Thailand's Troubled Tax Computerization Project

Abstract

In Februrary 1992 the World Bank signed a loan agreement with the Thai government to carry out a large and complex Information Technology (IT) project in the Revenue Department of Thailand. This was not an e-government application, as the project focussed on back-end computerization without any element of online service delivery. (Tax computerization projects in Singapore and the Philippines provide an interesting contrast, and are documented on this website.) Still, the Thai case is instructive for the problems it encountered. The project was to be completed by the end of 1996. However, the software contractor defaulted on the contract in 1993, forfeiting its bond. Subsequent evaluations acknowledged the project as a failure. This case study attempts to draw lessons from that experience.

Application Context

During the period from 1986 to 1995, Thailand enjoyed one of the fastest economic growth rates in the world -- averaging 8.5% a year. Income taxes, consumption taxes, and international trade taxes were Thailand's primary sources of tax revenue, representing 28%, 49%, and 22%, respectively, of total tax revenues during this period. Tax revenues represented, on average, 91% of total government revenues and over 16% of GDP.

The tax systems used to collect and process these revenues were slow and inefficient. A lmost all taxes in Thailand are collected by three departments of the Ministry of Finance (MOF): the Revenue Department (RD), Excise Department (ED), and Customs Department (CD). Only about 38% of the 6.5 million tax returns received by the RD were computer-processed. There was little cross-checking, even of computer-processed returns, and collection and compliance verifications were rudimentary. There was almost no tax data retrieval capability to assist auditors in data processing or applying audit criteria to tax returns. Data processing staff was fully occupied in keeping old software running, leaving no time for development of new software, reports, or additional functionality.

A New Approach

In January 1992, the government introduced a major tax-restructuring program meant to increase the efficiency of the tax system, make the tax structure more competitive with regional neighbors, and prepare for a potential shift to indirect taxes as a larger share of revenues. Based on recommendations made by the IMF and World Bank teams in the late eighties, the business tax (a tax levied on gross receipts) was replaced by a Value Added Tax (VAT) and the effective rates on corporate and personal income taxes were lowered.

Simultaneously, the government embarked on a program to modernize the Revenue Department and sought World Bank financing to overhaul its existing computer system. The objective of the project was "to support tax administration through the provision of hardware, software, and systems for tax processing and analysis."

The goal of the project was to assist the RD in establishing a comprehensive, integrated data base management and processing system for all taxes it administers. These include the personal income tax, the corporate income tax, the business tax, VAT, petroleum income tax, foreign travel tax, stamp duties, and the withholding tax. A national computer system was expected to:

• decentralize tax administration;
• introduce a single Tax Identification Number (TIN) to integrate all taxes and returns for each tax entity; and
• provide computerized billing and collection of all taxes.

In turn, these changes were expected to produce:

• faster and more efficient processing and collection of revenues, and
• improved tax records, with better access and ability to conduct audits at each administrative level.

An Invitation for Bids (IFB) was issued in September 1990. In August 1991, the "turn-key" contract was signed with the "Revenue Department Consortium" (RDC). IBM was lead member of the Consortium, which included a software contractor and a number of Thai communications and computer services firms. (The Thai Revenue Department was not a party to the Consortium.) Revenue Department funds were used to begin work under the contract, while the World Bank loan agreement was signed in February 1992.

The IFB requirements specified the following phased schedule of implementation:

1. mainframe installation, 9 special data entry systems for regions, 67 stand-alone terminals for Bangkok, limited VAT and business tax software;
2. second mainframe in Bangkok, 9 smaller mainframes in the regions (all linked together), and implementation of all software;
3. installation of 48 small mainframes, and 175 microcomputers in provinces and districts;
4. installation of another 334 microcomputers in provinces and districts; and
5. installation of remaining 349 microcomputers and testing of overall system.

The project was to be completed within 3 years, with a provision of 2 additional years for facilities management and system maintenance. Final completion was scheduled for December 31, 1996.

Implementation Challenges

The project faced delays from the start. The RDC delivered the hardware, system software, database software, and telecommunications equipment as specified in the contract, but failed to develop adequate project specifications and was therefore unable to develop the applications software. When the RDC finally specifed and designed suitable applications software, it became clear that the hardware, operating system software, database software, and telecommunications software and hardware that already had been delivered would have to be replaced.

