Chapter 2    THE ECONOMIC EFFECT OF A LAND TAX ON AGRICULTURE

2.1 Introduction

Measuring the economic effects of a policy intervention such as the levying of a land tax presents a measurement problem of the kind that economists conventionally model with the use of mathematical programming algorithms. While these models provide powerful and rigorous evidence of these kinds of policy effects, they also suffer from a number of weaknesses. First, the models are as contemporary as the data that are used to portray the real world. In this respect, the available data bases for the rural areas in South African are less than reliable. For this reason, for example, the commercial farming sector is modelled here, although one case study from the former homeland areas has been included. Second, while the development of computer technology allows models of a complexity that were impossible only a few years ago, the real world cannot be duplicated, only modelled. For this reason, the models presented here focus largely on farm production, and less on the off-farm rural economy. Third, while the strength of such models is derived from their mathematical rigour, their resolution depends largely on a set of somewhat restrictive assumptions about the way in which markets, and biological processes, work. In this respect, dynamic modelling is introduced in some of the cases presented below in an attempt to make the exercise more realistic. Notwithstanding these caveats, the purpose of this Chapter is to report on the results of an analysis of the effects of a land tax. The focus will be on the effect of the following variables:

• Different levels of a land tax;
• The cost of the land tax to the owner of the land;
• Different tax bases (i.e. market value versus productive value); and
• Different ways in which the land tax is implemented (e.g. deductibility from income tax).

The purpose of this exercise is to present information to policy makers on the optimal way in which a land tax could be implemented.

2.2 Methodology and data used

Various tools and techniques exist to analyse the efficient use and allocation of scarce resources within the farm business. One of the most widely used of these, linear programming, is a technique for solving problems that share the characteristic of an objective function or goal that has to be maximised or minimised given certain resource limitations as well as numerous alternative means of using the resources. The typical product of a linear programming analysis of a farm situation, for example, is a plan of how much of a wide variety of commodities should be produced, and the quantities of different inputs to use in producing those commodities, in order to maximise profits.

Furthermore, the technique also provides information on the true or economic value of the resources used in the optimal solution. Thus, a computational by-product of the programming procedure is information concerning those resources that limit the income potential of the farm operation, which resources are in excess, and the cost of acquiring additional units of the limiting resources. For example, a typical linear programming model would indicate how much land is used and how much is unused in a whole-farm planning problem, and if land is limiting the potential for growing additional crops and increasing income, how much the operator could pay for an additional hectare of lands. The value of land created in this way is also called the shadow price, which represents the income foregone if a hectare of land is lost or withdrawn from production.

A further attribute of linear programming is that it can easily be used to evaluate the effect of a change in product prices or in technical efficiency, i.e. it can be used to measure the sensitivity or stability of the farm plan to external changes. For example, the linear programming procedure indicates how the whole-farm organisation will change if the overhead cost (e.g. a change in the land tax rate) of the farm increases or decreases. Other questions of the "what would happen if…?" variety, such as what would happen to income if a land tax is raised at 2 per cent on the market values of land, can also easily be evaluated.

Linear programming handles opportunity cost with ease. Opportunity cost reflects the income foregone in using a resource in an alternative enterprise. The process of pricing resources in the production of various products based on the income generation capacity or opportunity cost of that resource in alternative uses, is the heart of the programming procedure. This characteristic will be very useful to calculate the productive value of the land. The effect of a land tax raised at different levels on the productive value of the business can easily be determined by simply comparing the results before the introduction of the land tax with that after the introduction of the tax.

Linear programming can be done for a single year, in which case it would be a static model, or it can be done for a series of consecutive years (dynamic approach). Both these methods were used in the current investigation.

Dynamic linear programming

The time-dependant processes of a biological system, such as is characteristic of agricultural production, can typically be modelled using dynamic algorithms. As biological processes vary over time, it is difficult to achieve a realistic description without reference to the time dimension. The effect of short and long term liabilities can also be analysed when the time factor is brought into account.

In dynamic linear programming, plans are formulated for distinctive and consecutive production periods within the planning horizon. The periods are mutually combined and conjoined by transfers. This means that activities and resources for each year under consideration have to be defined, which places large demands on the data requirement and computational capacity. The activities, which are executed in each year, are constrained by the resources available in that specific year.

