Application of Accounting Software and Enterprise Productivity——Evidence from Unlisted Companies

doi:10.21203/rs.3.rs-2823216/v1

Download PDF

Research Article

Application of Accounting Software and Enterprise Productivity——Evidence from Unlisted Companies

https://doi.org/10.21203/rs.3.rs-2823216/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

The existing research on the economic consequences of accounting informatization mainly focuses on enterprise resource planning (ERP) systems, but in reality, most companies are still using relatively simple accounting software. This paper empirically analyzes the application of accounting software in unlisted companies and its impact on enterprise productivity for the first time with a large sample of tax return information. We found that 75% of unlisted companies use accounting software of varied functions, but only 6% of them use ERP systems. With other factors unchanged, it is found that the productivity of enterprises using accounting software is significantly higher than that of enterprises not using accounting software, and the performance of accounting software is significantly positively correlated with productivity. This correlation is reinforced by the ability of enterprises to utilize accounting information and their growth opportunities. This paper adopts methods such as the Granger causality test and instrumental variable regression to alleviate endogeneity problems. This paper will contribute to a more comprehensive understanding of the economic consequences of accounting information systems and provide an important decision-making reference for promoting high-quality economic development.

Accounting information system

Enterprise productivity

Unlisted companies

ERP

Management accounting

The Guidance of the Ministry of Finance on Comprehensively Promoting China's Accounting Informatization (CK [2009] No. 6) was issued in 2009, which was nearly ten years ago. Accounting informatization helps improve the quality of accounting information and the level of accounting management, thereby improving the decision-making effect of managers and the efficiency of enterprises. However, the current empirical research on the application of accounting information systems and their effect on enterprise performance is not enough for us to fully understand the economic consequences of enterprise accounting informatization.

First, limited by the availability of data, existing research mainly examines the impact of ERP software, while ignoring other accounting software. However, due to high ERP investment costs and high requirements for the internal management environment of enterprises (Zhang Zhe et al., 2005), only a few enterprises adopt ERP. The accounting informatization of most small and medium enterprises (SMEs) is reflected by the adoption of relatively simple accounting software. Therefore, research conclusions based on ERP are not applicable to most enterprises.

Second, previous large-sample empirical studies mainly focused on listed companies. However, unlisted companies are the main body taking up economic activities, which is crucial to economic growth and employment stability①. Moreover, there are systematic differences between listed companies and unlisted companies, and the existing conclusions based on listed companies may not be applicable to unlisted companies.

Third, although the existing empirical analysis based on the application of ERP in listed companies in China is conducive, there are some limitations since existing research relies on voluntary disclosures of ERP adoption by listed companies. However, the research of Zhang Ruijun et al. (2008) showed that the ERP information disclosure of Chinese listed companies is highly arbitrary and lacks longitudinal and horizontal comparability, which reduces the reliability of research data and causes a serious problem of selective disclosure bias.

This paper empirically analyzes the application of accounting software in unlisted companies and its impact on productivity for the first time using the data of the Annual Corporate Income Tax Return submitted to the tax bureau by tens of thousands of enterprises above the designated size in a Chinese province. The data in this paper include the descriptions of the accounting software used by these unlisted companies, which are diverse and not limited to ERP. As the reporting of this information is a mandatory requirement of the tax bureau for enterprises, it greatly alleviates the problem of voluntary disclosure bias and enhances the reliability of research data.

Based on the observations of 58,359 unlisted companies in the province in the years 2014 and 2015, we found that about 75% of the companies used accounting software of different types and functions, but only about 6% of them used ERP systems. This means that ERP-based research conclusions do not apply to most companies. The regression results show that there is a significant positive relationship between the adoption of accounting software and the total factor productivity (TFP) of enterprises; the more advanced accounting software (produced by more famous manufacturers with more complete functions), the higher the impact on enterprise productivity. The above conclusions are both statistically significant and economically significant and still hold under a series of robustness tests. In order to alleviate possible endogeneity problems, this paper adopts propensity score matching, the Granger causality test, and instrumental variable regression, and the conclusions remain unchanged. Further analysis finds that the effect of accounting software is more significant in companies with a higher ability to use accounting information and greater business growth opportunities.

This paper advances existing research in the following three aspects: First, this paper comprehensively examined the impact of different types of accounting software on enterprise productivity. Previous studies have neglected non-ERP accounting software and rarely compared the differences in the effects of accounting software with varying levels or qualities. Due to the significant differences between ERP and non-ERP software, the findings of this paper support the positive economic consequences of accounting informatization in a more general sense. Second, this paper examines for the first time a large sample of unlisted companies' accounting software usage and its effects. Given the large number and importance of unlisted companies, this paper plays a significant gap-filling role. Third, the data advantage makes the conclusion of this paper free from the noise and endogeneity problems caused by voluntary disclosure of enterprises.

