Ex-Ante: Preferences, Price Effects, Lobbying Costs and Informal Linkages
Preferences
I start by evaluating interest pressures that may arise from differences in preferences across interest groups. Although there is no data on preferences towards carbon taxation, there is indicative survey data on preferences towards environmental and economic policy across interest groups (Table 2). Climate change (Column 1) is considered the most important development priority in Vietnam among a large share of those in bilateral and multilateral agencies, the Office of the Minister and research institutions. In contrast, less than 15% of those in local government, private sector and NGOs consider climate change the most urgent development issue. Pollution (Column 2) is perceived as a less urgent issue across all interest groups. The high importance given to pollution by the media seems to reflect the perception of the general public that pollution is a serious issue; air pollution and contamination from coal power plant operations have had large environmental effects (Ha-Duong, Nguyen, & Trinh, 2016), with emissions from coal power plant operation alone having caused 4,250 premature deaths in 2011 (Koplitz, Jacob, Sulprizio, Myllyvirta, & Reid, 2017).
The preference data further show that private sector development (Column 3) is considered more important than climate change by almost all interest groups, including government.[1] The data indicate that largest pressures against carbon taxation emanated from three interest groups: local government, private sector and NGOs. Among those groups, the highest level of interest pressures is predicted to come from local government. Importantly, the data suggests that pressures regarding carbon taxation would altogether have been minimal, because preferences of national government were in line with groups preferences to prioritize economic over environmental development.
Interest Group
|
Issue perceived as most important development priority (in %)
|
|
Climate change
|
Pollution
|
Private Sector Development
|
Bilateral/Multilateral Agency
|
38%
|
3%
|
35%
|
Office of Minister
|
33%
|
0%
|
50%
|
Academia/Research Institute/Think Tank
|
25%
|
12%
|
23%
|
Media (press, radio, TV, web, etc.)
|
23%
|
18%
|
31%
|
Deputy of the National Assembly, National Assembly Office/Committee
|
20%
|
0%
|
60%
|
Employee of a Ministry, Ministerial Department or Implementation Agency
|
17%
|
6%
|
24%
|
NGO/Community Based Organization
|
13%
|
18%
|
31%
|
Private Sector Organization
|
10%
|
17%
|
31%
|
Local Government Office or Staff
|
9%
|
6%
|
33%
|
Table 2 - Interest Group Perceptions (2017)
Source: Prepared by author based on data from the World Bank Group Country Survey 2017 – Vietnam. Notes: percentages are rounded. Some interest groups, including mass organization, youth groups, the judiciary branch and the financial sector have been omitted due to limited responses (<5). Total sample: 632 observations.
It is plausible that differences in preferences can account for the majority of variation in carbon taxation. For example, the minimal carbon price, the difference in carbon price across fossil fuels, the difference in increases in carbon prices, and the decreasing share of revenue recycling towards climate measures can be the result of preferences for economic development (and therefore large pressures against carbon taxation) across all interest groups. To explore these observations, the next section examines whether price effects can detect at a more granular level differences in pressures across different stratifications of households and firms.
Price Effects
Which interest groups would have felt most the effects from carbon taxation, and raised pressures because of those effects, is dependent on how price effects passed through to interest groups. The most direct price effect would have been through two expenditures; fuels, and indirectly, electricity through fuels as main input. Fuel prices remained largely stable over time, although there were temporal price hikes – for example, in 2011, when prices increased by 17% as fuel subsidies were reduced (Zimmer, Jakob, & Steckel, 2015, p. 23). Similar price hikes also occurred in 2009 when government adopted a market-based fuel price and increased fuel imports, which in effect meant increased exposure to international market volatility. Imports amounted to 71% in 2016 and were increased because of growing fuel demand and lacking national production (Petrolimex, 2017b, p. 63). In terms of electricity, price effects would have transferred partially to households and firms given that most sources of electricity generation have not been carbon-taxed. Renewable energy has not been levied and accounted for 50% of total electricity generation in 2013 (EVN, 2014, p. 33).[2] Electricity prices have also been susceptible to volatility in international energy markets, in large part as imports of coal – the major non-renewable source for electricity generation in Vietnam – more than doubled between 2018 and 2019 (Vu, 2019). This led final electricity prices to exhibit large hikes, despite that government controlled price increases among producers and operated a price stabilization fund. In 2014, for instance, electricity prices oscillated between VND 462 and VND 1318 per kWh (EVN, 2017, p. 33). These data suggest that households and firms would have experienced at least some inflation in fossil fuel and electricity prices from carbon taxation, which was amplified through exposure to volatility in international energy markets.
