How AI and Remote Sensing Technology Can Help Boost Property Tax Collections
Published: 07 December 2020
Machine learning and satellite imagery can provide data needed by governments for effective property tax management more quickly and efficiently.
Property taxation is the main revenue source for local governments. If well designed, it can contribute to social equity because of its progressive nature.
However, this revenue potential is largely untapped in many countries. In developed countries, as represented by member countries of the Organisation for Economic Co-operation and Development, the ratio of the revenue from property taxes to the gross domestic product (GDP) in 2018 reached only 1.1%. In Southeast Asian countries, such as Cambodia, Indonesia, Malaysia, Philippines, Thailand, and Viet Nam, the revenue was much lower at less than 0.1%–0.4% of GDP.
The underperformance from property taxes is largely attributed to the high cost of establishing and maintaining a functioning tax register and challenges in collecting taxes effectively and assessing land values accurately.
Artificial intelligence (AI) and remote sensing technology can provide a solution to overcome these challenges by providing a systematic approach to property taxation through better data collection and faster analysis. In addition, satellite imagery and AI can reduce the costs and time required to collect data that traditionally required immense labor costs and time.
How does artificial intelligence in property taxation work?
AI can quickly produce vast building footprints from images captured by satellites or aircraft. But it first requires samples to show what phenomena to detect. These samples—often referred to as training data—allow AI to learn the complex relationships between inputs and desired outputs. Once learning is complete, AI is capable of generalizing the inputs.
Similar to how humans learn when taught something new, AI will learn a simpler phenomenon faster than a complex one.
In the case of detecting changed buildings, information is given to the AI system to learn about the changes in buildings. And then, it is taught to distinguish these changes through images that show seasonal variation in vegetation and the natural landscape.
Figure 1: Artificial Intelligence Building Detection for Pleiades Imagery.
As an example, two AI models were developed for detecting changes in buildings: (i) building detection model, which identifies where the buildings are located; and (ii) change detection model, which classifies if a building has changed. These models are capable of rapid and widespread detection of buildings and their changes as captured by satellite imagery.
Figure 2: Overall Process for Change Detection Using Artificial Intelligence
The AI change detection model allows governments to investigate changes on a wide range of buildings (showing whether these are newly constructed, demolished, or rebuilt) within a short amount of time.
In addition, if tax authorities have tax records linked to building footprint maps, they can obtain more detailed information of the tax status of each building. It allows governments to see which buildings are taxed and not taxed.
How can artificial intelligence and remote sensing technology address problems in current property taxation systems?
A major issue with current evaluation methods is the high cost of field survey methods to map properties. Remote sensing is able to capture the extent of individual structures (buildings or homes) at a specific time, allowing one to identify the building’s area on the land without on-site measurements.
Areas that are far from population centers and difficult to access can also be checked using this technology. Data collection without physical excursions greatly reduces labor costs and time required to conduct field surveys.
Evaluation of property across cities at a single point in time can be difficult with field surveys. Remote sensing—satellite imaging in particular—can capture large areas of land at once. The visualization of land with a single snapshot enables the evaluation of properties at the same point in time across cities. Property evaluation at a uniform period is important for a systematic and comprehensive property tax system.
The vast amount of land that remote sensing imagery covers makes it impractical to analyze each image manually. AI can process vast amounts of data with speed and accuracy. Tasks such as identifying buildings and calculating building size can be automated to streamline property taxation services.
Property taxation must be systematic and comprehensive in order to be practical in developing economies. The systematization of a single aspect of property taxation can increase the robustness of property taxation methodologies and reduce tax evasion and underreporting.
The Asian Development Bank is currently implementing a pilot project in several cities in Nepal to create parcel maps with accurate building footprints using high-resolution imagery and artificial intelligence. The data will be used to generate updated tax maps and property rolls. This can be used to compare the actual property tax revenue with a model of the potential tax yields. The model will use current and historical property valuations from recorded real property information and the physical changes extracted from satellite images.
D.A. Ali, K. Deininger, and M. Wild. 2018. Land Governance Policy Brief: Using Satellite Imagery to Revolutionize the Creation of Tax Maps. Washington, DC: World Bank Group.
D.M. West. 2018. What is Artificial Intelligence? Brookings Institute: A Blueprint for the Future of AI 2018–2019. 4 October.
K. Nemoto, et al. 2018. Classification of Rare Building Change Using CNN with Multi-Class Focal Loss. Proceedings of the IEEE International Geoscience and Remote Sensing Symposium. Valencia. pp. 4663–4666.
Organisation for Economic Co-operation and Development. 2019. Revenue Statistics 2019. Paris: OECD Publishing.
R. Hamaguchi and S. Hikosaka. 2018. Building Detection from Satellite Imagery Using Ensemble of Size-specific Detectors. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops. Valencia. pp. 187–191.
Leave your question or comment in the section below:
YOU MIGHT ALSO LIKE
The views expressed on this website are those of the authors and do not necessarily reflect the views and policies of the Asian Development Bank (ADB) or its Board of Governors or the governments they represent. ADB does not guarantee the accuracy of the data included in this publication and accepts no responsibility for any consequence of their use. By making any designation of or reference to a particular territory or geographic area, or by using the term “country” in this document, ADB does not intend to make any judgments as to the legal or other status of any territory or area.