The model assumes management is of high quality and that the manager will react immediately to counter any activity that will cause the objective function to decrease. In reality this might not be the case. Usually even better than average farmers do not always react in the manner and as fast as the model assumes. The effect on a farming unit will, therefore, be larger than implied by the model. This is magnified because of the long production periods in agriculture, i.e. a farmer has to wait for the current crop to be harvested and sold before he/she can alter the cultivation activities. In the application reported below, the optimum farm plan without a land tax is used as a base scenario from which the effect of a land tax is determined.

Different land tax rates are introduced in the model. The different tax rates are raised on both conservative and optimistic market values, and on agricultural use values of the land under question. For the Southern Cape study the conservative market values were taken as being one standard deviation or 17 per cent lower than the going rate and the optimistic market value as 17 per cent higher than the going rate. With reference to the Olifants river study the conservative market values were taken as 20 per cent lower than the going rate and the optimistic market value as 20 per cent higher. The two market values were introduced in the model to capture the effect of possible subjectivity on the part of land assessors. The shadow value of land as obtained by the model was used to reflect agricultural use values.

Due to the implicit reference to the time factor that characterises the dynamic method, it was necessary to use a discount rate to change future values to present values. For this purposes an 8 per cent discount rate was used. The rate is based on a long term average interest rate of 17 per cent and an expected inflation rate of roughly 10 per cent. According to this the real rate amounts to 6,3 per cent. An increment of 1,7 per cent was added to incorporate risk, which brings the total rate to 8 per cent.

Static linear programming

As noted above, such dynamic models require relatively large time and computational resources. For this reason, the rest of the study areas were modelled using static deterministic linear programming models. Comparison with the results of the dynamic models, however, enables a rough consideration of the effect of the time factor. Static models work on the same principles as the dynamic models, except that the time factor is not incorporated.

2.3 Measuring the effect of a land tax

To calculate the cost of a land tax to the owner of land, both the direct and indirect costs had to be taken into account. Non-linear models were used to estimate the indirect effects, whilst linear programming models were used to calculate the direct effects. The effect of a land tax on land prices was estimated by a structural model of land prices that provides a framework for analysing the relative importance of factors determining farmland prices. The model has previously been formulated for different agro-economic regions and for South Africa as a whole. The possible effect of a land tax was measured by imposing taxes on land market prices at different rates in a model whose objective function is to maximise farm profits, measured by net farm income (NFI). A land tax rate of respectively 1, 2, 3, 4, 5 and 8 per cent was imposed on average land prices for South Africa as a whole as well as for Gauteng province, as if the tax had been in effect since 1970. Gauteng was selected as farmers in that province are expected to capitalise more non-farm factors into land values because of the larger number of alternative uses available. The gap between the market and agricultural value of land is therefore expected to be larger. New entrants in the market will thus find it difficult to gain entry unless they have sufficient equity to overcome the gap between the market and agricultural value of land. This problem will assume even larger proportions if a land tax is imposed in the province, as, theoretically, the tax should have a relatively smaller effect on market values than on net farm incomes, because of the high component of non-farm factors in expected land prices in Gauteng.

2.4 Data used for the analysis

Representative farms were modelled in a number of regions of South Africa. These differ from the areas identified in the original project proposal, partly to ensure that the areas represent as wide a variety of farming conditions in South Africa as possible, and partly because of data problems. The following case studies were included:

• district, Olifants river basin, Beaufort West district;
• Free State: Bloemfontein magisterial district;
• KwaZulu-Natal: Cedara study group information;
• Mpumalanga: Nelspruit magisterial district, former Gazankulu area (case study of communal areas); and
• North-West: Potchefstroom magisterial district.

2.5 The effect on market values of land in Gauteng and South Africa

This section compares the results of the mathematical programming analysis of areas where average market values are hardly distorted by non-agricultural factors with areas where considerable distortion is found. This is followed by an analysis of the results of the dynamic linear programming in order to explain the longer-term effects of the tax, and then by an analysis and summary of the results of the static linear programming approach.