The extent to which increased prices materialized into price effects depends on the difference in response to marginal price increases across interest groups. For households, Table 3 reports on some of the parameters through which carbon taxation might affect various income groups.[3] The table shows, first, that household expenditure on electricity, fuels, transport and food (Column 1-5) is regressive across income – i.e. the richest quintile spends relatively less on it than the lowest quintile. This suggest that if producers would fully pass on price increases to consumers, the tax would disproportionally affect poorer households.[4] Second, the table shows that government subsidies (Column 6) – which include ‘other benefits’, such as on kerosene – are asymmetrically allocated, in favour of households in higher income quintiles. Electricity subsidies (Column 7), which in the table are weighted by electricity expenditure, are in relative terms received more by richer than poorer households. This shows that, although general fuel subsidies were reduced after 2012, subsidies on kerosene and electricity persisted, and were regressive. Furthermore, subsidies on fuels also continued because government kept control of petroleum prices to prevent large price increases. In 2017, for example, government increased and decreased the petroleum price both ten times (Petrolimex, 2017a, p. 50). Government also continued to provide electricity subsidies to poorer households. In 2013, for instance, it subsidized 1.84 million people in poor households by providing electricity at 65% of the standard tariff (EVN, 2013, p. 23). After 2014, poor households received a full reimbursement of electricity costs up to 30 kWh per month (UNDP, 2014, p. 33). This evidence predicts that lower-income households, which would experience largest price effects and received relatively limited subsidies to absorb effects, would express largest interest pressures against carbon taxation.[5]
Income Quintile
|
Electricity
|
Gas
|
Other fuels*
|
Transport**
|
Meals
|
Subsidies***
|
Electricity subsidy
|
As share of total expenditure (in %)
|
Share of group that received (in %)
|
Ratio subsidy-to-expenditure
|
Q1
|
1.6
|
0.6
|
0.1
|
23.1
|
14.7
|
1.5
|
0.05
|
Q2
|
1.9
|
0.6
|
0.1
|
19.1
|
16.0
|
3.1
|
0.01
|
Q3
|
2.0
|
0.6
|
0.1
|
19.8
|
15.6
|
2.8
|
0.17
|
Q4
|
1.6
|
0.3
|
0.2
|
21.9
|
13.7
|
6.1
|
0.83
|
Q5
|
0.8
|
0.2
|
0.1
|
20.5
|
11.4
|
9.2
|
0.46
|
Table 3 - Selective Household Expenditures and Subsidies (2015)
Source: Prepared by author based on data from the World Bank Household Registration Study 2015. Notes: numbers rounded. *includes oil, coal, woods. **includes fare, fuel, repairs. ***includes “other benefits”, such as for kerosene, but excluding education, health or food. Table excludes households without reported income, which could be from either employment, remittances, or government benefits. Total sample: 2,656 observations. Sample is representative of five provinces: Ho Chi Minh City, Ha Noi, Da Nang, Binh Duong and Dak Nong.
In similar ways, price effects from carbon taxation may unfold differently across firms. Table 4 provides data on firm expenditure indices for electricity and fossil fuels across sectors. It shows, first, that variation in electricity expenditure between sectors (Column 1) is large, e.g. a nine percent points difference between the food and retail sector. Second, differences in fossil fuel expenditure between sectors (Column 2) are less pronounced but remain significant, for instance, a six percent points difference between the mineral and metal products sector. Alternative statistics suggest much higher expenditure shares within specific sectors – e.g. fuel expenditure accounting for 33-59% of aggregate expenditure in the seafood sector (Viet Nam News, 2017a). This would mean that the price effect of carbon taxation would have been felt through increases in food prices. Third, data shows that there exists variation in electricity and fuel expenditure across firm size (Column 3,4) and firm ownership (Column 5,6). Further tests establish that these differences of means are not statistically different from zero.[6] Although these data do not cover all sectors, it is likely that price effects were most felt among firms in industry, rather than agriculture, because the industrial sector captures the largest share of total energy consumption – around 50% in 2017, followed by transport at 20% (IEA, 2020). The firm-level data does not provide information on firm subsidies, but qualitative evidence suggests that micro businesses, in particular farmers and fisherman, have received some support from 2010 onwards to cope with fuel price increases – e.g. subsidies on transport and diesel costs – although policies have not always been comprehensively implemented and sectors that are also characterized by relatively high fuel expenditures – e.g. transport and forestry – have not been supported (UNDP, 2014, p. 33). Farmers also did not receive a special electricity tariff with the carbon tax, despite that irrigation is fuel-intensive. This evidence predicts that the level of interest pressures against carbon taxation would have been largest among the food and mineral sector because of relatively high fuel and electricity expenditures, and that pressures would also have been high in the transport and agriculture sector because of limited subsidies to absorb price effects.
In addition to price effects through expenditure, carbon taxation may also affect households through changes in factor prices. Households that rely relatively strongly on income from sectors in which factor prices reduce relative to other factor prices because of carbon taxation will be more affected at the margin (Rausch, Metcalf, & Reilly, 2011). Reductions in labour price (or wage) is one way that household income is affected. Table 4 shows data on average income per employee (Column 7) as a crude estimate of labour price. It shows that sectors with relatively high shares of expenditure on electricity (e.g. food, services) are characterized by moderate labour income, while sectors most dependent on fossil fuels (e.g. mineral products, garments) are best described by low levels of labour income. This data suggests that factor price effects would have been experienced most by households in lower income quintiles, because, price effects propagated most through fuel rather than electricity expenditures. This evidence anticipates that lower-income households would have expressed the highest level of interest pressures against carbon taxation, because of highest exposure to price effects as well as declining factor prices.