In 1993, the software contractor defaulted on the contract, forfeiting its bond, and leaving to the other RDC partners responsibility for the delivery of the applications software. Then, in September 1994, a major restructuring of the project was proposed. This plan included significant changes in hardware (improvements in processing capacity, more online disk storage capacity, a more suitable operating (UNIX) system, a relational database system (ORACLE), and a non-proprietary network operating protocol (TCP/IP)); but these changes extended the project schedule by two years. The new plan introduced better system specifications, but did not solve the problems of project management. Applications software delivery continued to fall behind schedule.

Due to the lack of centralized milestone reporting, there was no single point from which management could obtain a clear picture of project progress. Without a precise measurement of work to be done (there were thousands of tasks and sub-tasks requiring many thousands of work hours and the co-ordination of hardware, systems software, and network installation and support activities), task times, and resource estimates, underlying schedules tended to be inaccurate and of little operational value.

The RD assigned several senior managers to the project, as well as a dozen support staff. However, management responsibilities were not clearly defined (clear written terms of reference) and some key roles such as the Implementation Manager had remained unfilled. The Implementation Agency team lacked skills in systems integration engineering, facilities management, applications development, and software engineering.

The RDC project team was under-resourced but pressured to achieve optimistic time scales, which may have contributed to the loss of certain key project staff. The head of Facilities Management (and four of his immediate subordinates), as well as the head of Applications Development, left the project within three months of each other. Throughout the project, a major challenge to the RDC management was finding and retaining suitably skilled and experienced staff to work in Bangkok. The Team Leaders may not have had sufficient experience in the use of LBMS, CASE, and Oracle or in the development of distributed database applications.

No formal business process review and rationalization was conducted prior to or as part of developing project specifications. Such a review was finally completed in October 1993, three years after initiating computerization. Not surprisingly, the review was driven by the computerization exercise rather than the other way round.

An independent audit report was issued in February 1995. The World Bank placed the project under "intensive supervision" and 3 missions were undertaken in 1996. In November 1996, an extension of the project to the end of 1997 was requested by the Thai government. But in view of the poor prospects for the delivery of useable software specifications in 1997, the World Bank decided not to grant an extension to the project.

Benefits and Costs

The related estimated project costs included US $41.6 million for hardware and related peripherals, $11 million for software, $1.2 million for a large training program package for the RD staff, and $2 million for communications and utilities.

The project was deemed unsuccessful by the closing date, December 31, 1996. A comprehensive database management and processing system had not been established. There was also very little application software development -- limited VAT, limited Small Business Tax (SBT), and email had been established at a few locations. A full-function VAT, SBT, Corporate Income Tax, Personal Income Tax, and MIS were still to be delivered, and the database conversion resulted in the loss of much important data.

Key Lessons

The lack of an appropriate and well-specified objective for the project was the beginning of all the troubles. The project's proponents were overly eager to purchase and install hardware and devoted inadequate attention and patience to project preparation. There was surprisingly little emphasis on developing the RD as an institution, in increasing the efficiency of collection and improving the quality of auditing.

Modern tax systems are based on individual accounts so that the full picture of a taxpayer can be constructed. Designing a system to work on many different platforms can pose problems of data integrity. The architecture seemed to facilitate the work of informatics people at the cost of efficiency and effectiveness of business procedures. Certainly data migration needs to be planned more carefully than it was in this case, as it is always a problem to automatically migrate non-normalized data to a normalized database.

Although the development team was under-resourced, it was thought that pressure on the team could compensate for under resourcing. In fact, a lack of human resources and experience contributed to a poorly prepared bidding document. The IFB included requirements that were met by only one bidder. The system specifications produced by the RD were either too specific (e.g., specifying main frames, X.25 network protocol, etc.) or too generic (e.g., software requirements were not specific enough).

In implementing the project, proper project monitoring tools were not used; and the complexity the software development process was neither understood nor assessed. An implementing agency needs to have expertise in contractor management, including knowledge of legal aspects of contract administration, to help determine such issues as change-of-scope and compliance with contract terms.

In summary, the tasks of requirement specification, analysis, and design must be carried out with due diligence and skilled resources done at the initiation stage. Should gaps remain, the implementation authority should be able to spot the weaknesses in a timely manner through the use of appropriate project management tools.

Case study authors: Robert Schware and Subhash Bhatnagar
Information used to develop the case: World Bank project documents and comments from World Bank staff.
Date submitted: February 22, 2001