① In the initial sample of this paper, unlisted companies account for 99.57% of the total number of enterprises, contributing 92.77% of the operating income, 91.47% of the added value, and 96.31% of the employment of all enterprises.

I) Literature review

Driven by the trend of accounting informatization, most companies today have adopted computer software for accounting and analysis. Accounting software, especially ERP, can bring high costs and benefits to enterprises, so it is necessary to understand their functions and requirements for enterprises. Since the 1990s, researchers have conducted a series of empirical studies on how enterprise information management systems (mainly ERP) affect enterprise performance.

Several studies have examined whether systems can improve the perceived performance of managers. For example, Shang and Seddon (2002) found that managers believed that the system implementation improved enterprise resource management, quality of enterprise planning, and operational efficiency. According to the questionnaire survey by Spathis (2006), managers believed that information systems can bring about "more flexible information generation", "higher quality report", "shorter reporting time", "more timely and reliable accounting information for better decision-making", and so on. Chapman and Kihn (2009) found through a questionnaire survey that the integration of financial information and non-financial information can better support management control elements, so that managers perceive system effectiveness and performance improvement.

Some other studies examined the impact of the system on the objective performance indicators of enterprises, which are mainly financial indicators, including the return on sales (ROS), return on assets (ROA), and asset turnover. Empirical research on foreign listed companies generally found that the existence and implementation of information management systems can improve the actual performance of enterprises. For example, taking a sample of 54 US-listed companies that publicly disclosed ERP implementation information, Poston and Grabski (2001) found that the unit cost of sales decreased after three years of ERP adoption. Hitt et al. (2002) found that the adoption of SAP by US-listed companies has a significant positive relationship with their per capita sales, ROA, asset turnover, enterprise productivity, and other performance indicators. Nicolaou (2004) found that the implementation of ERP in US-listed companies improved their financial performance in the long run. Hen-dricks et al. (2007) found that the adoption of ERP improved the ROA of enterprises to a certain extent.

However, studies on Chinese listed companies were not consistent, although some studies found that the existence or implementation of ERP can improve the financial performance indicators of the companies (Wang Liyan and Zhang Jidong, 2007; Zeng Jianguang et al., 2012), more studies did not find that ERP implementation has a significant effect on financial performance, and even found that enterprise performance declines after ERP implementation (Rao Yanchao, 2005; Zhao Quanwu et al., 2008; Zheng Chengde et al., 2008; Xu Wenjuan and Ou Peiyu, 2011; Lou Runping and Xue Shengjia, 2011). This inconsistency may be due to the small differences in the adoption of ERP by listed companies, or the opaque and incomparable disclosure of ERP usage by Chinese listed companies.

Finally, to the best of our knowledge, previous large-sample empirical studies did not focus on accounting software other than ERP, nor did they take unlisted companies as research objects.

II) Hypotheses

One of the core accounting functions is to record and reflect the economic transactions of enterprises according to certain rules, so as to provide decision support for accounting information users. Simon's (1955) bounded rationality holds that human cognition and information processing capabilities are limited, and only a small information subset can be used in decision-making. Therefore, one of the important sources of accounting information usefulness in decision-making lies in its generalization: each stage of processing from massive original vouchers to account books and then to reports makes economic activity information of enterprises more generalized, and these generalized accounting figures can provide key information to decision-makers more quickly and intuitively. However, the process of accounting information generalization also leads to information loss②. In addition, general-purpose financial reporting information has a single set of rules and cannot provide more detailed and multi-dimensional information required for managerial decision-making. Enterprises are likely to flinch from the arduous work of re-aggregating and generalizing information in more dimensions, and eventually give up using this information.

Accounting information systems effectively alleviate the above-mentioned problems. After entering original information into the database, users are allowed to classify, extract, and analyze the information with varied rules, which improves the utilization of the original accounting information. ERP also connects enterprise financial data with a large number of non-financial business data (such as order status, inventory, and customer information) in real time, enabling managers to extract useful information for decision-making from a comprehensive perspective by considering all business activities and the entire business cycle. Therefore, accounting information systems cut the information processing cost of managers' decision-making, improve the quality of information on which decisions are based, and reduce wrong decisions caused by low-quality information (Poston and Grabski, 2001).

In addition, the decision-making rights theory of Jensen and Meckling (1995) pointed out that human cognitive limitations lead to information transmission and reception costs. The costs of transmitting this information to top managers are high, and decision-making empowerment at the bottom creates agency problems, so companies need to compromise between the two. In enterprises with many organizational levels or departments, the accounting system can also reduce the information transmission cost, thereby facilitating information transmission to top managers and reducing the internal agency cost.