One explanation for opposition to an agricultural land tax by farmers is that current landowners believe that the cost to them may end up being considerably more than the cost of tax alone. Farmers will not only have to pay the tax, but may also incur a loss in land values, since the prospect of future tax payments depresses current land prices. In this section an attempt is made to quantify the extent of this capital loss. In the model, a land tax at different rates is raised on the market value of land. The area covered includes South Africa as a whole to determine what the average effect of a land tax will be, and Gauteng to determine the effect in an area where market values are distorted by non-farm factors.

The results are shown in Table 2.1 below. According to Table 2.1 a land tax of 1 per cent, if an additional tax, will have an average negative effect of 5,3 per cent on predicted real land prices in South Africa as a whole. It should be noted that these results represent a worst case scenario, as it is assumed that landowners cannot shift the tax burden to consumers or tenants through higher prices or rentals, etc. However, it is entirely possible for the latter to occur, namely that the burden of the tax may be shifted down to tenants by the landowners.

It is often argued that the tax-induced decline in land prices encourages access to land for new farmers such as participants in a land reform programme, young people or tenants wishing to own land. However, if the tax is fully capitalised into reduced land values, this implies that land values fall by the same proportion as the fall in profits. If the ratio of after tax profits to land values remains the same, there is no additional incentive to buy land. Table 2.1 shows further that a 2 per cent tax rate, if an additional tax, would have resulted in a decline of 12,32 per cent in land prices in 1991. Thus the average real value of a hectare of farm land in South Africa would have declined from R297,75 to R218,40. The land tax payable on R218,40 per hectare at 2 per cent would have been R4,37 per hectare, which would have resulted in a decline of 11,28 per cent in net returns, i.e. from R38,72 to R34,35.

This differential in the reduction of land values and net returns is mainly caused by variable weather conditions. Therefore, these results support the argument that although land will be cheaper, farmers will not necessarily be better off with the tax, and the land market will not favour new entrants. Another interesting observation is shown in the final column, which shows that land must be taxed at more than 8 per cent before its value falls to zero, in contrast with the theoretical argument of a 4 – 5 per cent rate. The difference in these two rates can be explained by the reaction of farmers, who discount the negative effect of a land tax into lower land market values, resulting in a declining base from which the tax is calculated.

A 1 per cent land tax rate in Gauteng, if an additional tax, will have an average negative effect of 2,5 per cent on predicted real land prices, which is about half the extent of the effect for the country as a whole. This result is to be expected, as landowners there will discount the probability of their land becoming rezoned for urban use, and therefore of much higher value. Furthermore, the demand for agricultural land near urban areas is higher because of its potential recreational value. Thus, even if it remains zoned as farmland, it has a higher value.

Average returns per hectare in Gauteng might also be higher because of a higher degree of intensification, which is a direct result of the proximity to the market. A 2 per cent land tax rate on land in Gauteng, if an additional tax, would have resulted in a decline of 2,88 per cent in land prices in 1991, implying that the average real land value would have fallen from R1 526,77 to R1 482,44. At a tax rate of 2 per cent, this would have amounted to R29,65 per hectare, or a 20,38 per cent decline in net returns, i.e. from R145,43 to R115,78. This higher than expected fall in profits is again the result of the capitalisation of non-farm factors in land values. The result is that farmers who are dependent on the land are, therefore, in a highly vulnerable position in such areas.

Table 2.1: The effect of a land tax on real land prices in Gauteng and South Africa (%)

It is often argued that the prospects of future income or commodity taxes will have the same effect on land prices. However, Table 2.1 shows that even at a rate of 0,5 per cent, if the land tax is substituted for income tax, average farmland prices in South Africa will decline. The land tax will exceed income tax in most years. The reasons for this might be:

(i) a lower rate of return on South African farmland;

(ii) widespread evasion of income tax in South Africa; and/or

(iii) a low average income tax rate applicable to farmers.

