Harnessing the Forces of Urban Expansion: The Public Economics of Farmland Development Allowances

The Composition of Provincial Production Activities

Consider M otherwise identical provinces, i = 1...,M. Each province comprises two types of production activities, referred to henceforth as agriculture (subscript “a”) and nonagriculture (subscript “n”). Accounting for all other inputs and costs of production, land use in agriculture measured in efficiency units generates value added A().¹² A() is taken to satisfy standard properties, with A(0) = 0, and A_T() > 0 and A_TT() < 0. Nonagricultural land and capital input jointly enable nonagriculture production activities, generating value added N(Kⁱ ,). N(Kⁱ,) is strictly increasing and concave in both arguments, and N_κτ(Kⁱ,) > 0. All local governmental efforts (inclusive of land reclamation, land consolidation, and land rehabilitation) in amassing new and enriching existing land resources for agricultural use will be collectively referred to in what follows as land development . Let denote the extent of the conversion of agricultural land to support nonagricultural production. Furthermore, let and be the initial stocks, respectively, of agricultural and nonagricultural land. We have, therefore,

and

Thus, reflects the net increase in nonagricultural land use, while reflects the net increase in agricultural land use. Rearranging terms,

[1]

and

[2]

It follows that in order for total land use to exceed the baseline level , regardless of how the allocation is split between agriculture and nonagriculture, land development is indispensable, .

We assume that the minimum cost schedule of land conversion is strictly increasing and convex in the scale of land conversion, with γ(0).¹³ Similarly, assume that the minimal cost schedule of land development δ = δ() is strictly increasing and convex in the scale of land development, with δ(0) = 0.¹⁴

Fiscal Decentralization, Local Revenue, and Expenditure

Both local and central governments balance their own budgets. At the local level, sources of revenue include (1) (distortionary) taxation of value added generated from production activities, (2) land lease fees on nonagricultural enterprises that obtain land use rights, and (3) all other sources of nondistortionary taxation levied in the province , or net transfers from the central government , both of which will be taken henceforth as given exogenously.

With two sectors (agriculture and nonagriculture) generating value added, total tax take at the local and the central levels depends on a variety of factors. First, the tax rate τ on value added is mandated by the central government. For the case of China, since the Provisional Regulation of the People’s Republic of China on Value Added Tax in 1994, τ stood at 13% for agricultural products, while export-producing private enterprises receive full rebate on the value-added tax so that τ is effectively zero (Su and Zhao 2004; Beijing Local Taxation Bureau 2003). Furthermore, as emphasized by Qian and Roland (1998), foreign firms are able to evade taxes in host countries by transfer pricing schemes. To highlight this dichotomy in the tax treatment and tax take on agricultural and nonagricultural production activities, we assume henceforth that the value-added tax τ is strictly positive in agriculture and zero in nonagriculture.¹⁵

As a second source of revenue, local government levies land lease fees on the transfer of land use rights to nonagricultural enter-prises.¹⁶ As discussed, the Law of Land Administration of 1998 stipulates that these fees are expected to be market determined via bid tendering, auctions, or listing, and the evidence shown in Section II also concurs. Accordingly, we let land lease fee per land unit be given by the marginal product N_T(Kⁱ ,) in each province i. Total revenue from the land lease in province i is thus N_T(Kⁱ,).

The assignment of tax revenue between the central government and the local governments is taken to follow a simple guiding principle (Su and Zhao 2004; Beijing Local Taxation Bureau 2003): taxes that involve national interest or the macroeconomy are assigned to the central government (e.g., import tax); taxes pertaining to local economic activities come under the full control of subnational governments (e.g., local-level land lease fees), while the value-added tax is shared at a given rate determined by the central government. Accordingly, let υ be the share of local value-added tax revenue retained by the province, and 1 -υ the share repatriated to the central government. υ parameterizes the extent of fiscal decentralization with υ = 1 reflecting full fiscal decentralization and full tax revenue retention within each province. According to the Beijing Local Taxation Bureau (2003), value-added taxation revenue follows the sharing rule of 75% for the central government and 25% for local governments, or υ = 0.25.

The revenue that a local government collects in province i is thus

while tax revenue of the central government include collections from all M provinces is

Budget balance at the local level requires

[3]

Budget balance at the central government level requires

[4]

where Z is a public good at the national level.