As the quality of information on which enterprise decisions are based improves and the agency cost within the enterprise decreases, enterprise performance will also improve accordingly. TFP is an important measure of enterprise performance, which is a comprehensive reflection of enterprise efficiency (Schoar, 2002) and the basis for enterprises to achieve long-term future profitability. Therefore, we expect that the application of accounting software will bring higher enterprise TFP. The first hypothesis of this paper is as follows:

H1: When other conditions are the same, compared with enterprises that do not use accounting software, enterprises that use accounting software have higher enterprise productivity.

As mentioned earlier, the functions of different types of accounting software vary enormously. Basic accounting software can provide only financial information, while ERP also integrates information in various fields such as supply chains, production, sales, and human resources with accounting information, enabling managers to use both accounting information and non-accounting information for decision-making. This significantly enhances decision-making efficiency (Spathis, 2006) and even leads enterprises to advanced management accounting methods (Booth et al., 2000).

In addition, the difference between high-end ERP and low-end ERP is also significant, which is not only reflected in the number of modules, but also in the number of functions available within each module and the quality of implementing specific functions. ERP with more complete functions and higher efficiency can provide better support for management decision-making, thus playing a more significant role in promoting business efficiency (Ranganathan and Brown, 2006). Therefore, the second hypothesis of this paper is as follows:

H2: When other conditions are the same, enterprises that use more advanced accounting software have higher enterprise productivity.

However, there are also costs associated with the use of enterprise accounting software, especially ERP, and it is unclear whether managers of unlisted companies have the motivation and ability to take advantage of the higher-quality information that accounting software can provide. Therefore, the effect of accounting software on enterprise productivity still needs to be verified by empirical tests.

② For example, the generalized inventory information ignores the differences between assets such as raw materials, products in process, and finished products, which are quite different in nature.

Data in this paper are mainly the non-public data of the Annual Corporate Income Tax Return (main table and attached table) of enterprises above the designated size in a province of China from 2011 to 2015③. The key research variables in this paper, "whether the enterprise adopts accounting software" (Has_software) and "type of accounting software" (Software_type), are determined according to the "Accounting Software" field in the enterprise basic information table④. Specifically, if an enterprise fills in this field with "none", "no", "never", "not used", "manual", or words with similar meanings, it means that the enterprise does not use any accounting software. In this case, Has_software is set to 0; otherwise, it is set to 1. Furthermore, we divided the various accounting software used by enterprises into five grades from high to low according to their functions: ERP of top international manufacturers (such as Oracle and SAP) has the highest level, and Software_type is set to 5; ERP made by Chinese manufacturers has the second highest level, and if the word "ERP" is included in the field of accounting software or the manufacturer/model information is enough to determine that it is a Chinese ERP (such as UFIDA NC, Kingdee K/3, etc.), Software_type is set to 4; Considering the leading position of UFIDA and Kingdee in the accounting software market in China, non-ERP accounting software of UFIDA/Kingdee has the third highest level, and Software_type is set to 3; Non-ERP accounting software of other well-known Chinese accounting software manufacturers (such as Suda and Aerospace Information) has the fourth highest level, and Software_type is set to 2; The accounting software other than the above software is the basic financial software, and Software_type is set to 1. If accounting software is not used, Software_type is set to 0.

From the 212,737 observations from 2011 to 2015, this paper first eliminated samples with missing industry classifications, financial industries, samples with missing employee information or less than 5 employees, and samples with missing variables required to calculate enterprise TFP. The observations of the highest and lowest 0.1% of each variable of the production function were deleted to mitigate the impact of extreme values, and the industries with fewer than 20 observations annually were eliminated to obtain 147,429 observations for estimating TFP. As the accounting software field only exists in 2014-2015, this paper only retains the TFP figures for these two years. After further eliminating the samples of listed companies and the samples with missing data of the principal regression variables, the samples finally adopted in this paper are 58,359 observations of unlisted companies in the years 2014-2015.

The types of accounting software adopted by enterprises in the final samples are given in Table 1. Nearly half of the sample enterprises used UFIDA/Kingdee's non-ERP accounting software (Software_type = 3); only 6.41% of the enterprises have explicitly adopted ERP (Software_type ≥ 4); about 25% of the companies did not use accounting software.