2.6 The results of the Heidelberg case study

The underlying reason for the use of the dynamic linear programming model is to illustrate the long-term effects of a land tax. Tables 2.2 and 2.3 show the direct and indirect effects of a land tax in the Heidelberg district of the Western Cape province. In all the scenarios the model has chosen only wheat and barley as cash crops as a result of the limited alternatives available to farmers. When a land tax is introduced on the shadow value of the land, no changes occur to the farming structure. However, the net present value of the income stream for the six-year period declines at 1,1; 1,7 and 2,3 per cent when a land tax of 1; 1,5 and 2 per cent respectively is raised. Labour hours remain constant because of the static cost structure. The largest impact of the tax is on the liabilities of the farmer. Farmers will require more debt to operate their farms at an optimal level.

Table 2.2: Effects of different land tax rates if levied on shadow prices of land (R/ha)

Notes: LSU = large stock unit
NPV = net present value

Table 2.3: Effects of different land tax rates if levied on market values of land (R/ha)1

Notes: 1Base year values the same as in Table 2.2
LSU = large stock unit
NPV = net present value

Modelling for conservative and optimistic market values is for purposes of accommodating valuers’ possible bias in a comparable sales method environment.

For the first year short-term liabilities increase by 3,7 per cent when a tax of 1 per cent is raised, and by 7,4 per cent when a 2 per cent tax rate is levied. In year 3 the increase in short-term debt, when compared to the base, rises to 131,5 per cent when a tax of 1 per cent is raised, and 262,9 per cent when a tax of 2 per cent is raised. Here it is important to note that R50 000 of short-term credit is used in the base year. An increase of 131,5 per cent thus implies an amount of about R119 000 credit needed to operate the farm at an optimum level. The situation is worse when the land tax is levied on market values. The year 1 short-term liabilities increase by 9,3 per cent when a 1 per cent land tax is levied, and by 18,9 per cent with a 2 per cent land tax rate. In year 3 short-term liabilities increase by 328,7 per cent and 652,2 per cent respectively with a 1 per cent and 2 per cent land tax rate. When land tax is levied at a 2 per cent rate, financial assistance is also needed in the fifth year. Farmers exposed to this situation may incur liquidity problems and eventually face insolvency. Tables 2.2 and 2.3 clearly show that the impact of a land tax raised on shadow values differs considerably from the scenario where it is raised on market values.

2.7 The Olifants river basin results

While models have been constructed for all the case studies under consideration, only the case study of the Olifants river basin in the Western Cape province is discussed in full because the results of the other case studies show the same trends. A comparison between the different areas is, however, made, and this presents significant insight on the effect of different tax rates and tax bases. Summary results of this case study are given in Table 2.4, while complete results for the separate areas can be found in the original research report.

Table 2.4 shows that when a non-deductible land tax of 0,5; 1; 1,5 and 2 per cent is levied on the shadow prices of land, the land tax per hectare amounts to R33, R65, R98 and R130 per hectare respectively. The objective function (profit) falls by 0,59; 1,19; 1,80 and 2,42 per cent respectively.

When a land tax is raised on the market value of land the taxed amount per hectare increases sharply as the tax rate increases. The land tax per hectare at the various rates is R200, R400, R600 and R800 per hectare respectively, resulting in a proportionate change in the net farm income of 3,77; 7,84; 12,23 and 17,00 per cent respectively. When the tax is made deductible from income tax (i.e. at a marginal rate of 43 per cent), the situation alters significantly. Using the productive value as base, and rates of 0,5; 1,0; 1,5 and 2,00 per cent, the amount of tax payable reduces to R19, R37, R56 and R74 per hectare, accounting for 0,34; 0,68; 1,02 and 1,36 per cent of the respective net farm income values per hectare. When market values are used as base, the magnitude of these changes is much higher.

When the land tax is levied on market values, the effect on net farm income is much larger than the tax rate, for example when a non-deductible rate of 2 per cent is levied on market values, the land tax consumes 17 per cent of the net farm income per hectare. Therefore, a land tax raised on market values does not account for the ability to pay principle, unlike a tax on the shadow value of land. The small differences between the effect of the latter on the net farm income and the tax rate can be attributed to data problems.

Table 2.4: Summary of the Olifants river case study

Note: A marginal tax rate of 43 per cent was used for illustrative purposes.