Sector
|
Share of total expenditure (in %)
|
Labour income***
|
|
All firms
|
Small firms*
|
Private firms**
|
|
Electricity
|
Fuels
|
Electricity
|
Fuels
|
Electricity
|
Fuels
|
Food
|
11.9
|
5.7
|
12.6
|
7.7
|
9.4
|
5.7
|
81.1
|
Other services
|
9.1
|
1.0
|
5.3
|
0.3
|
3.6
|
1.2
|
118.0
|
Non-metallic mineral products
|
5.1
|
9.0
|
4.4
|
15.5
|
4.1
|
9.2
|
62.1
|
Garments
|
4.6
|
7.3
|
4.3
|
2.7
|
3.2
|
4.5
|
69.7
|
Other manufacturing
|
4.2
|
3.4
|
2.9
|
3.0
|
3.2
|
3.3
|
80.4
|
Fabricated metal products
|
3.2
|
3.0
|
2.1
|
2.4
|
2.1
|
3.0
|
84.6
|
Retail
|
3.0
|
2.5
|
0.4
|
0.5
|
0.7
|
2.5
|
73.7
|
Obs.
|
714
|
261
|
660
|
907
|
Table 4 - Firm Expenditure on Electricity and Fossil Fuels (2015)
Source: Prepared by author based on data from the World Bank Enterprise Survey 2015. Notes: mean value by sector, rounded. ‘Other Services’ includes transportation services. *<19 full-time, permanent employees, **full private ownership, ***average labour costs in million VND per full-time permanent employee. Excludes firms without information on electricity and fuel expenditure, or information on number of employees or ownership status. Total representative sample: 996 firms.
The firms that likely experienced most the effects of incoming carbon taxation are those in the electricity and fuel sector, which in Vietnam are almost entirely public.[7] Firms would be most affected through reductions in subsidies. Before the implementation of carbon taxation, fossil fuel and electricity producers and distributors received the largest part of all government subsidies, with consumers receiving a minor share of subsidies through direct transfers. Subsidies to SOEs occurred mainly through price controls and other provisions, including discounted or subsidized resources, preferential loans, loan guarantees, and tax breaks and concessions (UNDP, 2014, p. 4). The main electricity producer EVN received inputs frequently below market prices; the price for coal paid in 2012 was 60% of the export price and 70% of production cost. Similar subsidies were provided to state-owned coal producers PNV and VINACOMIN. Subsidies on fossil fuels constituted US$ 4.33 billion in 2011 (equivalent to a subsidization rate of 15.5%), of which 69% were subsidies on electricity and 31% subsidies on oil, gas and coal. After 2012, fossil fuel and electricity subsidies were phased out and reduced to US$ 0.1 billion in 2016 (T. C. Nguyen & Chuc, 2018, p. 15). However, government continued to subsidize SOEs despite reductions in fuel subsidies. EVN, for example, received preferential loans and other concessions to compensate for financial losses due to increased input prices, exchange rate depreciation, and for maintaining a below-cost tariff (H. Nguyen, 2018, p. 132). Furthermore, the coal and oil sector likely expressed largest pressures against carbon taxation because they captured 30% and 25% of total energy production in 2017 (IEA, 2020). This suggests that SOEs in the electricity and fuel sector, and in particular the coal and oil sector, would express largest interest pressures against carbon taxation.
These results suggest that three interest groups would have experienced largest price effects from carbon taxation, and as a result, probably exerted largest pressures regarding carbon taxation: (i) households in lower-income quintiles, (ii) private firms in the food and mineral sector, and (iii) public firms in the fossil-fuel and electricity sector. This could explain, first, why carbon taxation was set well below the international lower-bound of carbon pricing across all fossil fuels in 2012 and why some fossil fuels, such as diesel and coal, exhibited low carbon prices after 2012. It cannot explain well why there have been different carbon prices across fossil fuels, e.g. that in 2012 petrol was priced at US$ 22 per tCO2, diesel at US$ 9 per tCO2 and coal at US$ 0.2 per tCO2. While firms in the mineral sector and public fossil-fuel and electricity firms were predicted to exert large interest pressures and did receive relatively low carbon prices on their main inputs, lower-income households, which were anticipated to express large pressures due to high expenditure on petrol, received relatively high carbon prices. The persistence of subsidies from government to SOEs may explain why revenues from the carbon tax increased even though the share of revenues towards climate measures reduced; i.e. because it would be of interest for SOEs to have revenues recycled into the general budget to ensure the continuation of subsidies. While price effects can explain some of the variation in carbon taxation, the next section examines whether variation in lobbying costs can also account for some of the variation in carbon taxation.
