The World Inequality Lab's A Wealth Tax For South Africa - A Critique II - The Proposal

The first Brief considered the estimation of the distribution of wealth in South Africa as set out in a companion paper to the wealth tax proposal from the World Inequality Lab. This Brief considers the proposal itself.

We had committed ourselves to a wealth tax; but in five years I found it impossible to draft one which would yield enough revenue to be worth the administrative cost and political hassle.

- Denis Healey, Labour Chancellor of the Exchequer, United Kingdom, 1974-79

This is the second of a pair of Briefs updating the HSF’s work on wealth taxes.

The proposal[1]

The proposal is for a tax on the net wealth of individuals in the top 1% of the wealth distribution. All assets less debts are to be taxed. Three levels of taxation are considered as possibilities:

Table 1 – Marginal rates of wealth tax

Category

Wealth threshold

Low rate

Moderate rate

High rate

Top 1%

R 3 820 000

1%

3%

3%

Top 0.1%

R 30 350 000

2%

5%

7%

Top 0.01%

R 146 890 00

3%

7%

9%

Note: Adults are individuals age 20 or older.

Source: Chatterjee et al, 2021: Table 2

The proposal estimates, using a benchmark 30% evasion rate, that the low rate would yield 1.6% of GDP, the moderate rate would yield 2.8% of GDP, or R 134 billion, and the high rate 3.3% of GDP. A wealth tax could take the form of a capital levy (a tax imposed once only) or an annual wealth tax. The proposal raises the possibility of a capital levy at a high rate followed by an annual wealth tax at a lower rate.

This Brief will point out some of the difficulties with the proposal. But first a definition: a wealth tax is confiscatory if, given measured rates of capital appreciation over the long term, it results in a drop in the real value of a quantum of wealth held by an individual. If an individual’s quantum rises over time there is the possibility of a non-confiscatory wealth tax, and there will be a value of a wealth tax rate which represents the crossover from a non-confiscatory to a confiscatory level. It is worth knowing what this value is. It is also worth knowing whether a wealth tax is intended to be confiscatory or not, an issue not discussed in the proposal.

The difficulties

1. The wealth tax threshold is very low. The proposal invites us to consider the Colombian wealth tax, in a country with a similar GDP per capita in purchasing power parity dollars to South Africa. But the Colombian wealth tax is very unstable. It is on its sixth version of the tax since 2002, and the tax base has varied dramatically as have threshold levels of taxable wealth. Tax revenue has also varied greatly, between 0.1% and 0.6% of GDP, all much lower than the even the low tax rate proposed for South Africa.

More coherent proposals have emerged from Elizabeth Warren and Bernie Sanders, and these are presented in Table 2. Both are considered radical in the United States.

Table 2

Elizabeth Warren

Bernie Sanders

South African equivalent

Wealth threshold
(households)

Marginal tax rate

Wealth threshold
(married couples)

Marginal tax rate

Wealth threshold
(individuals) (Rand)

Marginal tax rate
(Sanders)

   

$ 32 million

1%

21,400,425
(top 0.15%)

1%

$ 50 million
(top 0.1%)

2%

$ 50 million

2%

33,438,164
(top 0.067%)

2%

   

$ 250 million

3%

167,190,822
(top 0.003%)

3%

   

$ 500 million

4%

   

$ 1 billion

6%

$ 1 billion

5%

   
   

$ 2.5 billion

6%

   
   

$ 5 billion

7%

   
   

$ 10 billion

8%

   

It would not be reasonable to simply translate the US thresholds into Rand. Allowance must be made for the fact that the US is a much richer country. Also the Warren and Sanders proposals are for households and married couples. Sanders cuts his thresholds in half for individuals, and the same is done for South African equivalents. The SA equivalent thresholds bear the same ratio to the SA national income per capita as the Sanders thresholds to US national income per capita. Doing that raises the initial threshold to R 21.4 million, a much more reasonable figure than the R 3.8 million in the proposal. There is no point in going beyond the third level in South Africa – the top category consists of only 1 155 people.

2. Non-confiscatory wealth tax rates are low across all asset classes, given capital appreciation over the last ten years. Annexure 1 sets out the top nominal annual rates of capital appreciation necessary for wealth taxes not to be confiscatory. Assuming consumer price inflation at the midpoint of the Reserve Bank target range, they vary from 5.56% for a wealth tax of 1% to 14.84% for a wealth tax of 9%, the top rate considered in the proposal. Note that capital appreciation is, in general different from the total return to wealth.