In theory the effect on net farm income should equal the tax rate, since the shadow price represents the opportunity cost of the land, which should be equal to its net income raising capability. It should be noted, however, that the productive value of land can be measured in many different ways. In this study the shadow price of land was used, which represents the opportunity cost of land. Productive values calculated in the conventional way (i.e. net farm income / discount rate) will yield higher effects. In this case the land tax rate should be divided by the same denominator to establish the effect of the land tax on the net farm income.

It is also clear that these impacts tell only a part of the story. The forward and backward linkages between agriculture and the rest of the economy also have to be accounted for in assessing the total impact of the imposition of a land tax. These multiplier effects include the effect of reduced income for farmers and farm workers on important variables such as total value added in the region and job creation. The multiplier effects of a land tax in terms of the different land tax regimes are shown in Table 2.5 below. For example, when a tax of 2 per cent is levied on the productive value of land in the Olifants river basin, GDP will decline by R3,7 million at a loss of 212 job opportunities, and a loss of R0.8m in revenue from income tax foregone, as shown in the second last row. Alternatively, if a tax of the same magnitude is levied on the market value of land, its GDP contribution will decline by R22,7 million, government revenue will decrease by R4,7 million, and 1 305 jobs will be lost in exchange for a revenue of R12,6 million from the land tax.

It is however important to remember that the money removed from agriculture through the tax will be re-invested into the community by the local government, with the result that these figures represent the gross losses. In addition, local authority spending may have a redistributive effect that is beneficial to the community as a whole. On the other hand, the multipliers for agriculture are larger than for the other sectors in the economy.

Table 2.5: Effect of different land tax rates and bases on the Western Cape economy

2.8 Summary of the results from the other case studies

In this section the results of all the areas modelled are compared to ascertain the extent to which the patterns described above are repeated. The results are compared with regard to the effect of a land tax on the net farm income of different areas. This is followed by a brief comparison between land taxes and RSC levies.

Table 2.6 presents the effect of different land tax rates on the net farm income value when the tax is raised on market values. From the table it can be seen that the effect is dependent on the scenario followed. When a non-deductible land tax of 2 per cent is introduced, the effect differs significantly between the areas. In the Great Karoo, for instance, such a tax absorbs 7,80 per cent of the net farm income per hectare, while in Bloemfontein it is a mere 1,85 per cent. This may be because average market values per hectare in Bloemfontein are relatively low in comparison. In addition, the area is characterised by a wide variety of land uses, ranging from extensive grazing to irrigated production. Thus, average values are skewed in this case.

Nonetheless, when a land tax is raised on average market values the effect on net farm income differs considerably across different regions, negating the equity principle. It is also very difficult to quantify what the effect will be on the net farm income because one will first have to identify the non-farm factors that are capitalised in the market values. Raising a land tax on market values can therefore be seen as distortion, as the imposition of the tax makes certain areas more profitable relative to others.

When a land tax is levied on the shadow prices as tax base, the effect on the net farm income of different regions is more neutral, even though the actual tax per hectare paid differs considerably. Table 2.7 shows the results. When a non-deductible land tax is levied at a 2 per cent rate on the shadow price of land, the effect on net farm income falls in the narrow range of 1,84 per cent to 2,36 per cent. When the tax is made deductible at the marginal income tax rate of 43 per cent, the range narrows even further, to between 1,05 per cent and 1,35 per cent. Thus, the effect of the tax is more predictable, and more equitable.

If a land tax is raised as a provisional tax, i.e. if it is made fully deductible from income tax payable, it will not have any effect on the shadow price of land or on the income potential of the land. However, it might still have a negative effect on the short-term availability of credit, as the owner has to pay the land tax before it can be claimed against income tax payable.