Interregional Competition for Capital

Capital is mobile across provinces. The marginal return to capital in province i is given competitively by

The nationwide aggregate stock of capital is . Arbitrage made possible by capital mobility implies that returns to capital are equalized across provinces, and thus for all i,

[5]

Assume henceforth that each province takes the nationwide returns to capital, r, as given. Equation [5] illustrates the tight link between land allocated to nonagricultural uses in province i, and the ability of province i to attract foreign capital. In particular, let

It is straightforward to check that an increase in encourages capital inflow, since N_KT(Kⁱ,) > 0, with elasticity

Finally, the economy-wide return to capital r implicitly solves the following equality, which requires that the sum total add up to equal the size of the aggregate stock ,

Each local government takes the structure of production, the fiscal contract between the provinces and the center, as well as the presence of interjurisdictional competition for capital as given. The objective of each local government in i is to maximize the welfare function Wⁱ in i,¹⁷ where

[6]

subject to the budget constraint in [3]. The expression in square brackets indicates provincial-level economic well-being, as given by the value added generated in i, A() + N(Kⁱ,), net of taxes and land lease fees

τA() + N_T(Kⁱ,) + . U(Z) denotes citizen benefits from a national-level public good Z that the central government finances through central government tax revenue. The objective function of the central government is .

Nationwide First-Best Allocation

We begin our analysis by setting out a firstbest baseline. From [4], the central government chooses capital allocation Kⁱ, land use and in each i, along with public good provision Z to maximize W:

[7]

subject to the budget constraint that total (local plus central) government tax revenue equals total expenditure:

[8]

and . Let superscript “o” denote these first-best outcomes. Equations [7] and [8] together imply that the central government’s maximization problem is simply

[9]

With M otherwise identical provinces,

[10]

with symmetric allocation of across the M provinces. Public good provision equalizes the marginal cost with marginal benefits aggregated across provinces:

In addition, land development in each province equalizes the marginal gains in agricultural value added and the marginal cost of land development:

[11]

Land conversion in a first-best equilibrium yields marginal gains in nonagricultural value added at Nτ(K^o,Tl). In terms of marginal costs, two forms are incurred, including first the direct marginal cost of land conversion y_t(to), in addition to the marginal cost of land development δ_t(to) to replenish an otherwise declining usage of agricultural land T_a:

[12]

Equations [11] and [12] jointly determine the first-best levels of land use, respectively, in agriculture and in nonagriculture (Figure 3). Schedule D⁰ D⁰ in Figure 2 traces the combinations of T_a and T_n that equate the marginal benefits of land development to the corresponding marginal cost. Note first of all that D^o D^o slopes downward. Intuitively, an increase in land use in nonagricultural production, all else constant, raises the marginal cost of additional land development for agricultural uses by convexity of δ. This discourages land use in agriculture.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

First-Best Land Development and Land Conversion

Schedule C^o C^o in Figure 3 in turn traces the combinations of T_a and T_n that equate the marginal benefits of land conversion to the corresponding marginal cost. Note that C^o C^o also slopes downward. Intuitively, an increase in land use in agriculture, all else constant, raises the marginal cost of additional land development for nonagricultural uses by convexity of δ from [12]. This discourages land use in nonagriculture.

The equilibrium land use allocation in the two sectors is given by the intersection of the D^oD^o and the C^o C^o schedules.¹⁸ These findings establish a set of benchmarks, based on which we will evaluate three additional policy regimes.

Incomplete Fiscal Decentralization and Interregional Competition for Capital

Contrary to the first-best outcome, Chinese provinces confront decentralized decision making with incomplete revenue retention υ < 1, and interregional competition for capital (Kⁱ (,r)). As noted, each provincial government maximizes the posttax income generated in the province, subject to a budget constraint in [3], taking as given the nationwide returns to capital r. Using [3] and [6], the problem of each provincial government i simplifies to

[13]

In a symmetric equilibrium, Kⁱ, , , and zⁱ are identical across regions. Let a superscript “c” denote these outcomes with interregional competition for capital. We have

[14]

so that the aggregate stock of capital is equally allocated once again across the M otherwise identical provinces. Thus,

Land development in each province equalizes the marginal gains in agricultural value added and the marginal cost:

[15]

Comparing [11] with [15], where the latter exhibits incomplete revenue retention parameterized by v < 1, provincial government underestimates the marginal benefits of land development relative to the first-best regime. In Figure 4, the DⁱDⁱ schedule reflects this difference by once again tracing out the combinations of land use in agriculture and nonagriculture that maximize provincial welfare, indicating a scaling back of agricultural land use relative to the first-best regime for all .

