A. Evolución económica del impuesto predial 2009-
5.1.3 Estimación del modelo
Assume that some selection bias is generated by enterprise-specific time-invariant charac- teristics and other time-invariant factors (e.g., region, industry). To estimate the effect of privatization on performance, I use the following baseline fixed effects model specification:
yit =α1P SRit+α2F SRit+βlnHHIjt+γXit−1+θi+δt+µkt+ϵit (2.1)
wherei,j,k, andtrepresents SOEiin industryj, located in regionkin yeartrespectively. The outcome variableyit includes a list of performance variables of interest: productivity
indicators, profitability indicators, innovation indicator, operation scale indicators, and profitability sources indicators.
Profitability indicators include lnP rof it_Lit, lnP rof it_Salesit, and
lnP rof it_Assetsit. I use operating profit in all variables. 6 lnP rof it_Lit is profit
per employee. lnP rof it_Salesit is profit per unit of sales revenue. lnP rof it_Assetsit is
profit per unit of total assets. All values are in log.
6By definition, operating profit equals the sum of sales profit of main products and profit of other business,
minus administrative expenses and financial expenses. Sales profit equals sales revenue minus costs of good sold, sales expenses, and sales tax.
Productivity indicators include lnV A_Lit, lnV A_Kit, and lnSales_Lit. lnV A_Lit is
value added per employee, which represents labor productivity. lnV A_Kit is value added
per unit of capital, which reflects capital productivity.7 lnSales_L
it is sales revenue per
employee. All values are in log.
Innovation indicator is lnN ew_Salesit, which is new product output value per unit of
sales revenue. Another potential indicator of innovation is R&D expenses per unit of sales revenue. But data on R&D expenses is not available in most of the sample years in the CIED data. Thus,lnN ew_Salesitis used to represent innovation. The value is in log.
Operation scale indicators include lnAssetsit, lnSalesit, and lnEmpit. lnAssetsit is
total assets, which represents enterprise size or scale. lnSalesit is sales revenue. lnEmpit
is the number of employees. All values are in log.
Profitability sources indicators include lnSalesP rof it_Salesit and
lnOtherP rof it_Salesit (two sources of profit), and lnAdminExp_Salesit and
lnF inExp_Salesit (two sources of cost). By the definition of operating profit, these
four components are the four sources of profitability. Specifically,lnSalesP rof it_Salesit
is sales profit of main products per unit of sales revenue.lnOtherP rof it_Salesitis profit of
other business per unit of sales revenue. lnAdminExp_Salesit is administrative expenses 7Here, capital is measured by net fixed assets of an enterprise.
per unit of sales revenue.lnF inExp_Salesitis financial expenses per unit of sales revenue.
All values are in log.
On the right hand side, the main independent variables of interest are the privately- owned share of capital (P SRit) and foreign-owned shared of capital (F SRit). In China,
the components of an enterprise’s total capital can include state-owned capital, collectively- owned capital, legal-person-owned capital, privately-owned capital, Hong Kong, Macao, and Taiwan (HMT)-owned capital, and foreign-owned capital. Total capital is the sum of these components.P SRit, which represents private share, is calculated as the percentage of
private-owned capital in total capital. It reflects the private control right of the enterprise.
F SRit, which represents foreign share, is calculated as the percentage of foreign-owned
capital in total capital.
Other covariates includelnHHIjt, which is the Herfindahl-Hirschman index, indicating
the level of market concentration or competition in industryj in yeart. It is calculated by taking the sum of squared market share of sales of each enterprise operating in a 3-digit industry category. lnHHIjt can range from close to 0 (pefect competition) to 10,000
(monopoly). HigherlnHHIjt means that the market of industry j in yeartis closer to be-
ing a monopoly, and that market is more concentrated (less competition). The value is in log.
A list of lagged enterprise characteristics Xit−1 is also incorporated. It includes indicators of an enterprise’s access to finance: lnSubsidy_Lit−1, subsidy per employee
(in log), which reflects an enterprise’s access to direct government financial assistance;
D_Ait−1, debt-to-asset ratio, which reflects an enterprise’s direct access to borrowing; and Sof tBudgetit−1, the interaction term of lnSubsidy_Lit−1 and D_Ait−1. These three variables together captures the softness of an enterprise’s budget constraint (see Footnote 5). In addition, I include expshit−1, the export share of an SOE, which is calculated as the share of export sales in total sales. It captures an enterprise’s export propensity. Since these enterprise characteristics are one-year lagged variables in yeart−1, they will not be affected reversely by the privatization decision or enterprise performance in yeart.
I also include a list of fixed effects. θiis firm fixed effects, which controls for unobserved
time-invariant enterprise characteristics that can affect enterprise performance. δt is year
fixed effects, which controls for changes in external environment over time that can affect performance and are common for all enterprises, such as macroeconomic fluctuations and changes in government policy over time. µkt is region-year fixed effects, which is the
grouped dummy (interaction) of regionk and year t. Region k is a dummy for the 3,555 counties where enterprises are located.
2.3.2.2 Binary Privatization Indicator
The above specification (Equation (2.1)) can estimate the intensity of privatization (measured by the continuousP SRit) on performance. Alternatively, I use the following specification
yit =α1P rivit+βlnHHIjt +γXit−1+θi+δt+µkt+ϵit (2.2)
where P rivit is an indicator for whether an SOE is privatized. It equals 1 if SOE i is
privatized in year t (or has been privatized in earlier years). All the other independent variables and outcome variables are the same as in Equation (2.1).