We now turn to our central question: what are the drivers of upgrading? I categorize drivers into three groups, which can be understood with reference to the general framework above: (1) output-side drivers: factors that affect product demand curves (the D(·) functions); (2) input-side drivers: factors that affect input-supply curves (the S(·) functions); and (3) drivers of capabilities: factors that affect the the “know-how” of firms (the Λijkt, Jit, and Kijkt). This categorization is necessarily somewhat loose - some drivers fit in more than one category, and some not quite in any - but the grouping is helpful to organize the review.
3.2 Input-Side Drivers
We turn now to drivers on the input side, beginning with factors influencing imports of inputs and then considering factors that influence the prices and availability of domestic inputs.
3.2.1 Imported Inputs
Above we observed that firms in developing countries appear on average to sell higher-quality varieties on international markets than on domestic markets. It also appears that firms tend to buy higherquality inputs on international markets than on domestic ones. In Colombian data, for instance, Kugler and Verhoogen (2009) document that plants systematically pay higher prices for imported inputs, controlling for detailed product fixed effects.[1] One possible explanation is that there is a home-market effect in the production of quality, such that firms in richer countries specialize in producing higher-quality goods to appeal to richer local consumers, as for instance modeled in Fajgelbaum et al. (2011).
If inputs available on the international market tend to be of higher quality than domestic inputs (or have lower quality-adjusted prices for higher-quality varieties), then we would expect a reduction of tariffs on inputs to lead developing-country firms to upgrade the quality of their inputs. Given the fact (discussed above) that high-quality inputs appear to be a key requirement for high-quality outputs, we would expect to see upgrading on the output side as well. Bas and Strauss-Kahn (2015) provide evidence for this mechanism in Chinese trade-transactions data. Comparing processing firms (which are exempt from tariffs) to ordinary firms, constructing firm-specific tariff reductions based on firms’ import mixes, and controlling for firm-product (and a number of other) fixed effects, they find that tariff reductions lead Chinese firms to increase the prices they pay for inputs and to increase the prices they charge for outputs, consistent with a quality story. The results are primarily driven by firms that import most of their inputs from, and export most of their outputs to, developed countries. Results are similar if they use the Khandelwal et al. (2013) methodology to construct measures of input and output quality. A roughly contemporaneous paper by Fan et al. (2015) also finds that Chinese firms responded to reduced tariffs on imported inputs by raising export prices and quality, and that this effect is stronger in more differentiated sectors. (See also Feng et al. (2016) and Abeberese (2016).) An obvious limitation of trade-transactions data is that they include only international transactions, which may not be representative. However, Bas and Paunov (2019) find broadly similar results with representative data from Ecuador (plant census and customs data), and also find that the importedinput-driven upgrading is associated with increases in skill intensity.
In an interesting extension of this line of work, Fieler et al. (2018) argue that there is an amplification effect in upgrading: tariff reductions on inputs lead firms to upgrade the quality of outputs, which in turn increases their demand for other high-quality inputs, which gives incentives for local suppliers to upgrade, which gives local final-good producers further incentives to upgrade. Empirically, the authors calibrate their model to pre-liberalization data and do counterfactual simulations. Now that datasets with firm-to-firm links are becoming available, a promising line of research would be to investigate this sort of mechanism in a less theory-dependent way.
Tariff reductions not only improve access to high-quality imported inputs, they also expand the variety of inputs available, which may in turn enable firms to produce new outputs. Focusing on India’s liberalization in the early 1990s, Goldberg et al. (2010) provide evidence that the increased availability of imported inputs led firms to expand their set of output varieties. They document a reduced-form relationship between import tariff reductions and product innovation and impose a simple theoretical structure to separate the price and variety effects of the tariff reductions, finding that a substantial share of the increase in product scope is driven by the expansion of imported input variety. Although Goldberg et al. (2010) do not have access to information on inputs at the firm level, Bas and Paunov (2019) directly observe both inputs and outputs of Ecuadorean firms and confirm the findings that import tariff reductions lead firms to use more inputs and expand product scope.
There also appears to be a robust causal relationship at the firm level between reductions of tariffs on imported inputs and increases in standard measures of revenue TFP. This relationship has been documented for instance by Schor (2004) in Brazil, Amiti and Konings (2007) in Indonesia, Topalova and Khandelwal (2011) and Nataraj (2011) in India, Brandt et al. (2017, 2019) in China, and Bas and Paunov (2019) in Ecuador.[2] In a recent review, Shu and Steinwender (2019) observe that papers that have considered tariffs on outputs and inputs separately have tended to find stronger effects of input-tariff reductions than of output-tariff reductions, and I share their view. (Refer to Section 3.1.3 above.)