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Decentralized Land Development and Land Conversion

Now, in sharp contrast to the case of agricultural land development, local government overestimates the marginal benefits net of marginal costs of land conversion:

[16]

where

and and are the shares of capital and land in nonagricultural production. Thus θ captures both the sensitivity of capital inflow Kⁱ to land use in nonagriculture and the impact of capital inflow on nonagricultural production. θ accordingly reflects the strength of the impact of interjurisdictional competition for capital on land conversion incentives.

Now compare [12] and [16]; competition for mobile capital effectively shifts the CC curve outward. As shown in Figure 4, the revised CⁱCⁱ schedule along with the DⁱDⁱ schedule jointly determine the equilibrium land allocation decision of local governments. Incomplete fiscal decentralization and competition for capital thus reinforce one another in augmenting local officials’ incentives to increase land use in nonagriculture, at the expense of agricultural production. We have the following proposition:¹⁹

Proposition 1. With incomplete fiscal decentralization and interregional competition for capital, there is overprovision of nonagricultural land by local governments and underprovision of agricultural land relative to the first best: . This is made possible by an increase in land conversion effort, but a decrease in overall land development effort relative to the first-best regime if local tax revenue retention is sufficiently poor, or with 1 -υ, is sufficiently large: .

Since by definition, nonagricultural land use greater than the first-best level necessitates land conversion effort greater than the first best. Now, , and as such, overall land development effort depends jointly on land allocation in both agriculture and nonagriculture, which move in opposite directions relative to the first-best allocation due, respectively, to incomplete tax revenue retention and competition for capital. If the former (tax retention) effect dominates, agriculture suffers doubly due to land conversion and an overall reduction in land development effort.

Land Development Allowances

In order to prevent unchecked urban expansion and the decline in agricultural production due to land conversion, various policy options are available. At one extreme, a ban on land conversion sets . Meanwhile, a comprehensive system of agricultural land use quotas with a ban on uncompensated land conversion requires that agricultural land use must never be fall below a given quota, and as such every unit of land conversion must be compensated by land development of at least the same scale. The Circular of Further Strengthening Land Management and Cultivated Land Protection, jointly issued by the Chinese Communist Party Central Committee and the State Council on April 15, 1997, is one such policy that specifies a target of dynamic balance of total cultivated land (Liang and Huang 2000). Yet another alternative policy puts checks on land conversion by a policy of land development allowances, whereby expansion in nonagricultural land use is allowable only as a fraction of newly added agricultural land. The Law of Land Administration enacted in 1998 (Chapter 4, Article 18) officially sanctions the use of land development allowances, where 60% of agricultural land newly added through land consolidation can be used as the compensation quota for cultivated land occupied for construction.²⁰

Each of these policy options is a special case of the following general formulation of a land development allowance policy, which links allowable land conversion to land development efforts. Formally, a land development allowance policy is a pair (ρ, ∆), which requires that land conversion in each province i be no greater than²¹

[17]

where ρ ∈ [0,1] gives the size of land development allowance permitted as a fraction of total land development. ∆ is a minimal land development quota, and land conversion is permitted under this policy when land development exceeds ∆. This policy imposes a constraint on the feasible land use choices in agriculture and nonagriculture facing each provincial government. The constraint is shown in Figure 5 as the upward sloping schedule ∆∆. Anywhere along the schedule, the constraint is binding with , or equivalently, . A ban on land conversion is thus the simplest case where ρ = ∆ = 0, and ∆∆ is thus a horizontal schedule at .

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Land Development Allowance Policy,

Alternatively, a law that stipulates a dynamic balance in the stock of agricultural land at no less than requires that ρ = 1 and ∆ = whenever ∆ ≥ 0. Once the quota ∆ is reached, land conversion is permissible so long as a “dynamic balance” in at the specified level is maintained by compensating each unit of agricultural land loss through land conversion with the same degree of land development (ρ = 1).