In fact,

Total return from wealth = Capital appreciation + plus income from wealth.

Income from wealth is already taxed: interest and dividends from property income form part of taxable income. Dividends from other shares in domestic companies are currently subject to a withholding tax of 20%, and company profits from which dividends are paid are subject to corporate tax of 28%. Dividends from foreign companies are included in a person’s gross income but may qualify for a full or partial exemption from normal tax, such that the maximum tax rate is 20%.

Now consider non-confiscatory wealth tax rates by asset class (equities, bonds, housing, pensions, life insurance, business assets and currency) given capital appreciation over the ten years from 1 January 2011 to 31 December 2020.

Annexure 2 Table 1 sets out the consumer price index in each December from 2010 to 2020. The average annual inflation rate over the ten year period was 5.07%.

Equities

Consider a hypothetical individual who invested a rand amount equal to the FTSE-JSE all share index on 31 December 2010 into a fund which perfectly tracked the index, and who sold the holding exactly ten years later. Annexure 2 Table 2 shows what would have happened to the capital over the period. After capital gains tax[2] had been paid, the average annual nominal capital appreciation rate would have been 5.43%, representing a real return of 0.36%. Now suppose a wealth tax is imposed, payable at the end of each year and financed by an appropriate sale from the investment. Capital gains tax would be payable on each sale (or a credit if the return in a particular year is negative) and on the final sale. The break-even annual wealth tax rate would have been 0.46%. In that case the real capital appreciation would have been zero. It is probable that the rich could have done better than that, but passive investing has become popular precisely because active investing yields returns that are not much higher, while being riskier.

Bonds

Consider a hypothetical individual who invested a rand amount on 31 December 2010 into a ten year bond sufficient to yield a capital sum of R 10 000 at maturity. There are two cases to consider: a fixed interest rate bond and an inflation linked bond. In the former case, coupon payments are taxed as income and in the latter case coupon payments plus the inflation increment in the face value of the bond is taxed as income. Annexure 3 Tables 3A and 3B set out the approximate situation over the ten year period. In both cases, the real appreciation rate on the bond capital was less than zero, so any wealth tax would have been confiscatory. Real appreciation on corporate bond capital will also be less than zero.

Houses

Annexure 2 Table 4 shows that average house prices have increased just short of inflation between December 2010 and 2020, so that the real capital appreciation was less than zero, making any wealth tax confiscatory.

Pensions

South Africa pensions are taxed on an exempt-exempt-tax basis. The first exemption takes the form of a deduction from taxable income, limited to the lower of R350 000 or 27.5% of taxable income before the inclusion of a taxable capital gain. The second exemption is from investment returns over the pre-retirement period. After retirement, income taken from the fund is fully taxed, whether as withdrawals from a living annuity or an annuity income from a life annuity.

Survey after survey has shown that South Africans do not save adequately for retirement. OECD pension statistics indicate that South Africa has by far the lowest net replacement rate[3] of all 37 countries at 18.5%[4]. Under these circumstances, it would be folly to tax pension wealth.

Life insurance

The method for assessing the net present value of contributions to life insurance policies compared with the net present value of payouts is described in Annexure 3. For the example chosen, the latter is lower than the former, so any wealth tax on interests in life assurance would be confiscatory.

Business assets

Data on the financial assets of unincorporated business are not available, but they (a) are likely to be held for precautionary purposes to cover working capital needs and (b) are a small proportion of the wealth of the richest people. The contribution of any wealth tax on them would be at most very small.

Currency

Except in circumstances where the consumer price index is falling, currency is a depreciating asset in real terms, so any tax on it would be confiscatory.

3. The proposal is insensitive to the life cycle. The proposal recognizes that some taxpayers may be asset-rich, but income-poor. It sees this as a temporary liquidity problem, and do not relate it to the life cycle, even though Figure 4 of the companion paper shows a standard life cycle pattern in age-specific wealth. A life cycle view predicts that the wealth of an individual peaks at the age of retirement and then declines as wealth is liquidated to fund consumption at a time when there are no longer earnings to do so. A wealth tax would hit retired persons particularly hard, the more so since pension replacement rates are so low. It would also hit the ability of individuals in the pre-retirement phase to accumulate assets for old age, especially in a low growth environment.

Conclusion

The rich will resist a wealth tax, and they will fight like demons if it is set at a confiscatory rate.