At the same time, a recurrent question in the literature on imports and productivity is to what extent the results reflect changes in markups or some other source of bias in measured TFP, for instance quality. In an influential contribution, De Loecker et al. (2016) develop a methodology to tease apart the contributions of technical efficiency, markups, and quality in multi-product firms. At the core of the exercise is a formula for calculating markups at the firm-product level, discussed in the context of Garcia-Marin and Voigtländer (2019) in Section 3.1.1.3 above. The formula requires information on input expenditures as a share of output revenues and on output elasticities at the product level. The authors’ strategy is to focus on single-product firms, where the mapping from inputs to outputs is clear, and to do a selection correction to address the fact that single-product firms may not be representative. In the output-elasticity estimation, which follows Ackerberg et al. (2015), the authors put physical output on the left-hand side and use output prices to proxy for input prices and input quality to address potential input-quality bias. They find that import tariff reductions cause a reduction of marginal cost that is only partially passed through to consumers. That is, product prices decline, but by less than marginal costs decline, and hence markups rise. This suggests that the estimated effects of import tariff reductions on standard TFP measures - which incorporate both technical efficiency and markups - overstate the true effect on technical efficiency. Quibbles can be raised about the output-elasticity estimation (which is subject to the identification concerns raised by Gandhi et al. (forthcoming)) and about whether putting physical output on the left-hand side in the production-function estimation adequately addresses the possibility of output-quality bias. But it is clear that this paper is an important contribution and has become a key point of reference for the literature.
3.2.2 Domestic Inputs
Several papers have investigated how changes in the cost of labor, capital, or other inputs on the domestic market affect firms’ upgrading decisions. Supply shocks of workers of different skill levels are one possible driver. Some of the best work on this topic is from the US: using a shift-share instrument for immigration, Lewis (2011) shows that US manufacturing firms in regions with greater inflows of low-skilled migrants were less likely to adopt advanced technologies, and Hornbeck and Naidu (2014) show that greater outflows of low-skilled workers from the US South, in response to a major flood in 1927, led farms to increase mechanization.[3] In a similar vein in a developing-country context, Imbert et al. (2019) use agricultural price shocks combined with historical migration patterns in China as a source of exogenous inflows of low-skilled migrants to urban areas. Firms in areas that receive more low-skilled migrants are less likely to file domestic patents and tend to shift toward products with low human-capital intensity (defined as the average share of the workforce with a high-school degree among firms that produce a given product).[4]
Two recent papers using city-level minimum-wage variation in China provide evidence that minimum wage regulations, which raise the relative cost of less-skilled labor (in addition to raising wage costs overall), can have effects similar to an increase in relative supply of more-skilled labor. Mayneris et al. (2018) find that firms more exposed to the minimum-wage hikes (in particular, those whose average wage in the previous year was below the new minimum wage) saw increases in productivity relative to less-exposed firms. Hau et al. (forthcoming) also find that firms more affected by minimum wage changes (in the sense that their average wages are closer to the minimum) tended to see increases in measured TFP and shifted to more capital-intensive production, with some heterogeneity based on firm characteristics. The usual caveats about TFP estimation apply, but broadly these papers suggest that higher wages overall (which induce firms to substitute capital for labor) and/or higher relative costs of low-skilled workers (which induce firms to substitute high-skilled for low-skilled labor) can lead firms to upgrade.[5]
The literature on access to capital as a driver of upgrading in larger firms in developing countries remains thin and somewhat mixed. There have been influential studies of the effect of capital in microenterprises.[6] There have also been careful studies of the effects of capital-supply shocks on other (i.e. non-upgrading) outcomes among larger firms, in both developed and developing countries (e.g. output: Banerjee and Duflo (2014); use of alternative credit sources and financial distress: Khwaja and Mian (2008); exports: Amiti and Weinstein (2011), Zia (2008), Paravisini et al. (2014), Kapoor et al. (2017); employment: Chodorow-Reich (2014), Brown and Earle (2017)). But there have been relatively few studies linking credit shocks directly to firm-level productivity, quality, technology adoption, or other upgrading outcomes among larger developing-country manufacturing firms.[7]
Perhaps surprisingly, the few papers that have focused on the effect of increased capital supply on productivity have largely failed to find evidence of such an effect. Bau and Matray (2019) examine the effect of a policy reform in India that removed some restrictions on foreign investment, arguably increasing the supply of capital, in a staggered way across industries. They primarily focus on misallocation, but they also estimate the impact of the reform on TFPR, and find no evidence of an effect. They caution that they also find a decline in product prices, likely reflecting reduced capital costs, and that the price decline may in part be responsible for the lack of an observed effect on TFPR. Also in India, Rotemberg (forthcoming) examines the effects of a 2006 broadening of the set of firms in India eligible for subsidies to small and medium-sized businesses, similar to an earlier change studied by Banerjee and Duflo (2014). The affected firms became eligible for a range of programs, but the most important (70% of the budget for such programs) appears to have been subsidized credit. Rotemberg focuses primarily on quantifying simultaneously the direct and indirect effects of the subsidies and their contributions to aggregate productivity, but he also examines direct effects of the subsidies on firm-level TFPQ and finds no evidence of an effect. Cai and Harrison (forthcoming) study a reform in China that reduced the value-added tax (VAT) on investment goods, with the goal of encouraging technology adoption. They find an increase in capital intensity but no effects on fixed investment, product introductions, or productivity.[8] Arráiz et al. (2014) study the effect of a Colombian government loan-guarantee fund, using a propensity-score matching estimator with fixed effects, and find impacts on output and employment but not investment, productivity, or wages. By contrast, Eslava et al. (2012), also using a combination of matching techniques and fixed-effect estimators, find that loans from a publicly owned development bank to Colombian manufacturing firms generated significant positive effects on productivity as well as output, employment, and investment.