Table 1 Accounting software types

*Software_type* Value (Software Type)	Number of Observations	Percentage
5 (ERP of international manufacturers)	308	0.53%
4 (ERP of Chinese manufacturers)	3,432	5.88%
3 (Non-ERP software of UFIDA/Kingdee)	28,648	49.09%
2 (Non-ERP software from other well-known manufacturers)	3,640	6.24%
1 (Basic financial software)	7,851	13.45%
0 (No accounting software used)	14,480	24.81%
Total	58,359	100.00%

I) Regression variables and models

In this paper, the logarithmic Cobb-Douglas production function is first estimated by OLS, and TFP is calculated as follows:

lnVA = a + b_klnK + b₁lnL + Fixed Effects + e 1)

where, VA is the added value of the enterprise calculated according to the income method, K is the fixed capital stock of the enterprise, and L is the labor input of the enterprise. The specific measurement method is shown in Table 2. Fixed Effects are annual and industrial fixed effects. e is the enterprise productivity TFP_ols obtained from the regression. Another measure of enterprise TFP TFP_lp is also obtained in this paper by the LP method (Levinsohn and Petrin, 2003).

The main regression model in this paper is set as follows:

TFP_t = a₀ + a₁Software_t + bX + Fixed Effects + e 2)

where, TFP_t is the dependent variable TFP_ols or TFP_lp estimated by model 1). Software_t is the key research variable Has_software or Software_type, whose specific definition is presented above. Fixed Effects are annual and industrial fixed effects.

X is a control variable. In this paper, the size of the enterprise (Size) is first controlled, because the size affects the resources that the enterprise can utilize and productivity. This paper adds the growth rate of enterprise sales revenue (Growth) to control the effect of market demand growth on productivity. This paper also controls the historical profitability of the enterprise, including the total return on assets of the previous year (ROA_t-1) and "whether there was a loss in the previous year" (Loss_t-1). External financing provides important support for enterprise development, so this paper controls the dummy variable (Finance) of "whether there is debt financing". Excessively high financing costs will increase the burden on enterprises, so this paper controls the range of net interest expenses (Finance_cost). Considering that companies making foreign investments may have more resources, this paper also controls the existence of long-term equity investment (Investment) and the occurrence of equity investment gains and losses (Invest_gain). Government subsidies are another important resource that enterprises may have, so this paper controls the proportion of government subsidies to the total assets of enterprises (Subsidy). A large number of studies have found that corporate R&D investment is closely related to productivity (Doraszelski and Jaumandreu, 2013), so this paper controls corporate R&D expenditure (R&D) and high-tech enterprise certification dummy variables (High_tech). The competition of enterprises in the international market can also promote the improvement of efficiency (Loecker, 2007). Considering that export sales will cause foreign currency exchange for enterprises, this paper controls the dummy variable of "whether there is exchange gain or loss" (Exchange). Finally, the level of urban economic development and financial development also affects productivity, so this paper also controls the per capita GDP (City_GDP) and the amount of local loans of financial institutions (City_loan) in the prefecture-level city where the enterprise is located. Table 2 lists the specific definition and calculation method of each variable.

Table 2 Definitions of variables

Variable	Definition
Has_software	Dummy variable. If an enterprise uses any accounting software (including ERP), it is 1; otherwise, it is 0.
Software_type	The value ranges from 0 to 5 according to the type of accounting software adopted by an enterprise. The more advanced the software, the higher the value (see above for the assignment rules).
TFP_ols	The total factor productivity of an enterprise estimated by OLS. It is the residual obtained by estimating the following equation: lnVA_t = a + b_klnK + b₁lnL + Fixed Effects + e, where Fixed Effects are annual and industrial fixed effects. Based on the industry classification of the national economy, this paper adopts the second-level classification for the manufacturing industry (first-level for other industries) and merges individual industries with a strong correlation.
TFP_lp	The total factor productivity of an enterprise estimated by LP. The dependent variable of the semiparametric regression is lnVA, the state variable is lnK, the free variable is lnL, and the proxy variable is lnM. The annual and industrial fixed effects are controlled during calculation.
VA	The added value of an enterprise under the income method. It is calculated by the current depreciation and amortization + business taxes and surcharges + employee compensation paid + operating profit + interest net expenditure and flattened by the ex-factory price index of the province's products of the current year.
K	The fixed capital stock of an enterprise. It is calculated by the historical costs of the enterprise's fixed assets and intangible assets and flattened by the fixed asset investment price index of the province of the current year.
L	The labor input of an enterprise. It is calculated by the number of employees in the enterprise.
M	The intermediate product input of an enterprise. It is the difference between the total output value (replaced by sales revenue) and the added value and is flattened by the provincial ex-factory price index of the current year.
Size	The natural logarithm of the average total assets of an enterprise.
Growth	The business revenue growth rate of an enterprise.
ROA_t-1	The average of total pre-tax profit/total assets of an enterprise in the previous year.
Loss_t-1	Dummy variable reflecting whether the total pre-tax profit of an enterprise in the previous year is negative. For a negative profit, it is 1; otherwise, it is 0.
Finance	Dummy variable reflecting whether an enterprise has a net interest expenditure. If there is such an expenditure, it is 1; otherwise, it is 0.
Finance_cost	The ratio of the enterprise net interest expenditure to the operating income.
Investment	Dummy variable reflecting whether an enterprise has a long-term equity investment at the end of 2013. If there is such an investment, it is 1; otherwise, it is 0.
Invest_gain	Dummy variable reflecting whether an enterprise has equity investment income in the current year. If there is such an income, it is 1; otherwise, it is 0.
Subsidy	Government subsidy income/total assets of an enterprise.
R＆D	Sum of R&D expenditure that can be added and deducted in the past three years/operating income of the previous year.
High_ tech	Dummy variable reflecting whether an enterprise has a high-tech enterprise certificate in the current year. If there is such a certificate, it is 1; otherwise, it is 0.
Exchange	Dummy variable reflecting whether an enterprise has exchange gains and losses. If there is such a gain or loss, it is 1; otherwise, it is 0.
City_ GDP	The natural logarithm of the per capita GDP (yuan) of the prefecture-level city where the enterprise is located.
City_ loan	The natural logarithm of the year-end balance of local loans from financial institutions (100 million yuan) of the prefecture-level city where the enterprise is located.