The only asset for which a non-confiscatory wealth tax can be imposed is equity. It has been shown that the maximum non-confiscatory wealth tax for an investment in a JSE all share tracker fund would have been 0.36% in the decade up to December 2010, assuming no transaction fees. It was also observed that the rich would have probably been able to get a better return than obtainable on the JSE. Assume that the margin to be 1% for the bottom wealth tax bracket, 2% for the middle bracket and 3% for the top bracket. Using data from Table 1 and interpolating within and extrapolating from Table 10 in the companion paper, one can estimate the share of equity in total wealth for each bracket (as defined for South Africa in Table 1 above). Table 3 sets out the results:

Table 3

Bracket

Individuals

Tax on equity

Share of equity in wealth

Incremental wealth tax rate

Collections[5]

(R billion)

Bottom

62 494

1.41%

24.7%

0.35%

4.7

Middle

22 697

2.35%

42.3%

0.99%

9.7

Top

1 155

3.27%

44.2%

1.45%

5.1

All

86 346

     

19.5

The differences between the estimate in Table 3 and the proposal are:

(a) the wealth threshold is much higher in Table 3. The proposal would weigh heavily on the upper middle class as well as the seriously rich.

(b) the analysis here indicates that only equities could bear a non-confiscatory wealth tax. Tax on any other category would be confiscatory. This incidentally has been the conclusion that the Colombians have come to. In 2019, their wealth tax was reformed as an equity tax, with 1% rate on net equity over COP5bn ($1.67 million or R 25 million).

Against the collections would have to be set the administrative costs for the state and, from a social point of view, the costs on all other agents who would have to co-operate, not to mention legal battles. It would be much more sensible to raise the revenue net of the costs of collection by adjustment of existing tax rates. Denis Healey was right.

Charles Simkins
Head of Research
charles@hsf.org.za

Annexure 1 – Calculation of annual break even nominal rates of capital appreciation

Denote the rate of nominal capital appreciation by r, the wealth tax rate by t, and the rate of inflation by c. Then, in order for real wealth to be constant between this year and the next, each rand of wealth this year must be worth 1+c rand next year. If the rate of capital appreciation is r, the value of wealth next year is 1+r before tax and (1+r)(1-t) after tax. So the breakeven rate of nominal capital appreciation can be found from the equation:

(1+r)(1-t) = 1+c

Making r the subject of the equation,

r = (c+t)/(1-t).

Assume that c = 4.5%, the midpoint of the long-standing inflation target given to the Reserve Bank. Then r can be calculated for varying levels of the wealth taxation rate.

Wealth taxation rate

Break even nominal capital appreciation rate

1%

5.56%

2%

6.63%

3%

7.73%

5%

10.00%

7%

12.37%

9%

14.84%

 

Annexure 2 – Non-confiscatory wealth taxes by asset class: tables

Table 1 - Consumer price index

     

December

   
     

2010

70.7

 

2011

74.2

 

2012

78.4

 

2013

82.9

 

2014

88.0

 

2015

92.0

 

2016

97.8

 

2017

103.0

Average

2018

107.8

annual

2019

112.2

rate

2020

115.9

5.07%

Source: Statistics South Africa, Consumer Price Index, Statistical Release P0141, December 2020, CPI History Table

Table 2

     

FT-JSE index

     
 

Without

With wealth tax

31-Dec

wealth tax

Before

After

       

2010

32119

32119

 

2011

31986

31986

31839

2012

39250

39069

38890

2013

46256

45831

45621

2014

49771

49087

48862

2015

50694

49768

49539

2016

50653

49499

49271

2017

59505

57881

57615

2018

52737

51062

50827

2019

57084

55017

54764

2020

59409

56994

56732

       

Capital gains tax

4912

4100

 

Wealth tax

0.00%

0.46%

 

Proceeds

54497

52632

 

Nominal return

5.43%

5.06%

 

Real return

0.36%

0.00%

 

Source: Wall Street Journal, Market Data

Table 3A

       

Government Bonds

       
 

Vanilla

 

Coupon

6.75%

     

Yield

8.74%

     
 

Market price

Face value

Coupon

Cash flow

 

Ex-dividend

     
         

2010

8706

10000

 

-8706

2011

8792

10000

675

675

2012

8886

10000

675

675

2013

8988

10000

675

675

2014

9098

10000

675

675

2015

9219

10000

675

675

2016

9350

10000

675

675

2017

9493

10000

675

675

2018

9648

10000

675

675

2019

9816

10000

675

675

2020

10000

10000

675

10675

         