More generally, varying ρ ∈ (0,1) and Δ ≤ 0 allows the central government to finetune the land allowance policy. The question remains as to whether the policy can be designed to exploit local governments' incentive for urban spatial expansion to benefit agricultural land development more generally. To see this, consider once again the local governments' problem, now augmented with a centrally mandated policy of land development allowance:

[18]

subject to [17] or equivalently,

and

In a symmetric equilibrium, Kⁱ, , and thus Tnⁱ continue to be identical across regions. Let a superscript “r” denote these outcomes with land development allowance, fiscal decentralization, and interregional competition for capital. We have

[19]

With the land development allowance as given by [17], an increase in agricultural land use made possible by land development additionally impacts the ability of the local government to expand nonagricultural land use, given ρ and Δ. At an interior equilibrium, we have

[20]

It follows that the marginal welfare impact of an increase in land use in agriculture is now a weighted average of (1) the marginal welfare impact of an increase in , and (2) the marginal welfare impact of an increase in in the absence of the allowance policy, . Furthermore, as long as the land development allowance policy is binding(), the combinations of land use in agriculture and nonagriculture that satisfy [20] lie between CⁱCⁱ and DⁱDⁱ,as shown in Figure 5 as RⁱRⁱ .²² Note that RⁱRⁱ ends in the Cⁱ Cⁱ schedule as ρ tends to 1, and DⁱDⁱ otherwise as ρ tends to 0. The intersection of the upward sloping land development allowance constraint ΔΔ and the RⁱRⁱ schedule gives the local government choice of land use in agriculture and nonagriculture.

We now show that an appropriately designed land development allowance scheme can replicate the first-best outcome:

Proposition 2. A land development allowance policy (ρ^r,Δ^r) that replicates the first-best land allocation and takes the form

[21]

and

[22]

The result is intriguing for it shows that a land development allowance policy can reinvigorate rural land development efforts while reducing urban land use. From [21], and [22], the intuition behind this result is as follows. Note that land development allowance ρ^r as displayed in [21] essentially eliminates the disincentive for agricultural land development due to incomplete fiscal decentralization τ(1 — υ)A_T by pairing it with the heightened incentive for urban spatial expansion in the presence of mobile capital θN_T. Next, by accordingly calibrating ∆^r based on the firstbest land allocation and ρ^r,²³ an appropriately designed land development allowance policy can indeed replicate the first-best outcome. Doing so increases land use in agriculture and tames excessive urban spatial expansion in each province. Furthermore, relative to the decentralized outcomes (), the land development allowance policy strictly increases the total land development for , as shown in Proposition 1, as long as local tax revenue retention is sufficiently poor. Jointly these findings are consistent with the experience of land policy reform and the massive increase in land consolidation efforts in China as discussed earlier.

Proposition 1 also provides clear guidance on the determinants of the size of the land development allowance ρ^r, and the size of the quota ∆^r. Specifically, upon an increase in the extent of interjurisdictional competition for capital, due to an increase in the elasticity θ, for example, the land development allowance policy should be stricter, requiring a reduction in ρ, but simultaneously a relaxation of the agricultural land development quota through a reduction in ∆. Meanwhile, a change in the fiscal contract, such as a reduction in υ that worsens local tax revenue retention even more, should call for a increase in ρ to encourage local governments to take advantage of the land development allowance policy, while at the same time increasing the quota ∆^r to ensure that the first-best allocation can be reached.

Note furthermore that the land development allowance policy (p^r,∆^r) is feasible only if the land development quota ∆^r = To — Ta — {[θ NT(CK/M),To)]/[r(1 — v)AT(To)]} (Tn — T_n) is positive. A priori, the expression in [22] can be positive or negative, depending on the relative magnitude of the first-best allocations, and the marginal value-added impact of land in the two sectors N_T and A_T. Alternatively, therefore, consider a subsidized land development allowance policy (ρ,S). Specifically, let s denote a subsidy transfer from the central government to province i for each unit of land development carried out. The cost of land development in province i is thus . With an earmarked subsidy replacing the role of the land development quota, the land development allowance policy restricts land conversion via the constraint .

It can be shown that first-best land allocation can be replicated here as well. Consider once more the local governments’ problem, now augmented with a centrally mandated policy of land development allowance with earmarked subsidy:

[23]

subject to

and

Let a superscript “s” denote these outcomes with subsidized land development allowance, fiscal decentralization, and interregional competition for capital. We have

[24]

Furthermore, at an interior equilibrium,

[25]

where the impact of a subsidy s is highlighted. We thus have the following:

Proposition 3. A land development allowance policy with earmark subsidy (ρ^s,s^s) that rep-lic_oates the first-best land allocation and takes the form

[26]

and

[27]

Thus, the land development allowance policy requires an earmarked subsidy to achieve first best if the land development disincentives induced by incomplete fiscal decentralization 1 — υ are sufficiently high. Alternatively, if competition for capital is sufficiently intense, the first-best replicating land development allowance policy can in fact be a source of central government tax revenue for s^s < 0, from [27].