Lobbying Costs
Whether interest groups are able to exert interest pressures depends on overcoming free-riding and coordination problems (Olson, 1965). I explore four variables that estimate the ability and cost of organizing lobby, following Shapiro (2019). The four variables – firm concentration, firm size, firm dispersion, capital intensity – provide the following hypotheses. First, industries with high concentration, or industries with a few large firms, may have lower lobbying costs because greater concentration allows to coordinate interest pressures more easily (Bergstrom, Blume, & Varian, 1985; Bombardini & Trebbi, 2012). Second, having larger firms and larger firm dispersion – larger standard deviation in firm size – may increase lobbying costs because greater size and dispersion limits coordination and increases free-riding (Bombardini, 2008). Third, capital intensity increases firm concentration and therefore may lead to reduced lobbying costs.
Table 5 reports on these four variables across sectors. Firm concentration (Column 1) is highest in the food sector and lowest in the fabricated metal sector, meaning that the food sector faces lowest lobbying costs, and therefore highest lobbying ability, compared to other sectors. Mean firm size (Column 2) is highest in the garment sector, which suggests that the sector experiences higher lobbying costs than other sectors. In contrast, the retail and metal sector are characterized by lower firm size and accordingly bear relatively lower lobbying costs. Firm dispersion (Column 3) is largest among the food and ‘other services’ sector, of which the latter includes services other than retail. It means that firms in the food and services sectors are relatively divergent in firm size and thus endure higher costs to organize lobby than sectors with lower firm dispersion, e.g. retail and metal sector. Capital intensity (Column 4) is the capital share in gross output. It is highest for the metal and ‘other manufacturing’ sector, of which the latter includes manufactures other than food, garment, metal and minerals. This implies that the lobbying costs for the metal and other manufacturing sector are in relative terms lower than for sectors with lower capital intensity, e.g. the food and retail sector.
Sector
|
Firm Concentration*
|
Firm Size**
|
Firm Dispersion***
|
Capital Intensity****
|
Food
|
0.79
|
344.0
|
1660.8
|
0.06
|
Garments
|
0.49
|
365.9
|
1088.6
|
0.31
|
Non-Metallic Mineral Products
|
0.32
|
143.4
|
331.8
|
1.29
|
Fabricated Metal Products
|
0.31
|
81.7
|
222.3
|
44.54
|
Other Manufacturing
|
0.52
|
148.7
|
560.6
|
15.52
|
Retail
|
0.45
|
53.3
|
118.8
|
0.11
|
Other Services
|
0.52
|
228.7
|
1467.9
|
7.96
|
Obs.
|
980
|
991
|
991
|
800
|
Table 5 - Ability and Cost of Lobbying (2015)
Source: Prepared by author based on data from the World Bank Enterprise Survey 2015. Notes: *the share of total sector output accounted for by the five largest firms in the sector, **includes full-time employees, ***standard deviation of firm size, ****the value of the capital stock (e.g. machinery, vehicles and equipment) divided by gross output (sales). Total representative sample: 996 firms.
In addition, I examine lobbying costs by using trade association membership at sector-level to proxy for political organization, following Ludema and Mayda (2013). The latent assumption is that unionisation is an important determinant of political organization and therefore lobbying costs. Table 6 reports on political organization by sector. It shows to be the largest for the “other services” sector and lowest for the fabricated-metal and retail sectors. If assuming that sectors are equally divisible, the results suggests that political organization is highest – and therefore lobbying costs lowest – for the “other services” and “other manufacturing” sectors and lowest (highest) for the fabricated-metal and retail sector. These findings are in line with the findings above, which also suggest high lobbying costs for firms in the services sector.
Sector
|
Political Organization Ability
|
Food
|
11
|
Garments
|
5
|
Non-Metallic Mineral Products
|
3
|
Fabricated Metal Products
|
1
|
Other Manufacturing
|
17
|
Retail
|
1
|
Other Services
|
26
|
Table 6 - Lobbying Costs
Notes: Online Appendix G provides references to sources and lists all associations by sector. The above table excludes associations that support all sectors (13 in total).
These results predict that lobbying costs are lowest (and therefore interest pressures highest) in the retail and metal sector, and highest (lowest) in garment and services sector. Can this variation in lobbying costs explain variation in carbon taxation? The first observation is that consumers, in comparison to producers, are often poorly organized (Olson, 1965). This difference in lobbying ability could explain why lower-income households, despite facing large price effects, did not achieve the low carbon price that producers using diesel and coal received. This would also explain why difference in carbon prices persisted over time. Lobbying costs do not explain why carbon prices increased over time, because the data suggest that no large lobbying costs existed among those producers that were predicted to experience the largest price effects. Lobbying costs could justify why tax revenues increased disproportional to carbon price increases. For example, the three-fold increase in carbon taxation in 2015 across all fossil fuels increased the carbon price in absolute terms most for fossil fuels which had lower shares in energy production (e.g. oil), while it increased carbon prices least for fuels with higher shares (e.g. coal). Against this background, the next section evaluates to what degree these interest pressures would have emerged as informal or formal pressures to government.
