Doctor of Philosophy / Agricultural Economics / Allen M. Featherstone / Christine Wilson / Farmland values have reached all-time highs and have significantly risen over the last few years. This has caused much debate about whether farmland prices are currently on a bubble and ready to burst, much like the earlier 1980s. Much research has been done on farmland values; however, work done outside of agricultural economics, looking at general asset values, can be incorporated into models of farmland value. Information cascades, or herding, are phenomenon where information in the market is sent between investors and this information is bid into the asset price, thus resulting in boom and bust periods. By using a Vector Autoregression (VAR) model, farmland price dynamics are modeled and analyzed for spatial dependencies from one region to the next. VAR allows for no a priori specification of network typology. This allows for the examination of the existence of information cascades and what form the network takes among spatially located farmland markets. This method is then compared to two other spatial estimation techniques. The first is a Spatial Autoregressive (SAR) model where network typology is imposed prior to estimation. The second is a VAR model where no network is modeled, and only the region’s own asset prices can influence future periods. It is found that information cascades exist and network typology is somewhat random. These results caution the current direction of the literature of imposing network or spatial structure. However, due to data requirements, SAR models are easier to estimate since they require less data and if network structure, which the SAR model inherently imposes by the weight matrix, could be determined by an autoregressive process instead of an adjacency rule it could prove to be the most accurate forecasting method.

Submitted by ANTONIO DIAS FIGUEIREDO FILHO (toni.figueiredo@terra.com.br) on 2016-02-26T04:39:17Z No. of bitstreams: 1 antonio_dissertacao_fgv_final.pdf: 794189 bytes, checksum: 2016ed8e754ff01d5d0ee3db29d7a691 (MD5) / Approved for entry into archive by Fabiana da Silva Segura (fabiana.segura@fgv.br) on 2016-02-26T14:04:50Z (GMT) No. of bitstreams: 1 antonio_dissertacao_fgv_final.pdf: 794189 bytes, checksum: 2016ed8e754ff01d5d0ee3db29d7a691 (MD5) / Made available in DSpace on 2016-02-26T14:12:27Z (GMT). No. of bitstreams: 1 antonio_dissertacao_fgv_final.pdf: 794189 bytes, checksum: 2016ed8e754ff01d5d0ee3db29d7a691 (MD5) Previous issue date: 2016-02-03 / A avaliação de imóveis é uma prática normatizada pela ABNT que vem desde a década de 60, quando surge a necessidade de se fazer grandes desapropriações. A normatização para imóveis rurais ganha corpo na década de 70 e desde então a ABNT NBR 14653-3 organiza e padroniza os conceitos da engenharia de avaliações. O método utilizado para se avaliar imóveis rurais se baseia em preços de mercado e este apresenta algumas fragilidades. Neste contexto, o objetivo deste trabalho é de propor o uso de uma metodologia, baseada no valor presente da renda do imóvel rural, que complemente a metodologia de preços de mercado, incorporando assim geradores de valor que são importantes para a gestão na alocação de capital em propriedades agrícolas. Discorre-se também ao longo deste trabalho a exemplificação da metodologia proposta por meio de um estudo de caso, abordando a caracterização do imóvel, premissas utilizadas para análise e a valoração do hectare da propriedade, comparando-a com a avaliação pelo método de preços de mercado. Acredita-se que esta abordagem baseada em fundamentos econômico-financeiros, aliada ao levantamento de preços de mercado trarão mais robustez para a avaliação de imóveis agrícolas com propósito gerencial, trazendo para discussão outras variáveis de relevância para a gestão agro imobiliária. / Real estate appraisal is a well established discipline in Brazil, which is standardized by ABNT, a Brazilian private entity, member of the International Organization for Standardization (ISO). Rural real estate appraisal standardization starts to build up from the 70´s and since then ABNT NBR 14653-3 organizes the appraisal discipline. The commonly used method to value land in Brazil is the sale value estimates from comparable sales, but it offers some fragilities. In this context, the objective of this research paper is to suggest a methodology capable of incorporating value drivers into farmland price, assessing the agricultural potential available at the property being analyzed. This method takes into account farmland rents to estimate its value, which adds to the current method applied by appraisers. This research paper also develops a case study to test the proposed methodology, approaching issues such as property agronomical characteristics, economic and financial assumptions to realize the property value. It is believed that this approach, based upon economic and financial fundamentals, coupled with sale value estimates will bring together a robust and comprehensive appraisal toolkit.

This dissertation consists of three essays wherein tools of financial econometrics are used to study the three aspects of farmland valuation puzzle: short-term boom-bust cycles, overpricing of farmland, and inconclusive effects of direct government payments. Essay I addresses the causes of unexplained short-term boom-bust cycles in farmland values in a dynamic land pricing model (DLPM). The analysis finds that gross return rate of farmland asset decreases as the farmland asset level increases, and that the diminishing return function of farmland asset contributes to the boom-bust cycles in farmland values. Furthermore, it is mathematically proved that land values are potentially unstable under diminishing return functions. We also find that intertemporal elasticity of substitution, risk aversion, and transaction costs are important determinants of farmland asset values. Essay II examines the apparent overpricing of farmland by decomposing the forecast error variance of farmland prices into forward looking and backward looking components. The analysis finds that in the short run, the forward looking Capital Asset Pricing Model (CAPM) portion of the forecast errors are significantly higher in a boom or bust stage than in a stable stage. This shows that the farmland market absorbs economic information in a discriminative manner according to the stability of the market, and the market (and actors therein) responds to new information gradually as suggested by the theory. This helps to explain the overpricing of farmland, but this explanation works primarily in the short run. Finally, essay III investigates the duel effects of direct government payments and climate change on farmland values. This study uses a smooth coefficient semi-parametric panel data model. The analysis finds that land valuation is affected by climate change and government payments, both through discounted revenues and through effects on the risk aversion of land owners. This essay shows that including heterogeneous risk aversion is an efficient way to mitigate the impacts of misspecifications in a DLPM, and that precipitation is a good explanatory variable. In particular, precipitation affects land values in a bimodal manner, indicating that farmland prices could have multiple peaks in precipitation due to adaption through crop selection and technology alternation.