II) Main variable description statistics and correlation coefficients

The descriptive statistics of the main variables in this paper are shown in Table 3. All continuous variables were Winsorized at the 1% and 99% quantiles.

In terms of the Pearson correlation coefficient of the main variables, the correlation coefficients between Has_software (Software_type) and TFP_ols and TFP_lp are 0.112 and 0.153 (0.144 and 0.200), respectively, and both are significantly positive at the 1% level, indicating that there is a significant positive correlation between accounting software adoption and enterprise productivity.

Table 3 Descriptive statistics of main variables (N = 58,359)

Variable	Mean	Standard Deviation	Minimum	p25	Median	p75	Maximum
Total assets (million)	115.073	349.187	0.841	7.998	19.882	63.000	2,716.85
Operating income (million)	72.810	166.057	4.194	9.975	20.391	53.993	1,175.69
Number of employees	100.128	167.752	5	20	45	100	1,137
Software_type	1.994	1.379	0	1	3	3	5
Has_software	0.752	0.432	0	1	1	1	1
Location_len	5.674	4.417	2	3	3	7	21
TFP_ols	0.003	0.730	-2.221	-0.366	-0.022	0.351	2.347
TFP_lp	10.84	0.776	8.578	10.398	10.808	11.265	13.101
Size	17.005	1.606	13.642	15.895	16.805	17.959	21.723
Growth	0.142	0.737	-0.738	-0.166	-0.001	0.196	5.045
ROA_t-1	0.032	0.093	-0.291	0.001	0.015	0.046	0.511
Loss_t-1	0.213	0.410	0	0	0	0	1
Finance	0.492	0.500	0	0	0	1	1
Finance_cost	0.015	0.034	0.000	0.000	0.000	0.014	0.231
Investment	0.098	0.298	0	0	0	0	1
Invest_gain	0.035	0.183	0	0	0	0	1
Subsidy	0.003	0.011	0.000	0.000	0.000	0.000	0.078
High_tech	0.049	0.217	0	0	0	0	1
R&D	0.007	0.029	0.000	0.000	0.000	0.000	0.187
Exchange	0.110	0.313	0	0	0	0	1

III) Main regression results

Table 4 lists the results of the main regression in this paper. The results in columns (1) and (2) show that regardless of TFP_ols or TFP_lp as the dependent variable, the coefficient of the key research variable Has_software in this paper is significantly positive at the 1% level, indicating that the adoption of accounting software has a significant positive effect on enterprise productivity. In columns (3) and (4), the coefficients of Software_type are also significantly positive at the 1% level, indicating that the level of accounting software adopted by an enterprise also has a significant positive effect on enterprise productivity. In order to further compare the effects of different kinds of accounting software on productivity, the values of Software_type are added for regression as dummy variables. From the results in columns (5) and (6), we found that not only the coefficients of each dummy variable are significantly positive at the 1% level, but also the coefficient difference tests for adjacent levels of software show almost all significant results, indicating that a higher level of accounting software has a significantly higher impact on productivity than low-end software⑤.

