Developing codes from the transcripts is the initial step in the multistage analysis of qualitative data
(DeCuir-Gunby et al., 2011). Ryan and Bernard (2003) indicate that typically codes can be theorydriven where a priori codes are developed from existing theory or concepts or data-driven in which
new codes emerge from the raw data. As previously mentioned, the purpose of using the case study
is to explore whether the conditions and factors identified from the literature are applicable in a
particular case of OGD citizen-led engagement and to identify those missing from the literature but
emerge from the case. Therefore, we followed recommendations by DeCuir-Gunby et al. (2011) for
developing both theory-driven and data-driven codes.
To develop theory-driven codes, firstly the first author created a codebook containing a list of codes
and their meanings from the results of the literature review, as recommended by Saldaña (2013).
Secondly, the author checked the existence of these codes against the texts from different sources of
information including interview transcriptions. If a code exists, then its descriptions provided by
interviewees or other sources of information were labeled. Finally, the codes were shared among two
other authors and validated based on discussions. Code reviews and revisions and reliability
evaluations were conducted in this final step.
The data-driven code development was started by splitting a meaningful group of sentences related
to conditions and factors of OGD engagement of Kawal Pemilu from its founder transcriptions, as
recommended by Saldaña (2013). The first author assigned potential themes to these subsamples of
information. Then, the texts of other sources of information were evaluated to find similarities with
the developed themes. Next, the first author created codes based on the emergent similar themes
and tagged the texts with the codes. Finally, the codes were shared among three other open data
researchers and the reliability was assessed. The codebook can be found here: [link removed for blind
peer review].

4. Case study background
The Indonesian presidential elections are organized and managed by the Election Commission or
Komisi Pemilihan Umum (KPU), at the national level, and its branches at provincial (KPU Provinsi or
KPUP), municipal (KPU Kabupaten/Kota or KPUD), district (Panitia Pemilihan Kecamatan or PPK),
village (Panitia Pemungutan Suara or PPS), and polling station level (Kelompok Penyelenggara
Pemungutan Suara or KPPS) respectively (Komisi Pemilihan Umum, 2014b). Votes are manually cast
and tallied by the KPPS members at polling stations in the C1 form. The tallies are then aggregated
manually at higher administrative levels until reaching the national level from C1 to D1 (village), DA1
(district), DB1 (city/municipality), DC1 (province), and finally DD1 forms (national). Figure 4 illustrates
the hierarchical processes of the presidential election tabulation.