Energy inputs are often measured reasonably well in manufacturing surveys in developing countries, and a small literature has investigated the role of shocks to energy supply or prices on firm-level upgrading outcomes. Abeberese (2012, 2017) examines the relationship between electricity prices and various dimensions of firm behavior, using arguably exogenous variation in coal prices interacted with the initial share of thermal generation (which uses coal) in states’ electricity generation. She finds that higher electricity prices induce firms to shift their product mix toward products that are on average produced by firms that use less electricity. Although specific technologies are not observed in the Indian data, it seems plausible that less electricity-intensive processes are also less technologically advanced. She also finds a negative (although not significant) relationship between electricity prices and the level of productivity, and a negative and significant relationship between electricity prices and the growth rate of productivity.[9] A subsequent paper by Allcott et al. (2016) pursues a related strategy. Using rainfall at higher elevations (which determines hydro-electric power generation capacity) as an instrument for shortages (rather than electricity prices) in India, they find that shortages lead firms to contract in terms of both sales and input purchases but they do not find a significant effect on TFPR. Simulations suggest that there is more of a negative effect for firms that do not already have generators, which are smaller on average.[10]
[1] Importing plants also pay more on average for their inputs than non-importing plants, even for domestic inputs, consistent with the ideas that there are fixed costs of importing and that more-capable plants use imported inputs, which tend to be higher-quality, to produce higher-quality products. See also Blaum et al. (2019).
[2] See also the studies by Tybout and Westbrook (1995), Lopez Cordova (2003), Kasahara and Rodrigue (2008) and Halpern et al. (2015), which find positive contributions of imported inputs to productivity. An exception is Muendler (2004), which finds that imported inputs make only a minor contribution to productivity, if any.
[3] See also Clemens et al. (2018) and San (2020).
[4] Related work by Bustos et al. (2019), with data at a regional level in Brazil, suggests that such shifts into low-skillintensive manufacturing may have lock-in effects with negative growth consequences in the long run.
[5] To be clear, although higher minimum wages appear to have spurred upgrading in these cases, they are likely to have reduced profits for individual firms. The point from Section 2.1 that upgrading may or may not be profit-maximizing is worth recalling here.
[6] See e.g. de Mel et al. (2008), McKenzie (2017), and the reviews by Banerjee et al. (2015), Woodruff (2018) and Quinn and Woodruff (forthcoming).
[7] There are small literatures on credit constraints and technology adoption in agriculture (see e.g. Giné and Klonner (2008) and the review in Jack (2013, Section 5)) and households (see e.g. Berkouwer and Dean (2019)).
[8] Liu and Lu (2015) find an effect of the same reform on exports by Chinese firms.
[9] In related work in Chinese data, Fisher-Vanden et al. (2015) find that firms respond to higher electricity prices by outsourcing more inputs; at the same time, they find muted effects on productivity. Related contributions not focused on firm-level upgrading outcomes include Rud (2012) and Cole et al. (2018).
[10] Relatedly, Abeberese et al. (forthcoming) find negative impacts of outages on productivity among small and mediumsized Ghanaian firms (see also Hardy and McCasland (forthcoming), which focuses on microenterprises) and Ryan (2019) finds that randomized energy audits in Indian manufacturing firms, which appear to have increased energy efficiency, led firms to expand their use of energy. In related work on the role of infrastructure, Hjort and Poulsen (2019) examine the reduced-form relationship between the arrival of fast internet and skill upgrading in Africa, but also presents evidence that fast internet led to productivity improvements in Ethiopia (as well as increases in exports from several countries.)