The main result is also economically significant. According to TFP obtained under the OLS (LP) method, the average TFP of enterprises using accounting software is 0.071 (0.081) higher than that of enterprises not using accounting software, which is approximately equivalent to 10% of the standard deviation of the enterprise TFP; The coefficient of the dummy variable "Software_type = 3" shows that the TFP of enterprises adopting UFIDA/Kingdee’s non-ERP accounting software is 0.076 (0.090) higher than that of enterprises not using accounting software, which is approximately equivalent to 10% (12%) of the standard deviation of the enterprise TFP; The coefficient of the dummy variable "Software_type = 4" shows that the average TFP of enterprises adopting ERP is 0.166 (0.183) higher than that of enterprises not adopting accounting software, which is about 23% of the standard deviation of enterprise TFP⑥; Enterprises adopting basic accounting software even have average TFP that is 0.037 higher than those not using any accounting software, which is approximately equivalent to 5% of the standard deviation of the enterprise TFP (or 1/5 of the ERP effect).

Table 4 Main regression results

		(1)	(2)	(3)	(4)	(5)	(6)
Variable	Expected symbol	*TFP_ols*	*TFP_lp*	*TFP_ols*	*TFP_lp*	*TFP_ols*	*TFP_lp*
Has_software	+	0.071***	0.081***
		[10.371]	[12.550]
Software_type	+			0.030***	0.036***
				[13.047]	[16.314]
Software_type = 5	+					0.388***	0.492***
						[7.827]	[11.311]
						(p < 0.001)	(p < 0.001)
Software_type = 4	+					0.166***	0.183***
						[10.391]	[12.399]
						(p < 0.001)	(p < 0.001)
Software_type = 3	+					0.076***	0.090***
						[10.204]	[12.809]
						(p < 0.001)	(p < 0.001)
Software_type = 2	+					0. 043***	0. 045***
						[3.438]	[3.813]
						(p = 0.669)	(p = 0.557)
Software_type = 1	+					0.037***	0.037***
						[3.823]	[4.057]
Size	+	0.140***	0.244***	0.137***	0. 240***	0.136***	0.239***
		[46.796]	[84.720]	[45.609]	[83.014]	[45.140]	[82.339]
Growth	+	0.161***	0.143***	0.160***	0.143***	0.160***	0.143***
		[29.287]	[27.288]	[29.241]	[27.230]	[29.216]	[27.195]
ROA_t-1	+	1.876***	1.927***	1.868***	1.918***	1.857***	1.905***
		[40.237]	[43.166]	[40.107]	[43.046]	[39.890]	[42.776]
Loss_t-1	-	-0.262***	-0.234***	-0.264***	-0.236***	-0.264***	-0.237***
		[-29.017]	[-27.046]	[-29.236]	[-27.332]	[-29.298]	[-27.413]
Finance	+	-0.002	0.051***	-0.001	0. 052***	-0.000	0.053***
		[-0.321]	[7.397]	[-0.202]	[7.548]	[-0.027]	[7.784]
Finance_cost	-	-0.440***	-1.219***	-0.443***	-1.221***	-0.430***	-1.205***
		[-2.969]	[-9.123]	[-2.988]	[-9.147]	[-2.902]	[-9.033]
Investment	+	0.026**	0.069***	0.025^*	0.067***	0.025^*	0.067***
		[2.015]	[5.745]	[1.918]	[5.605]	[1.927]	[5.631]
Invest_gain	+	0.187***	0.166***	0.185***	0.164***	0.183***	0.163***
		[8.629]	[8.776]	[8.538]	[8.669]	[8.463]	[8.598]
Subsidy	?	-3.312***	-1.748***	-3.379***	-1.829***	-3.394***	-1.843***
		[-10.101]	[-5.461]	[-10.315]	[-5.725]	[-10.373]	[-5.779]
High_tech	+	0.062***	0.127***	0.059***	0.122***	0.056***	0.118***
		[3.401]	[7.166]	[3.198]	[6.922]	[3.035]	[6.694]
R&D	+	0.609***	0.774***	0.590***	0.750***	0.610***	0.779***
		[4.440]	[5.881]	[4.311]	[5.719]	[4.448]	[5.924]
Exchange	+	0.065***	0.100***	0.063***	0.097***	0.062***	0.096***
		[7.144]	[11.658]	[6.944]	[11.397]	[6.842]	[11.289]
City_GDP	+	0.049**	0.076***	0.056**	0.084***	0.062***	0.090***
		[2.155]	[3.533]	[2.442]	[3.896]	[2.733]	[4.209]
City_loan	+	0.049***	0.029***	0. 046***	0.025***	0.045***	0.024***
		[6.275]	[3.944]	[5.884]	[3.439]	[5.765]	[3.329]
Annual and industrial fixed effects		Yes	Yes	Yes	Yes	Yes	Yes
N		58,359	57,858	58,359	57,858	58,359	57,858
Adj. R²		0.245	0.408	0.247	0.409	0. 248	0. 410

Note: (1) Continuous variables have been Winsorized at the 1% and 99% levels. ***, **, and * represent two-tailed test results p < 0.01, p < 0.05, and p < 0.1, respectively;

(2) The t value in square brackets is based on the robust standard error of the company-level cluster;

(3) The p value in the parentheses between the coefficients of every two rows of Software_type in columns (5) and (6) is an F test for the difference between the regression coefficients of two adjacent types of software.