Informal Pressures
The historical background on Vietnam (online Appendix A) suggests that interest pressures occur through both formal and informal linkages. To approximate informal linkages, I consider three proxies: the frequency and relative amount of informal payments between private firms and government, and the share of SOEs in sectors. These proxies provide the following hypotheses. Sectors with higher informal payments and sectors with larger shares of SOEs express more pressures through informal than formal linkages, because such payments and connections provide firms with an additional and potentially more effective means to influence government. The hypotheses predict that sectors with high informal payments and connections (a) express less public pressures and (b) can pressure government more effectively, compared to sectors with low informal payments and connections.
Table 7 reports on the three proxies of informal linkages. It shows, first, that informal payments during tax collection (Column 1), which is a proxy of frequency of corruption, are persistent across all sectors, but highest in the metal and service sector and lowest in the garment and manufacturing sector. For instance, 24.2% of firms in the metal sector was requested a gift or informal payment during tax inspections, while 12.9% of firms in the garment sector. This predicts that firms in the metal expressed pressures more by means of informal channels and that those firms were more effective in expressing pressures to government than firms in the garment sector. The data on the percentage of total sales paid in informal payments (Column 2) indicate large informal payments across all sectors (note that the data reports on share of sales not of profits) and that those payments are highest in the services and retail sector, and lowest in the food and mineral sector. For example, firms in the service sector spend on average 5.8% of sales on informal payments, while firms in the food sector spend 1.0%. It suggests that the firms in the service sector pressured government more using informal linkages and more effectively than firms in the food sector.[8] Data on the share of firms with state-ownership (Column 3) shows that SOEs are most common in the mineral and food sector, and least in the ‘other manufacturing’ and retail sector, meaning that firms in the mineral and food sector likely pressured government most by means of informal connections and did so more productively than firms in other sectors. It is possible that variation in these three proxies is driven by differences in firm size across sectors; additional calculations show that this is not the case for informal payments, but is for share of SOEs.[9] To circumvent this issue, I weigh the share of SOEs by firm size (Column 4). It shows slightly different results from the unweighted proxy: that the presence of state-ownership is largest in the garment and food sector, and lowest in the ‘other manufacturing’ and services sector.
This evidence predicts, first, that pressures from the metal and service sector are in comparison to other sectors in large part exerted by means of informal linkages between ‘connected’ firms and government. Second, it predicts that interest pressures from the garment and food sector compared to other sectors are largely expressed through informal linkages between SOEs and government. This suggests that the pressures from the food and mineral sector, which were predicted to exert largest pressures regarding carbon taxation, likely surfaced differently: the former using informal channels such as SOE-government linkages, while the latter using formal channels. Moreover, the fact that informal payments and connections persisted across all sectors may explain why the share of tax revenues towards climate measures decreased. That is, if firms with informal linkages use such channels to attract or maintain subsidies or other favourable policies (see online Appendix B), it would be in the benefit of firms to persuade government to use carbon taxation to generate public revenue and use revenues to maintain favourable policies.
Sector
|
Informal charges
|
State-owned Enterprises
|
during tax inspections*
|
as share of sales**
|
as share of firms
|
as weighted share of firms***
|
Food
|
20.6%
|
1.0%
|
5.6%
|
13.2%
|
Garments
|
12.9%
|
3.6%
|
3.4%
|
15.2%
|
Non-Metallic Mineral Products
|
14.5%
|
2.1%
|
7.8%
|
4.7%
|
Fabricated Metal Products
|
24.2%
|
5.0%
|
2.8%
|
4.7%
|
Other Manufacturing
|
13.9%
|
3.0%
|
0.7%
|
0.2%
|
Retail
|
16.7%
|
5.6%
|
0.9%
|
2.8%
|
Other Services
|
23.4%
|
5.8%
|
4.2%
|
1.5%
|
Obs.
|
450
|
187
|
996
|
991
|
Table 7 - Informal Pressures (2015)
Source: Prepared by author based on World Bank Enterprise Survey 2015. Notes: percentages rounded. *Percentage of firms that was requested a gift or informal payment during tax inspections – 52% non-response rate (515 firms); of sample 6% was unable or refused to respond (31 firms). **Weighted average percentage of total sales paid in informal payments annually – 42% non-response rate (420 firms); of sample 86% unable or refused to respond (389 firms). ***weighted by firm size (i.e. full-time employees). Total representative sample: 996 firms.