Table 2.6: The effect of a land tax on net farm income per hectare: market valuation

Table 2.7: The effect of a land tax on net farm income per hectare: shadow values

2.9 Land taxes and RSC levies

Due to time constraints it was not possible to compare the effect of a land tax and of RSC levies for the whole of the South African agricultural sector. However, a case study was done of the Olifants river area that compares the potential revenue from these two sources. RSC levies are levied as a percentage of the turnover of a farm as well as a percentage of the wages paid out. According to the District Council, the establishment (turnover) levy amounts to 0.15504 per cent of the turnover (including VAT) in this area, whilst the service levy, which is charged on gross salaries, wages and drawings, amounts to 0,3876 per cent (including VAT) of these components. Administration costs account for 20 per cent of the revenue generated through these levies. Table 2.8 compares the expected revenue from these taxes with those from a non-deductible land tax. However, the amounts in Table 2.8 have been adjusted to exclude VAT.

Table 2.8 : RSC levies compared with different land tax regimes

It is clear from Table 2.8 that if a land tax is levied on market values, the cost of the land tax to the landowner will far exceed that of RSC levies. However, when the land tax is raised on the shadow prices of the land the amount of the revenue is more comparable. As these RSC levies are deductible from income tax, the actual cost of the RSC levies to the landowner will be much smaller. However, it is clear that the revenue from RSC levies will be larger than the revenue from a land tax if the rate for the latter is closer to 0,5 per cent, both are deductible, and the land tax is raised on productive value. In all other cases the revenue from land taxes will be larger.

2.10 Conclusion

This chapter has shown that if a land tax is raised the following should be considered:

• The land tax base;
• The land tax rate; and
• The deductibility of the land tax from income tax.

The effects of a rural land tax when raised on market values of agricultural land are more severe than when raised on the productive potential (agricultural use value) of the land, and this effect differs between regions. The distortionary effects of using market values as the base for calculating land values are, therefore, larger than in the case where productive value is used. This will effect the profitability of different regions of the country, and, through its effect on profitability, may distort production patterns. This kind of distortion is most severe in those regions that are closest to the market, i.e. when non-market factors influence the market value of land the most. The way in which the productive potential of the land is calculated will determine the level of the tax rate.

One explanation for opposition to an agricultural land tax by farmers is that current landowners believe that the cost to them may end up being considerable more than the tax alone. This study has shown that farmers will also incur a loss in land values. In terms of a 1 per cent land tax rate levied on market values, a land tax will have a average negative effect of 5,3 per cent on predicted real land prices. In the case of a 2 per cent rate levied on market values, the effect resulted in a 12,32 per cent decline in the market values. It should, however, be noted that these results represent a worse case scenario, as it is assumed farmers cannot shift the tax burden to consumers or tenants through higher prices or rentals.

It is often argued that the tax-induced decline in land prices encourages access to land for new farmers such as participants in a land reform programme, young people or tenants wishing to own land. It has, however, been shown that the effects of a land tax will not only fall on the commercial farmer but also on the young and emerging farmer. It can thus be concluded that although land will be cheaper, profits will be lower and farmers will thus not necessarily be better off with the land tax and the land market will not favour new entrants.

This chapter also showed that, over the long term, the most severe effect of a land tax is on the level of short term liabilities of farmers, as farmers are forced to use more debt to operate their farms at an optimal level.

The effect of the forward and backward linkages between agriculture and the rest of the economy also have to be accounted for in assessing the total impact of a land tax. This study has shown that the multipliers in agriculture are the highest relative to the other economic sectors. Thus, although the funds removed from agriculture will be re-invested into the local community, there will be a loss in total welfare for the relevant rural communities.

In conclusion, the research shows that, when measured against the principles for an efficient and coherent tax system, the potentially distortionary effect of the rural land tax on the agricultural sector can be minimised by regarding the tax as a provisional payment of income tax.

Further, the research reported here has shown that the decision to value land according to its market or use value should rest on local circumstances, including the capacity of the municipalities to implement a specific assessment mechanism. However, it is clear that there should be bias towards use value methods for agricultural land, as market valuations cause considerable distortions in areas where the discrepancy between market and use value is the greatest. In this respect, the experience in the United States has been that most local authorities now value land on the basis of its use value.

See also Eckert, JB; Liebenberg, GF and Trotskie, DP 1997. Compiling an Agricultural SAM for the Western Cape. Department of Agriculture of the Western Cape, Elsenburg.

Contents    Foreword    Chapter 1   Chapter 2    Chapter 3     Chapter 4   Appendix   Home