IV) Robustness test

Several robustness tests have been carried out in this paper. First, in order to further control the scale effect of enterprises, we additionally controlled the natural logarithm of the sales revenue of the previous period. Second, this paper used the propensity score matching method to make sample enterprises using different accounting software more comparable. Third, in order to eliminate the impact of earnings management behavior on productivity measurement, we additionally controlled the proxy variable that reflects the possibility of corporate accounting manipulation, the corporate book-tax difference or the deviation between the distribution of the first digit of accounting data and "Benford's Law"⑦. Finally, we only took manufacturing enterprises as samples for analysis. The unreported results showed that the conclusions of this paper still hold in the above robustness analysis.

③ The data obtained in this paper has been desensitized, and the enterprise name, shareholders, and other identity information have been hidden.

④ This field is based on Article 20 of the Law of the People's Republic of China on the Administration of Tax Collection, which states that "The financial and accounting systems or financial and accounting treatment methods and accounting software of taxpayers engaged in production and business operations shall be submitted to the tax authorities for filing". Therefore, its disclosure is mandatory and can avoid the noise and endogeneity problems of voluntary disclosure.

⑤ The coefficient difference between the dummy variables "Software_type = 1" and "Software_type = 2" is not significant, which may indicate that non-ERP software of the manufacturers other than UFIDA/Kingdee has little difference from basic accounting software.

⑥ This effect is also equivalent to directly increasing TFP from the 50th percentile level to the 62nd percentile level, or from the 25th percentile level to the 36th percentile level.

⑦ The corporate book-tax difference refers to the difference between accounting profit and taxable income. The larger the book-tax difference is, the greater the possibility of corporate accounting manipulation is (Hanlon, 2005). The probability that the first significant digit of a naturally occurring variable is d (a natural number with a value of 1-9) is close to log₁₀(1 + 1/d). This law of the distribution with a small first digit is called Benford's Law. Amiram et al. (2015) found that the greater the deviation between the first digit distribution of each subject in a corporate's financial report and Benford's Law, the more likely the corporate's financial report is wrong or manipulated.

I) Endogenous processing

The costs of adopting accounting software for an enterprise

include software purchase price, service fee, user training fee, and other adaptation costs. Enterprises with abundant resources can easily bear these costs, so they have a greater probability to use accounting software. At the same time, these enterprises may also have higher productivity, which will bring about endogeneity problems. Although the variables related to enterprise resource abundance have been controlled as much as possible in this paper, some missing variables may still cause coefficient estimation bias in the main regression of this paper. Therefore, we adopt two methods to further deal with endogeneity problems:

1. Granger causality test. Through regression comparison, this paper finds that after controlling the TFP of the previous period, the impact of accounting software adoption in the previous period on the TFP of the current period is still significantly positive at the 1% level; However, after controlling the adoption of accounting software in the previous period, the impact of the TFP of the previous period on the adoption of accounting software in the current period is not significant. This indicates that the adoption of accounting software affects the level of TFP, rather than the level of TFP affecting the adoption of accounting software.

2. Instrumental variable regression. In this paper, the degree of detail of the "file storage location" entered by an enterprise in the basic information table reflects the importance that the enterprise attaches to accounting information management, and this is used as an instrumental variable to perform 2SLS regression. "File storage location" refers to the specific location where an enterprise keeps its accounting files in accordance with laws and regulations. The more detailed an enterprise enters this information (the more it attaches importance to accounting information management), the more likely it is to seek ways to improve the level of accounting information management (such as using accounting software). At the same time, whether the information is filled out in detail is not necessarily related to the amount of enterprise resources, exhibiting its exogeneity. This paper uses two instrumental variables simultaneously, one is the text length of the file storage description field (Location_len), and the other is a dummy variable indicating simplicity (Simple). If the field only indicates that the location is in the company (without any further description), it is 1; otherwise, it is 0. In the first-stage regression, the coefficient of the instrumental variable Location_len (Simple) was significantly positive (negative) at the 1% level, and the positive (negative) sign was consistent with expectations; the weak instrumental variable test showed that the first-stage F value and Cragg-Donald Wald F statistics were both much higher than the Stock Yogo critical value. In the second-stage regression, the regression coefficient of the fitting values of the accounting software was still significantly positive at the 1% level. In the over-identification test, the p-values of the Sargan test were not significant at the normal level. The results of the above endogenous treatment further supported the conclusion of this paper.