This section examined the four observed policy changes in carbon taxation using predicted ex-ante interest pressures. I find that the fact that carbon taxes increased over time, although remained minimal, can be explained by clear preferences for economic rather than environmental development across all interest groups, including government. That taxes were unequal across fuels resulted from differences in lobbying costs across interest groups. Price effect data suggests that there should have been lobby to reduce carbon taxes across all fossil fuels because those most affected – lower-income households, private firms in energy-intensive sectors, and public firms in the fossil-fuel and electricity sector – either used one of the fuels. However, households, because of its size, likely faced coordination and free-riding problems, resulting in high lobbying costs and low ability, which may explain why carbon taxation on petrol increased more than on other fuels. That the share of revenue recycled towards climate mitigation decreased over time may be explained by large informal lobbying from public firms; because most public firms received firm subsidies, informal lobbying would allow them to persuade government to use carbon tax revenues to continue such subsidies. To develop these results further, the next section evaluates whether ex-post observed pressures can explain the remaining variation in carbon taxation and whether the ex-ante predictions can be confirmed.
Ex-Post: Observed Pressures and Policy Change
Discourse Analysis
In this section, I use insights from the discourse analysis to explain variation in carbon taxation. In terms of the differences in carbon price across fossil fuels, the results suggest that the ineffective pressures from academia and civil society, and the effective pressures from firms and households, can explain the divergent and low carbon prices across fossil fuels over time. The data show that local governments, academia and environmental NGOs strongly lobbied government to reduce the inconsistent tax rates across fossil fuels. For instance, during the 2010 government consultations, several environmental experts voiced that there should be a detailed evaluation of environmental impacts before setting the tax rates (Nam, 2010; VNS, 2010). Despite these pressures, government priced carbon differently across fuels in 2012 and maintained those differences despite recurring pressures in 2015, 2017 and 2018. The pressures also related to the minimal carbon prices that persisted across all fossil fuels. In 2015, environmental NGOs lobbied government regarding the limited level and coverage of the tax (Nam, 2015). In absence of any change in policy outcomes, however, the evidence suggests that these pressures were predominantly ineffective.
Firms and households were more effective in persuading government to adjust policy in their favour. The interest pressures from firms and households between 2010-2012 indicate to have been largely effective, because government adapted the 2012 policy such that it benefited mainly these interest groups. First, government proposed to levy the tax at a maximum of 25% of the sale price, while petrol and oil would no longer be levied an environmental fee, and tax rates would be set at the lowest level within each tax range (Viet Nam News, 2010). Second, the new policy exempted inputs frequently used by subsistence farmers, such as chemical fertilizers (Viet Nam News, 2010). Government also limited price increases in fuels and electricity by introducing carbon taxation whilst reducing other taxes (for fuels) and by regulating output prices to households and firms (for electricity). Similar interest pressures from the same interest groups in 2017 also suggest to have been effective, because after pressures about competitiveness and welfare effects to firms and households (Viet Nam News, 2017b), government disapproved the 2017 proposal and when it introduced an adapted policy in 2018 it increased carbon prices only minimally (Van & Tuc, 2018). In terms of public firms, there is evidence of formal but not of informal lobbying, although informal lobbying is unlikely to be covered in news releases. For example, in 2017, the Vietnam Petrol and Oil Association, which in effect represents state-owned enterprises, said to support the tax increase, but plead for lower tax rates (Viet Nam News, 2017c).
This evidence can explain why carbon prices for diesel and coal were minimal when the policy was introduced in 2012 and continued to remain minimal despite increases in 2015 and 2018. That the carbon price on petrol increased, even with interest pressures, can be explained by household lobbying inability. Household indirectly lobbied government through provincial government delegates, which provided comments in the National Assembly, but were unable to directly express interest pressures (Vietnam Plus, 2010). This is consistent with the ex-ante predictions that consumers face large lobbying costs and use institutions and associations to increase lobbying ability. That the tax rate increased substantially for petrol, but only to limited extent for diesel and coal, could then explain why tax revenues increased over time, although disproportional to increases in carbon prices.
Regarding the increase in carbon prices over time, evidence suggests that government was able to increase tax rates without large interest pressures because tax rates were increased strategically during decreases in other taxation. For example, in 2015, to compensate for reductions in public revenues – which were a product of declines in global oil price and reductions in import tariffs – government increased tax rates threefold (Duy, 2015). In effect, however, in terms of petrol prices, decreases in global prices and reduction in import tariffs more than offset the price increase from taxation. In fact, retail prices reduced from around US$ 1 per litre in 2014 to on average US$ 0.8 per litre in 2015 (Trading Economics, 2020). That government managed to increase carbon prices without increasing pressures, and that it did so by balancing increases with decreases in other taxation, may also explain why revenues from carbon taxation increased over time, both in absolute and relative terms.