CGT

233

     

Proceeds

9767

 

6750

 

Nominal appreciation rate

1.16%

     

Real appreciation rate

-3.91%

     

Final value

10000

     

 

Table 3B

           

Government Bonds

           
 

Inflation-linked

         

Coupon

1.75%

         

Yield

3.50%

         
 

Market price

Face value

Coupon

Face value

Taxable

Cash flow

 

Ex-dividend

   

increment

income

 
             

2010

4400

6100

     

-4400

2011

4673

6402

112

302

414

112

2012

5085

6764

118

362

481

118

2013

5533

7153

125

388

513

125

2014

6020

7593

133

440

573

133

2015

6552

7938

139

345

484

139

2016

7129

8438

148

500

648

148

2017

7757

8887

156

449

604

156

2018

8442

9301

163

414

577

163

2019

9187

9681

169

380

549

169

2020

10000

10000

175

319

494

10175

             

CGT

306

         

Proceeds

9694

 

1438

3900

5338

 

Nominal appreciation rate

1.56%

         

Real appreciation rate

-3.50%

         

Final value

10000

         

 

Table 4

 

House prices

 
   

31-Dec

Index

2010

332.4

2011

346.4

2012

366.4

2013

395.2

2014

420.1

2015

446.6

2016

467.4

2017

484.8

2018

504.2

2019

519.8

2020

533.3

Nominal capital appreciation

4.84%

Real capital appreciation

-0.23%

Source: First National Bank, House price index

Annexure 3 – Net present value of contributions to, and benefits from, life insurance

1. Analysing life insurance requires a life table. The starting point is the abridged life table for all South Africa for 2015-2020. Non-smoking holders of life insurance will have lighter mortality than the population as a whole. Brass showed that a family of life tables can be generated from a baseline life table using two parameters, α and β[6]. An α of zero and a β of one will reproduce the baseline table. The revised life table kept β at 1, while setting α at -0.314. This setting means that life expectancy among non-smoking policy holders at age 30 is 42.9, five years more than for the population as a whole. The survival curves are shown in Figure 1.

2. Life insurance premiums for cover of a million rand cost about R 475 per month, or R 5 700 per year[7] for a non-smoker starting a policy at age 30. To calculate the present value of a life time of contributions in prices at the start of the policy, one needs the following assumptions:

Inflation rate: 4.5%

Nominal premium escalation rate: 2.25%, half the inflation rate

For simplicity, it is assumed that premiums are paid annually in the middle of the year.

3. Life insurance premiums are paid into a fund from which payouts are made. The number of deaths per year are given by the life table and, for simplicity, they are assumed to be concentrated in the middle of the year. It is assumed that the fund earns a nominal rate of return of 7% net of all expenses.

4. Before the first premium is paid, the fund has a value of zero, and it returns to zero after everyone has died.

5. Under these circumstances, the value of payouts is 89% of the value of contributions in constant prices . The non-confiscatory wealth tax of interests in life insurance is therefore zero.


[1]AroopChatterjee, Leo Czajka and Amory Gethin, A wealth tax for South Africa, World Inequality Lab Working Paper No 2021/02, January 2021, available at file:///C:/economy/WorldInequalityLab_WP2021_02_SouthAfrica_WealthTax_2b.pdf

[2]All capital gains tax is calculated on the basis of a 45% marginal tax rate, and a 40% inclusion rate, with no exempt amount. This reflects the top personal income tax bracket and the rate at which capital gains are included in income in the 2020/21 tax year.

[3]The net replacement rate is defined as the individual net pension entitlement divided by net pre-retirement earnings, taking into account personal income taxes and social security contributions paid by workers and pensioners. It measures how effectively a pension system provides a retirement income to replace earnings, the main source of income before retirement. This indicator is measured in percentage of pre-retirement earnings.

[4] OECD, Net pension replacement rates (indicator). doi: 10.1787/4b03f028-en (Accessed on 20 February 2021)

[5]The evasion rate assumed here is 20% for the bottom bracket, 25% rising to 35% in the middle bracket, and 40% in the top bracket, giving an average evasions rate of 33.2%, close to the proposal’s assumption of 30%.

[6] W Brass, On the scale of mortality. In W. Brass (ed.). Biological Aspects of Demography.Taylor and Francis, 1971.

[7]One Life, Truth About Money,