II) Analysis of regulating variables

Accounting software can improve the information environment for managers to make decisions, but to what extent it can improve enterprise productivity depends on the following two factors: (1) the ability of enterprises to use accounting information; (2) the business growth opportunities for enterprises.

First, although information systems allow managers to obtain useful information for decision-making more conveniently, whether enterprises can gain a competitive advantage by virtue of them is also affected by the information utilization behavior of enterprises (Bendoly et al., 2009). The better an enterprise makes use of accounting information, the greater the role of accounting software in improving productivity. This paper uses the method adopted by an enterprise to account for bad debts of accounts receivable to reflect the enterprise's ability to use accounting information⑧. In the group test results, the accounting software regression coefficients of the enterprise group using the allowance method are higher than those of the enterprise group using the direct write-off method, and the differences are mostly significant at the 5% level or better. This is consistent with our expectations, indicating that the function of management accounting not only requires hardware resources such as accounting information systems, but also depends on the managerial skills of enterprise managers.

Second, when the market space of the enterprise is large and the business growth opportunities are great, the enterprise's decision-making quality enhancement is easier to translate into the improvement of productivity; If the whole industry is in a recession or in a mature period, the role of decision-making quality in improving productivity is limited. In this paper, the average sales revenue growth rate of enterprises in the same industry, year, and city represents business growth opportunities. In the group test results, the accounting software regression coefficient of the high-growth opportunity enterprise group is significantly higher than that of the low-growth opportunity enterprise group at the 1% level, indicating that the higher the business growth opportunity of the enterprise, the greater the effect of accounting software on productivity improvement.

⑧ The allowance method can reflect the true financial situation of an enterprise in a more timely manner to avoid information delay caused by confirming losses when the final accounts cannot be recovered. Therefore, an enterprise adopting the allowance method (direct write-off method) may have a higher (lower) ability to use accounting information. In this paper, 73.4% of enterprises are still using the direct write-off met

Based on the observations of 58,359 unlisted enterprises in a Chinese province in the years 2014 and 2015, this paper empirically tested the impact of unlisted enterprises' accounting software adoption on their productivity for the first time. We found that in unlisted enterprises, the adoption of accounting software has a significant positive effect on enterprise productivity, and the more advanced accounting software (produced by more famous manufacturers with more complete functions) has a more significant effect; the above effect is also modulated by the ability of enterprises to use accounting information and business growth opportunities. This paper uses methods such as propensity score matching, Granger causality tests, and instrumental variable regression to alleviate endogeneity problems, which proves that the conclusions still hold. The conclusions of this paper show that accounting software can improve the quality of decision-making by improving the information environment for managers to make decisions, thereby improving enterprise productivity. Under the background of the current emphasis on high-quality development and TFP improvement, the findings of this paper have important policy reference value.

Runping L, Xue Shengjia (2011) ERP and Corporate Profit Performance: Empirical Evidence from Shanghai and Shenzhen Listed Companies. Syst Eng Theory Pract 8:1460–
Rao Yanchao (2005) An Empirical Analysis of the Impact of ERP System Implementation on Enterprise Costs: Empirical Evidence from Manufacturing Listed Companies in Shanghai and Shenzhen. Financial Res 4:133–
Wang Liyan ZJ (2007) The Relationship Between the Implementation of ERP System and the Corporate Performance Growth: An Empirical Analysis Based on the Data of Listed Companies in China. Manage World 3:116–
Jianguang Z, Liyan W, Xu Haile (2012) ERP System Implementation and Agency Costs: Based on the Evidence of China's ERP Introduction Period. Nankai Manage Rev 3:131–
Zhang Ruijun S, Yuefan SB (2008) Research on Key Information Disclosure of ERP Investment in Chinese Enterprises. Acc Res 2:55–
Zhang Zhe H, Pei ZL (2005) Analysis of Key Success Factors of ERP Implementation in Chinese Enterprises: Multiple Case Studies. Manage World 12:137–
Zhao Quanwu H, Zhizhong BX (2008) An Empirical Study on the Performance Changes of Listed Companies Before and After ERP Implementation: From the Empirical Data of Shanghai Stock Market from 1993 to 2003. J Manage Sci 1:122–

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Application of Accounting Software and Enterprise Productivity——Evidence from Unlisted Companies

Status:

Version 1

Abstract

I. Introduction

Ii. Literature Review And Hypotheses

Iii. Data Source And Sample Description

Iv) Robustness Test

V. Endogenic Processing And Analysis Of Regulating Variables

Vi. Conclusions

References

Additional Declarations

Status:

Version 1