Evidence from the discourse analysis leaves unexplained why tax revenues increased while the share of revenue recycled towards climate measures decreased over time. In fact, the issue of revenue recycling remained unaddressed by government since the introduction of the policy in 2012, despite recurring interest pressures in 2015, 2017 and 2018. For instance, in 2015, local governments argued that tax revenues did not benefit the environment, indirectly voicing concerns from local consumers and producers (VietNamNet, 2016). In 2017, academics expressed concerns about the effectiveness of the policy to limit emissions (Petrolimex, 2017b; VietNamNet, 2017). In 2018, firm associations lobbied government on more transparent spending of revenues (Viet Nam News, 2018). Lacking any observed change in policy outcome, it suggests that the pressures – largely from academia and civil society, but at some point from all interest groups – were mostly ineffective. The next section examines whether insights from the interviews can help explain the unaddressed variance in carbon taxation.
Structured Interviews
In this section, I use insights from the interview questionnaire to verify insights from the discourse analysis and the ex-ante predictions regarding public and private pressures, and examine whether and how interest pressures resulted into policy change. Interview references are found in online Appendix D.
In terms of differences in carbon prices across fossil fuels and that those differences persisted over time, the results suggest that the limited lobbying ability and willingness of households can explain to large extent the relatively large increase in carbon price for petrol. On the one hand, households experienced increasingly environmental effects, such as air pollution, and therefore accepted carbon price increases on petrol (A1, H9). Generally, households accepted the policy because they understood that government needed to raise funds for broader economic development purposes. On the other hand, government used the lobbying inability of households, due to limited lobbying means, to increase the carbon price of petrol without creating interest pressures (A2, H1). Many households expressed to want to decrease the carbon price on petrol, but that the “residential groups”, which officially represent households, were only effective in bringing pressing issues to the attention of government and rather functioned as a top-down mechanism to inform households about new policy, although I do not find that such groups informed households about the carbon tax (H2-5, H7-8, H10, H13-14, A2-3, C1). This finding – that households have limited lobbying ability – is in line with the ex-ante predictions, although I find that lobbying inability is not a product of coordination or free-riding problems, as suggested by ex-ante predictions and in theory (Olson, 1965). Instead, limited ability is the product of inefficient representative institutions. In line with ex-ante predictions, the results provide evidence that exposure to the tax differed across income groups, although I find that it did not translate into different interest pressures because all income groups had limited lobbying ability and information about the policy.
The limited increases in carbon price for diesel and coal can be largely explained by the large and effective pressures by both private and public firms. Results suggest that private firms lobbied government effectively, both directly and through associations, using both formal and informal pressures, and working together among firms and associations to increase effectiveness (A3, F1-2). This finding may explain the limited increases in carbon price for diesel over time. In line with the ex-ante predictions, I find that firms in sectors with relatively high expenditure on fuel tended to lobby government more than those with lower expenditure, although I am unable to identify that the food, mineral, transport and agriculture sector provided largest pressures, or that the metal and services sector used informal lobbying most (F1, F3, F16, F18).
In line with the ex-ante predictions, public firms also expressed large and effective pressures regarding the policy and did so mainly using private connections to ministries and politicians. This may explain the low and non-increasing carbon price on coal. Actually, the evidence suggests that private firms more often used informal channels than SOEs, in part because public firms could position members of management in leading positions in government and thereby influence policy, while private firms could not (A3, F9, F11, C1, G2). Further, SOEs in coal needed to provide limited pressures because government already preferred to support coal to promote economic development (A3, G3, I3).
Academia and civil society expressed large pressures to increase the carbon prices on coal but were largely ineffective (C2-3, A3, I4). Lobbying ability among academia and civil society was limited in part because they went for advocacy rather than lobbying and in part because they had limited lobbying possibilities and resources (A4, A2, G2, I3, I2, G4, I1, C2-3). Evidence shows that international organizations also advocated for increases in the carbon price (for coal), but were ineffective in that they focused on technical assistance (I1-6). These observations – of the limited ability and willingness of household lobbying, the effective lobbying by firms, and the ineffective lobbing from academia, civil society and international organizations – can also explain why tax revenues increased over time, both in absolute and relative terms, but only disproportionally to increases in carbon prices.
In terms of the reduction in share the of revenue recycling towards environmental purposes, the results show that all interest groups preferred more transparency about revenue recycling and demanded a larger share of revenues to be directed towards environmental purposes (A4, H2, H4, H6, H13, I5, C1, C3, F3-4, F6, F12, F14, F15, F17-18). However, lobbying was not the main driver of changes in policy. Instead, policy changed, first, as tariff reductions from trade liberalizations urged government to find alternative sources of public revenues. Secondly, and in line with the ex-ante predictions, most interest groups agreed with government that economic development was superior to environmental development, which in turn resulting in reduced interest pressures (F2, F5, F8, F12, A4, C1, I2, I5, H7). Thirdly, the general law on budget management did not allow for earmarking, meaning that although revenues were intended for climate adaptation and mitigation, they were used within the general budget (A1).
The evidence also suggests that limited lobbying regulation reduced lobbying effectiveness for some groups (A2, F7, F9, G7, H3). For households, academia and civil society, limited regulation implied limited means to lobby government, while among private firms it incentivized informal lobbying, in part because public lobbying could lead to harassment (A4, F4, H7). While there exist formal lobbying procedures by which interest groups are consulted, in many cases they do not include all interest groups and do not always translate pressures well to government (G5, F1).
