H IDDEN DEBT
FIGURE 3.12 Total Liabilities of Loss-Making State-Owned Enterprises in India, Pakistan, and Sri Lanka, 2005–17 20
18
16 Percent of GDP
112
12
12
12
15
14
8 4 1
2
2
2
2
2006
2007
2008
2009
5
4
4
4
2010
2011
2012
2013
5
3
3
2014
2015
4
5
5 3
0 2005
India CPSEs
Pakistan
2016
2017
Sri Lanka
Source: Melecky, Sharma, and Yang 2020 (see details in annex 3A). Note: CPSEs = central public sector enterprises.
Indian CPSEs are not engaged in inherently more risky activities than private firms. Overall, the results in table 3B.3 show that CPSEs do not have significantly more volatile sales or profits than comparable non-SOEs.15 Adjusting for size is critical: even though the raw volatility of CPSE sales is significantly lower (column 1), this difference disappears when we include size as a control. Although individual SOEs do not face more volatile conditions than individual nonSOEs, the SOE sector as a whole could be more volatile because the shocks hitting SOEs are more correlated. But this hypothesis also is not confirmed: The volatility of aggregate CPSE sales has been similar to the aggregate volatility of total sales of private firms in the Prowess database in recent years.
Why Do SOEs Underperform Comparable Private Firms? Using the same basic regression specification, we next show that SOEs commercially underperform otherwise comparable private sector firms. We regress indicators of performance on the CPSE dummy and controls, such as size, age, and sector-year fixed effects. The results are shown in table 3B.4.
Indian CPSEs overemploy labor and capital. Controlling for size, age, and sector, the revenue-to-wage bill ratio for CPSEs is 85.8 log points lower and their revenue-to-fixedassets ratio is 21.5 log points lower (columns 2 and 4, respectively).16 Thus, CPSEs earn less per unit labor cost and per unit capital than comparable private firms. CPSEs also have a higher debt-to-asset ratio than comparable non-SOEs (column 6). We further compare CPSEs with other firms in terms of revenue-based productivity measures: revenue total factor productivity (TFPR); and the marginal revenue products of capital (MRPK), labor (MRPL), and material inputs (MRPM). The estimates of TFPR, MRPK, MRPL, and MRPM are based on the procedure outlined in Asker et al. (2014), which is based on a model in which firms produce differentiated products using a simple (industry-specific), constant-return CobbDouglas production function and face a demand curve that is constantly elastic. The details of the estimation are presented in annex 3C. TFPR measures sales per unit inputs and should not be equated with physical total factor productivity (TFP). Differences in TFPR across firms could reflect distortions such as input adjustment costs, markups, and