holding company may be regarded as



That is to say all the forces acting on NAV, plus the assumed value of head office fees and subsidies (HO)activity and of likely much greater importance the assessment markets of the net present value of additional investment and capital raising activity NPV. NPV or HO may be adding to or subtracting from the market value of the holding company MV.

A further force influencing the market value of the holding company would be any liability for capital gains taxes on any realisation of assets. Unbundling would no presumably attract any capital gains for the holding company. These tax considerations are not taken up here

IF we substitute equations 2 and 3 into equation one the forces common to 2 and 3 ML,MU,NDt cancel out and we can conveniently write the Discount as the ratio


Disc= – (H0+NPV)/(ML+MU-NDt ) ………………………………………..(4)


Clearly any change that reduces the numerator (top line) or increases the denominator (bottom line) of this ratio will reduce the discount. Thus an increase in the value attached to the Head Office or the value of future business will reduce the discount. ( These forces are preceded by a negative sign in the ratio) A large increase in the value attached to investment activity will also reduce the discount and might even turn the ratio into a negative value, that is a premium. Clearly should the market value of listed or unlisted assets rise or Net Debt decline (become less negative) the denominator would attain a larger absolute number, so reducing the discount. The implication of this ratio seems very obvious. If the management of a holding company wishes to add value for shareholders in ways that will reduce any discount to NAV or realise a premium then they would need to convince the market of their ability to find and execute more value adding positive NPV projects. Turning unlisted assets into more valuable listed assets would clearly serve this purpose


Some calculus might also help to illuminate the forces at work. Differentiating the expression would indicate clearly that the discount declines for increases in H0 or NPV


That is dDisc/dNPV or dDisc/dHO= -1/(ML+MU-NDt)

This result indicates that the larger the absolute size of the holding company the more difficult it will become to move the discount through changes in the business model


Differentiating for small changes in the variables in the denominator is a more complicated procedure but would yield the following result for dML or dMU or dDNt



For example dDisc/DML= -(H0+NPV)/(ML+MU-NDt)^2


Again it may be shown that the impact of any change in ML,MU or NDt will be influenced by the existing scale of the listed assets held ML. The larger the absolute size of ML the less sensitive the discount will be to any increase in ML. The same sensitivities would apply to changes in MU or NDT. This reaction function is illustrated below



Current Research

Money and economic activity

Friday, July 29th, 2011

In the eighties and early nineties a number of attempts were made to measure the relationship between various measures of money supply growth and the growth in GDP, GDE, Household Consumption Expenditure and Consumer Prices between 1966 and 1993. (82, 1984,1989, 1990b, 1993) This earlier work on the relationship between money economic activity and prices was concluded in 1990 with an attempt to separate monetary causes and effects. That is to estimate whether the money to expenditure and income link was stronger than the income to money link, given the accommodative nature of money supply responses. It was reported that the money to income link was stronger than the reverse income to money influence, using a vector auto-regression approach. (1990a). The purpose of this paper is to update this analysis to include the past twenty years of data to establish whether or not money still matters for the SA economy in the way it did.

The full paper is available here: Money supply and economic activity (2012)

Current Research

Study on Inflation and Inflation Expectations in South Africa

Friday, April 15th, 2011

Our study strongly supports the view that supply side shocks on inflation in SA are best ignored by monetary policy. The analysis infers that raising interest rates in the face of a supply side shock that pushes prices temporarily higher will reduce demand in the economy without affecting inflation in the short term or inflation expected in the short or longer term. We show very clearly that realised inflation has affected inflation expected to a modest degree in South Africa. But the reverse does not hold at all – inflation expected does not affect inflation. Thus in response to supply side shocks, especially those that emanate from net foreign capital flows and the exchange rate, a much better way should be sought to anchor longer term inflationary expectations in SA than raising short term interest rates. It would seem that raising interest rates to fight inflationary expectations or so called second round effects on inflation can impose large costs on the economy in the form of lost output to no useful purpose.

A preliminary draft of our study is available here: Study on Inflation and Inflation Expectations in South Africa

Other recent research on monetary policy:
Lessons from the Global Financial Crisis

The global forces that drive SA’s Financial markets from day to day – an analysis with the implications drawn for monetary policy

A full directory of my research on monetary policy is can be found here: Research Papers – Monetary and Financial Economics

Current Research

Lessons from the Global Financial Crisis

Monday, January 31st, 2011

The worldwide financial markets and the global economy have suffered from a financial crisis on a scale not experienced since the 1930s. But the crisis now appears to be over. Credit spreads have returned to more normal levels, activity in credit markets has recovered strongly, and the volatility of day-to-day movements in share prices has declined. Moreover, the recovery of the global economy, of which the U.S. is such an important part, now appears strong enough to suggest that the recession of 2008-9 may turn out to have been a mild one of short duration. The IMF is forecasting global growth of 4% in 2011 after recording a marginal decline of about 1% in 2009, and thus the global financial crisis does not appear to have led to an economic crisis.

Click to read the full article: Lessons from the Global Financial Crisis (Or Why Capital Structure Is Too Important to Be Left to Regulation)


Current Research

The global forces that drive SA’s Financial markets from day to day

Monday, January 31st, 2011
This study demonstrates with the aid of single equation regression analysis the role global capital markets play in determining the behaviour of the Johannesburg Stock Exchange(JSE ALSI) the Rand/ US dollar exchange rate (ZAR) and long term interest rates in South Africa on a daily basis represented by the All Bond Index (ALBI) or long term government bond yields represented by the R157. It will be shown that since 2005 the state of global equity markets, represented in the study bythe MSCI Emerging Market Index (EM) has had a very powerful influence on the JSE. The EM Index is shown to have had a less powerful yet statistically significant influence on the ZAR while it is also demonstrated and that conditions in global capital markets, and the ZAR have had some weak but statistically significant influence on the direction of long term interest rates in South Africa. It will be demonstrated that movements in  policy influenced short term interest rates, have had very little predictable influence on share prices, the ZAR or long termbond yields. The causes as well as the consequences of the ineffectiveness of policy determined interest rates for monetarypolicy are further analysed.