In sum, the discourse analysis and the representative surveys provide additional and contrasting evidence regarding lobbying inability of households, academia and civil society. It shows that coordination issues or free-riding are not the reason that households face large lobbying costs. Instead, household lobbying is limited due to inefficient representative institutions; “residential groups”, which should allow households to lobby and receive information about the policy, were ineffective. On the other hand, private and public firms strongly influenced government, and were effective in changing policy, as they had direct formal and informal connections to politicians, with many respondents describing that – in absence of formal lobbing regulations and without efficient firm associations – frequent interactions with politicians and connections to ministries were main ways to influence policy.
Measuring Interest Pressures
This section evaluates whether the proposed combination of methods is useful to measure interest pressures and its effects. I find that in absence of data on political contributions, lobbying costs or charitable giving, the given methods are a manageable empirical strategy to approximate interest pressures. The results also suggest that deploying methods individually might not capture interest pressures and their effects as well and accurately as a combination of methods. For example, I find that the ex-ante proxies – which used preference, consumption, lobbying-cost and corruption data – provide useful predictions of economic interests across different stratifications of consumers and producers and that findings from the discourse analysis and structured interviews can confirm most predictions. The proxies are less useful, however, to examine whether and how these differences in exposure or potential demand for lobbying translated into interest pressures, and subsequently, how pressures affected changes in policy.
The discourse analysis and structured interviews can address these issues. In particular, they provide context to the ex-ante predictions and importantly, provide insights why interest pressures, which in theory should have materialized, might not materialize. For example, interview evidence was useful in showing that despite high demand for lobbying by some households, pressures did not materialize, not because of coordination problems, but because of limited lobbying means. The qualitative approaches in some cases allow to measure pressures more precisely than the proxies, such as when distinguishing between different lobbying activities and its effects (see Section 2). Importantly, qualitative evidence can address to some extent the challenge of causal identification and can explain whether interest groups support politicians because they share similar preferences or because of influencing policy.
A general caveat of both methods is that while they allow to measure interest pressures, they are (i) not suited for statistical inference, (ii) unable to obtain pecuniary returns on outcomes, and (iii) impractical to estimate the return to interest pressures. While there are several issues that might cause bias from endogeneity in the interest proxies, or bias from validity threats in the interviews, there are respectively robustness checks and questionnaire guidelines that can reduce such biases (see Section 2). Generally, the proposed methods are beneficial because of their easy replicability and low computational burden, as they use publically accessible survey data for the interest proxies and offer a standardized questionnaire for straightforward and low-cost data collection.
[1] Qualitative evidence confirms this and shows that government perceives climate change as secondary to economic growth (Fortier, 2010, p. 234) and that carbon taxation should have limited welfare effects and be progressive across income groups (Haughton, Quan, & Bao, 2006, p. 230).
[2] By 2016, the share of hydropower had been reduced to 38% of total electricity generation (EVN, 2017, p. 12), in effect increasing the pass-through effect of carbon taxation on electricity prices.
[3] Household budget shares have shown to approximate well variation in consumer good price changes due to policy shocks (Deaton, 1989a, 1989b). It is important to note that price effects, here derived from expenditure indices, would be more accurately estimated by using longitudinal data to derive demand elasticities and then define household demand functions, but available statistics are insufficient for such an exercise. It is quite likely, for instance, that even though fuel constitutes a relatively large household expenditure, it is by comparison price inelastic, thus making it desirable to tax fuel more heavily than other (more elastic) expenditures (Deaton, 1984, p. 99). However, for the purpose of this paper, expenditure indices are sufficient because, from a political economy perspective, at first what is most relevant is how consumers and producers perceive, rather than experience, the effects of the incoming tax.
[4] Other data suggest that household expenditure on foods are higher, in the range of 35-50% of total expenditure (Anh, Thang, & Phuong, 2013, p. 39), suggesting larger price effects for poorer households.
[5] I cannot examine whether these findings are robust to lagged data (and thus evaluate whether results are biased by endogeneity) because older households survey data are not publically available.
[6] I use an independent t-test between small (<19 employees) and larger firms (> 18 employees), and an independent t-test between private (0% state-ownership) and non-private firms (> 0% state-ownership). This implies that – assuming that demand elasticities are similar across firm sector, size and ownership – carbon taxation would have had different effects across sectors because of different expenditure shares, but not across firms with different sizes or types of ownership
[7] In 2016, 75% of electricity generation, 100% of electricity transmission and 100% of electricity distribution was undertaken by SOEs (H. Nguyen, 2018, p. 129).
[8] Note that this data should be considered with care because many firms in the sample were unable or unwilling to report on the measure (see footnote Table 6).
[9] I perform an independent t-test between small (<19 employees) and larger firms (> 18 employees) and do not find that differences of means are statistically different from zero for informal payments, but find that this is the case for the share of SOEs.